Primary neural stem/progenitor cells expressing endostatin or cytochrome P450 for gene therapy of glioblastoma

Abstract

Despite recent technical improvements in surgical excision techniques and adjuvant radio- and chemotherapy, the clinical outcome of patients with grade IV astrocytoma (glioblastoma) remains very poor, with a median survival of less than 12 months. A promising approach to therapy employs gene-engineered neural stem/progenitor cells (NSCs) as a cellular therapeutic delivery system, to track glioblastoma cells and deliver anticancer molecules. However, most results on their tumor tropism have been derived by immortalized NSCs. We now report that primary murine gene-engineered NSCs displayed in vivo tropism for glioblastoma cells. Ten days after injection into the brain, many NSCs continued to express the transgene and the NSC marker, nestin. NSCs transduced with a retroviral vector co-expressing a secretable form of human endostatin and eGFP (NSC/endo-eGFP) released potentially antiangiogenetic concentrations of endostatin into the culture medium. Conditioned medium of NSC/endo-eGFP cells inhibited the growth of mouse pulmonary microvascular endothelial cells (PMVECs). A good correlation between endostatin levels and PMVEC growth inhibition was observed. In NSCs co-expressing cytochrome P450 2B6 (CYP2B6) and eGFP (NSC/CYP2B6-eGFP), the forced expression of CYP2B6 resulted in intracellular activation of CPA and subsequent cell death. In the presence of NSC/CYP2B6-eGFP, we observed CPA cytotoxicity to DsRed-expressing U87Mg glioma cells. In vivo treatment of intracranial GL-261 glioblastoma with NSC/endo-eGFP caused a 65% reduction in tumor size, compared to untreated control mice or mice treated with NSC/eGFP cells. Our data suggest that primary NSCs transduced with retroviral vectors expressing endostatin and/or CYP2B6 have a potential role in glioblastoma therapy.

Introduction

Gliomas, and in particular grade IV astrocytoma (glioblastoma), are associated with very poor prognosis and represent one of the greatest therapeutic challenges in oncology.¹ The extensive infiltration of malignant glioblastoma cells into adjacent normal neural tissue is at least partially responsible for the recurrent tumor growth and the frequent failure of conventional therapy. Gene therapy strategies, employing neural stem/progenitor cells (NSCs) armed with anticancer molecules to inhibit the growth and/or cause regression of glioblastoma, have recently been proposed. The advantage of employing gene-engineered NSCs as vehicle of anticancer agents is that NSCs have been shown to track in vivo glioma cells as well as other types of cancer cells.² The exceptional migratory ability of NSCs^3,4 is a property that has made them so useful in therapeutic paradigms demanding gene and cell replacement in various animal models of neurodegeneration, albeit usually in the newborn.⁵

In the past few years, NSCs have been used to deliver therapeutic genes and/or their products to tumor cells in the adult central nervous system (CNS). Genetically engineered NSCs expressing interleukin-4 elicited the systemic immune response and inhibited tumor growth.⁶ Aboody et al.³ showed that NSCs can deliver a therapeutically relevant molecule, cytosine deaminase, which causes tumor regression. When implanted intracranially at distant sites from the tumor (for example, into the controlateral hemisphere), NSCs migrated through normal tissue targeting cancer cells.^2,3 Recently, Danks et al.⁷ have reported that intravenous injection of HB1.F3.C1, an immortalized NSC cell line, engineered to deliver the cDNA encoding a secreted form of the CPT-11-activating enzyme carboxylesterase, inhibited neuroblastoma tumor growth in mice treated with systemic CPT-11.

However, most results on gene-engineered NSCs and on their tumor tropism have been derived by immortalized NSCs. Recent data question the reliability of immortalized NSC cell lines,⁸ thus favoring the use of primary NSCs as vehicles of anticancer agents. Primary NSCs are generally grown as floating cultures, to form heterogeneous structures comprised of undifferentiated cells and their progenies, called neurospheres. We have recently developed a highly efficient cell culture protocol⁹ based on adherent culture of NSCs from neonatal forebrains on recombinant fibronectin (rFN). rFN facilitates retroviral infection by co-localization of hematopoietic cells and viral particles on the same rFN molecule.¹⁰ Under protocol conditions, NSCs grow faster on rFN matrix than as neurospheres and do not lose their stem cell nature or multipotentiality; our protocol allows to transduce up to 90% NSCs at a multiplicity of infection (MOI) of 2 with no need for viral concentration or production of serum-free retroviral supernatants.⁹ Our studies have been the first to report gene transfer by an FMEV-derived backbone in NSCs: the FMEV retroviral vectors were optimized for efficient transgene expression in hematopoietic progenitor cells, and we found them similarly well suited for NSCs and derived cell types.⁹ In the present study, we have employed this retroviral-based transduction protocol to arm NSCs with the antiangiogenic molecule, endostatin or with the human prodrug-activating cytochrome P450 enzyme 2B6.

The expansive growth of malignant cells, including brain cancer cells, depends on the continuous outgrowth of new blood vessels, a process known as angiogenesis. Inhibition of angiogenesis has shown efficacy in several experimental tumor models,^11-14 thus validating angiogenesis as an important potential therapeutic target. Several antiangiogenic molecules have been discovered or synthesized. Among these are endostatin, a proteolytic terminal fragment of collagen XVIII generated by elastase, cathepsin or matrix metalloproteinases. In addition to being a potent inhibitor of tumor angiogenesis, endostatin inhibits many functions of endothelial cells, including proliferation and migration.^13,14 Endostatin also has direct effects on tumor cells.^15,16 A negative correlation between levels of endostatin and malignancy grade in human gliomas was observed,¹⁷ presumably reflecting the shift of a regulatory balance between promoters and inhibitors of angiogenesis toward increased neovascularization. These findings have resulted in testing of endostatin in clinical trials.¹⁸ However, several problems, such as the short biological half-life of endostatin, variability of endostatin preparations and the impaired intratumoral blood flow, have so far hindered its therapeutic success.¹⁹

CYP 2B6 metabolizes a number of drug substrates, that are usually nonplanar, neutral or weakly basic, fairly lipophilic with one or two hydrogen bond acceptors, on which it catalyzes various oxidative reactions. For cyclophosphamide and ifosfamide, these reactions represent major metabolic or activation pathways, and for their kinetics CYP2B6 function is of considerable importance.²⁰ The well-described bystander effect^21-23 provides a mechanism for extending the cytotoxic response to the surrounding tumor cells. In this study, we provide evidence of a potential anti-glioblastoma activity of NSCs transduced with retroviral vectors expressing endostatin and/or CYP2B6.

Materials and methods

Cell culture

Human U87Mg glioblastoma were cultured in RPMI-1640; murine 3T3 and GL-261 cells were cultured in Dulbecco’s modified Eagle’s medium. All media were added with 10% fetal bovine serum and 2 mm L-glutamine.

NSC preparation

NSCs were prepared from 1- to 3-day-old C57BL/6 mice as described by Machon et al.,²⁴ and cultured in serum-free Neurobasal-A medium containing B27, 2 mm L-glutamine, 10 ng ml⁻¹ basic fibroblast growth factor and 20 ng ml⁻¹ epidermal growth factor. Low passage (passage 2–5) NSCs were used throughout all the experiments. Untreated tissue culture 24-well plates were coated with 10 μg cm⁻² rFN (Retronectin; Takara Shuzo Co., Kyoto, Japan), according to the manufacturer’s instructions. For culture on rFN, neurospheres were incubated in the presence of 0.05% trypsin with 0.02% EDTA for 7 min at 37 °C. The single-cell suspensions were then pelleted by centrifugation at 400 g for 3 min, resuspended in complete medium, filtered through a 70-μm filter to remove eventual remaining cell clumps and added to rFN-coated plates. The cells were then incubated at 37 °C in an atmosphere of 5% CO₂ in air. Under protocol conditions, NSCs grow faster on rFN matrix than as neurospheres and do not lose their stem cell nature or multipotentiality. By this method, we consistently transduced 60–90% NSCs at an MOI of 1 with no need for viral concentration or production of serum-free retroviral supernatants.⁹

Drug sensitivity assay

The drug sensitivity patterns were determined using the (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-phenazine methosulfate (MTS/PMS) microtiter plate assay, as previously described.²⁵

Retroviral vectors

All vectors we have used are based on pSF91 (GenBank accession no. AJ224005) with the 3′-LTR of spleen focus-forming virus and the leader of the murine embryonic stem cell virus, which is based on dl587rev,²⁶ a backbone that we and others have previously shown to induce efficient expression of the transgene(s) in early hematopoietic and neural cells.^9,27-29 The human endostatin XVIII cDNA was removed from pBLAST-hENDOXVIII (Invivogen, San Diego, CA) by digestion with SgrAI and EheI; cohesive ends were filled with Klenow and moved into pBSL99. The γ-GCSh cDNA was removed from pSF91/GCSh-eGFP³⁰ by NotI digestion and replaced with the endostatin cDNA, which was removed from pBSL99-endostatin with NotI, thus obtaining pSF91/endostatin-eGFP. The CYP2B6 cDNA, a kind gift from Dr Gonzalez, was PCR amplified from the plasmid pBabe-2B6-puro³¹ with primers containing NotI sites and cloned in pBluescriptII SK⁺. The sequence of the cloned product was verified by double-stranded DNA sequencing and a 1485-bp NotI fragment was used to replace γ-GCSh of pSF91/GCSh-eGFP, thus obtaining pSF91/CYP2B6-eGFP.

Retroviral transduction

To generate retroviral producers, the Phoenix-gp packaging cell line was transfected by the calcium phosphate/chloroquine method³² with pSF91/DsRed, pSF91/eGFP, pSF91/endostatin-eGFP or pSF91/CYP2B6-eGFP and a plasmid expressing the ecotropic glycoprotein, or a plasmid expressing the vesicular stomatitis virus (VSV) glycoprotein. Culture supernatants containing viral particles were collected at 24–48 h after transfection, passed through 0.22 μm Millex GP filters (Millipore Co., Bedford, MA, USA) and stored at −80 °C. For transduction of NSCs, retroviral supernatants were preloaded onto rFN-coated plates and centrifuged at 950 g for 30 min at 4 °C, thus largely avoiding serum contamination. The operation was repeated a second time with fresh supernatant. The supernatant was then removed and the plates washed with phosphate-buffered saline (PBS) before addition of NSCs. We then amplified the three cultures and stored in liquid nitrogen cell aliquots for all the required experiments. For transduction of U87Mg cells with SF91/DsRed, cells were centrifuged at 850 g for 60 min at room temperature in the presence of VSV-pseudotyped retroviral vector particles. After 24 h, the retroviral vector-containing medium was replaced with virus-free fresh medium. A few days later, cells were cloned by limiting dilution. A clone displaying bright fluorescence and a doubling time identical to that of the respective parental cells was employed for all the required experiments.

Immunocytochemistry

For immunocytochemistry, cells were plated onto poly-L-lysine-coated chamber slides, fixed in 4% paraformaldehyde, washed with PBS, permeabilized in 0.2% NP-40 and blocked with goat serum. The primary antibodies were obtained from Chemicon (Hofheim, Germany).

In vivo migration studies

All procedures involving animals were approved by and performed according to guidelines of the Institutional Animal Care and Use Committee of the University of South Alabama. Athymic nude mice (nu/nu; 6- to 8-week old) were maintained in microisolator cages. Briefly, the mice were anesthetized by i.m. injection of a mixture of ketamine and xylazine, and placed in a small animal stereotactic frame (KOPF, Tujunga, CA). Here the mice received stereotactically guided injections into the forebrain. A sagittal incision was made through the scalp to expose the skull and a small burr hole was made into the right forebrain at the following coordinates relative to bregma: 1 mm anterior, 2 mm lateral of the midline and at a depth of 2.5 mm from the skull surface. In total 50 000 cells, suspended in 1.5 μl of PBS, were injected with a 26 g Hamilton syringe over 180 s. The needle was left in place for 1 min and then withdrawn slowly. The incision was closed with single use skin staples (Appose ULC, Tyco, CT). We used a light-emitting diode flashlight (LDP LLC, Woodcliff Lake, NJ) with an excitation filter (midpoint wavelength peak of 525 nm) and an emission 570 long-pass filter (both filters from Chroma Technology, Rockingham, VT) to follow the intracranial growth of DsRed-expressing U87Mg glioma by in vivo fluorescence imaging. The images were captured with a Nikon CoolPix P2 digital camera. A software program was used to calculate the number and intensity of pixels in each image and then convert the numbers to mm² for each tumor, as described by Yang et al.³³ The fluorescence intensity is an average value of all detectable DsRed signals.

Inoculation of intracranial GL-261 gliomas with NSC/endo-eGFP

C57Bl/6 female mice (6-weeks old) were anesthetized i.p. with ketamine and xylazine and stereotactically inoculated at the coordinates described above with 2.5 × 10⁵ mouse GL-261 glioblastoma cells. At day 4 postimplantation, the animals were divided into three groups and treated using the same burr hole and stereotactic coordinates with a single intracerebral inoculum of 3 × 10⁶ NSC/endo-eGFP cells (3 μl; n = 8), 3 × 10⁶ NSC/eGFP cells (3 μl; n = 6) or saline (3 μl; n = 8). Eight days after NSC inoculations, mice were euthanized using CO₂ asphyxiation, and their brains were harvested and frozen immediately on dry ice. Brains were sectioned using a cryostat into 10-μm-thick slices that were mounted on slides and then stained with hematoxylin and eosin (H&E) as per standard protocol. Tumor size was determined by making serial H&E-stained coronal sections spaced 150 μm apart. The ratio between tumor volume and total brain volume was determined on the sections bearing tumor with the maximum visible diameter.

Statistical analysis

The statistical significance of differences between two groups was validated by the Wilcoxon test. A 5% level of probability was considered significant.

Results

Using a previously established protocol,⁹ we transduced with high efficiency (over 60%) primary cultures of murine NSCs with SF91/eGFP, SF91/endostatin-eGFP and SF91/CYP2B6-eGFP. An overview of the vectors used is shown in Figure 1. We used an MOI of not more than 2, to reduce the possibility of potentially dangerous insertional mutagenesis. The growth rate of the three mass cultures was identical to that of untransduced NSCs. We have previously characterized our NSC preparations from neonatal C57Bl/6 mouse forebrain as containing over 90% nestin⁺/GFAP⁻/β-tubulin⁻ cells.⁹ Also NSCs transduced with SF91/endostatin-eGFP maintained a high level (>90%) of expression of the progenitor cell marker, nestin (Figure 2).

Figure 1.

Schematic representation of retroviral vectors used in this study. All expression vectors had identical control elements. SF91/eGFP contains only the eGFP cDNA (0.8 kb), whereas SF91/endostatin-eGFP and SF91/CYP2B6-eGFP contain the cDNAs of a secretable form of endostatin (0.6 kb) and CYP2B6 (1.5 kb), respectively, and of eGFP linked via an IRES from the encephalo-myocarditis virus (EMCV).

Figure 2.

Neural stem/progenitor cells (NSCs)/endo-eGFP express nestin, an NSC marker. After transduction, single cells were incubated with a primary monoclonal antibody against nestin (left) or an isotype control antibody (right), followed by a phycoerythrin-labeled secondary antibody, and counterstained with DAPI (4-6-diamidino-2-phenylindole).

In order to demonstrate that primary NSCs stably transduced with an eGFP-expressing retroviral vector can survive in vivo upon injection into the brain and continue to express the transgene, NSC/eGFP were injected into the brains of 4-week-old athymic nu/nu mice. After 10 days, the animals were killed and brain sections were analyzed by fluorescence microscopy. In five out of six cases, eGFP⁺/nestin⁺ cells could be detected near the point of injection (Figure 3), suggesting that at least part of the injected NSCs maintain their NSC phenotype and stably express the transgene(s).

Figure 3.

Expression of nestin in neural stem/progenitor cells (NSCs)/eGFP cells 10 days after injection into the forebrain of an athymic nu/nu mouse. (Left) Brain section incubated with a primary monoclonal antibody against eGFP, followed by an fluorescein isothiocyanate (FITC)-labeled secondary antibody and counterstained with DAPI (4-6-diamidino-2-phenylindole). (Right) Brain section incubated with a primary monoclonal antibody against Nestin, followed by a phycoerythrin-labeled secondary antibody and counterstained with DAPI (4-6-diamidino-2-phenylindole).

To demonstrate that retrovirally transduced primary murine NSCs can track brain tumor cells, we first developed a fluorescent glioblastoma model by transduction of SF91/DsRed retroviral vector into human U87Mg glioblastoma, a cell line that when injected into the brains of athymic nude mice develop into invasive angiogenesisdependent tumors.³⁴ After transduction, we established several clones by limiting dilution. A clone displaying bright fluorescence and a doubling time identical to that of the respective parental cells was used throughout the experiments (Figure 4). These cells have been cultured for over 2 months and have shown a strong, stable intracellular fluorescence in 100% of the cells throughout the observation period. Figure 5 shows serial images of a DsRed-expressing U87Mg glioma, taken at 9, 12, 14 and 16 days from the orthotopic implant of 50 000 cells. Comparison of in vivo fluorescence images with frozen brain sections analyzed by fluorescence microscopy, using a large-field fluorescence objective (× 1.25), revealed a good correspondence between the central areas of fluorescence, clearly visible in the serial images of Figure 5, and the actual tumor size (not shown). Then, three 4-week-old athymic mice were orthotopically implanted with 50 000 U87Mg/DsRed cells at the following coordinates relative to bregma: 1 mm anterior, 2 mm lateral of the midline and at a depth of 2.5 mm from the skull surface. After 1 week, 50 000 NSC/eGFP were injected at 2 mm lateral distance from the site of the tumor inoculum. After 2 more weeks, the animals were killed and brain sections analyzed by fluorescence microscopy. NSC/eGFP were found to have migrated toward the tumor (Figure 5), whereas in one mouse only NSC/eGFP, but not U87Mg/DsRed glioma cells were found.

Figure 4.

Fluorescence micrographs of DsRed-transduced human U87Mg glioma and eGFP-transduced primary neural stem/progenitor cells (NSCs). Both DsRed-transduced U87Mg and GL-261 were single clones obtained by limiting dilution.

Figure 5.

In vivo fluorescence images. (Left) Serial fluorescent images of an athymic mouse orthotopically implanted with 50 000 DsRed-U87Mg cells. (Right), Representative fluorescence micrograph of a brain section from an athymic nude mouse implanted with U87Mg/DsRed cells on day 0, and neural stem/progenitor cells (NSCs)-eGFP at 2 mm. lateral distance from the site of tumor inoculum on day 7; the animal was killed on day 21.

We used an enzyme-linked immunosorbent assay (ELISA) assay (Accucyte, Cytimmune, Rockville, MD) to measure the amount of endostatin released from NSC/endo-eGFP cells into the cell culture medium. The total amount of endostatin 48–120 h after plating 50 000 NSC/endo-eGFP cells (70% eGFP⁺) in 3 ml medium ranged from 250 to 390 ng, whereas no endostatin was detectable in the culture medium of NSC/eGFP (85% eGFP⁺) (Figure 6). Thus, the production of endostatin is sustained with time in culture. We then investigated whether endostatin secreted by NSC/endo-eGFP cells inhibited the growth of mouse pulmonary microvascular endothelial cells (PMVECs). To this aim, PMVEC were isolated and cultured using a method described by King et al.³⁵ Cultures were characterized using uptake of DiI-acetylated low-density lipoprotein and a lectin-binding panel. The cells stained positive for the lectins Glycine max and Griffonia simplicifolia, as well as for eNOS, PECAM-1 and VE-cadherin (not shown). PMVEC were seeded in 96-well plates (700 cells per well) and cultured for 3 days in complete medium conditioned for 96 h by different numbers of NSC/endo-eGFP cells (endostatin concentrations were measured by an ELISA assay). Representative phase contrast micrographs of PMVEC exposed to conditioned medium from NSC/endo-eGFP and control NSC/eGFP cells are presented in Figure 7. Growth inhibition was measured by an MTS/PMS assay (Figure 8). A very good correlation between the endostatin concentration masured by ELISA and the antiproliferative effect, measured by the MTT assay, was observed (R² = 0.9803).

Figure 6.

Neural stem/progenitor cells (NSCs)/endo-eGFP release endostatin into the extracellular medium. In total 100 000 cells were plated in a well containing 3 ml medium. Conditioned medium was collected after various times, and passed through a 0.2 μm filter. Endostatin concentration was measured by an enzyme-linked immunosorbent assay (ELISA) assay, and expressed as ng ml⁻¹ per 100 000 cells. Columns represent the mean of three separate experiments (bars represent s.d.).

Figure 7.

Effect of endostatin released by neural stem/progenitor cells (NSCs)/endo-eGFP on mouse pulmonary microvascular endothelial cells (PMVECs). In total 100 000 NSC/endo-eGFP or NSC/eGFP cells (as control) were plated in a well containing 3 ml medium. Conditioned medium (CM) was collected after 48 h, passed through a 0.2 μm filter and added to PMVEC cultures for 72 h. (Left) PMVEC incubated with CM from NSC/eGFP control cells. (Right) PMVEC incubated with CM from NSC/endo-eGFP control cells.

Figure 8.

Correlation between concentrations of endostatin released by neural stem/progenitor cells (NSCs)/endo-eGFP and growth inhibition of pulmonary microvascular endothelial cells (PMVECs). NSC/endo-eGFP were plated at different numbers, and their conditioned medium (CM) was collected after 48 h. Endostatin concentrations were measured by an enzyme-linked immunosorbent assay (ELISA) assay. PMVEC were exposed to CM for 72 h. Growth inhibition was measured by an MTS/PMS assay, and expressed as percentage of mock-treated cells. Points are the means of quadruplicate experiments. The correlation coefficient R² was 0.9803 (bars represent s.d.).

To prove that the retroviral-mediated expression of CYP2B6 in cells cultured in vitro activates CPA and that activated CPA metabolites are then released into the culture medium, we carried out the following experiment: we transduced murine 3T3 fibroblasts with SF91/CYP2B6-eGFP or SF91/eGFP, obtaining a mass culture comprising >99% transduced cells, as measured by flow cytometric analysis of eGFP⁺ cells; parallel cultures of 3T3/CYP2B6-eGFP and 3T3/eGFP (as control) were exposed to 0, 3 and 10 mm CPA for 24 h. The conditioned medium was then aspirated and centrifuged at 900 g for 5 min to remove any floating cells. U87Mg/DsRed cells, plated 3 days before at 5000 per well in six-well plates were incubated with a 1:3 mixture of CPA-conditioned medium and fresh culture medium, resulting in a real CPA/4-OH-CPA concentration of 0, 1 and 3.33 mm, respectively. To evaluate the efficiency of tumor-cell destruction, DsRed fluorescence was monitored by the fluoroscanner Ascent CF for 4 additional days. At that time, U87Mg/DsRed cells in control cultures reached ~80% confluency. Within 24h after the addition of conditioned medium from 3T3/CYP2B6-eGFP, a clear decrease in fluorescence intensity was observed, reflecting a cell-killing effect. No effect on cell growth was observed after the addition of conditioned medium from 3T3/eGFP (Figure 9).

Figure 9.

Expression of CYP2B6 in cells cultured in vitro activates CPA into cytotoxic species. 3T3 fibroblasts were transduced with SF91/CYP2B6-eGFP or SF91/eGFP, and exposed to 0, 3 and 10 mm CPA for 24 h. The conditioned medium (CM) was then added at 1:3 ratio (CM:drug-free medium) to U87Mg/DsRed cells, plated three days before at 5000 per well, resulting in a real CPA/4-OH-CPA concentration of 0, 1 and 3.33 mm, respectively (right). To evaluate the efficiency of tumor-cell destruction, DsRed fluorescence was monitored by the fluoroscanner Ascent CF for four additional days. No effect on cell growth was observed after the addition of conditioned medium from 3T3/eGFP (left). Data points represent the mean of two separate experiments, with a variability of <10%.

We then investigated whether the forced expression of CYP2B6 in NSCs resulted in intracellular activation of CPA and subsequent cell death. To this aim, we transduced primary cultures of murine NSCs with SF91-CYP2B6-eGFP or with SF91-eGFP. The transduction efficiency, measured by flow cytometry as percentage of eGFP⁺ cells, was 60 and 75%, respectively. Cells were plated at 2000 cells per well. CPA was added 24 h after plating, and the incubation with the drug was continued for 6 additional days. The sensitivity of cells to CPA was measured by the MTS/PMS assay and by the count of the number of total and eGFP ⁺ neurospheres (Figure 10). Both assays revealed a clear cytotoxic activity of CPA on NSC/CYP2B6-eGFP, but not on NSC/eGFP. Instead, an apparent growth stimulatory activity of CPA on NSC/eGFP was observed by the MTS/PMS assay only, for which we have no explanation at present. At 100–200 μm CPA, only the eGFP⁺ cells of the mass population transduced with SF91/CYP2B6-eGFP (60% of the total) were prevented to form neurospheres. From 250 to 750 μm, a progressive inhibition of neurosphere formation was observed for the eGFP-negative subpopulation, presumably for a bystander effect. No cytotoxicity was observed on either eGFP⁺ or eGFP⁻ cells of the mass culture transduced with SF91/eGFP. We then investigated whether in the presence of CYP2B6-transduced NSCs, CPA was cytotoxic to DsRed-expressing U87Mg glioma cells. Briefly, we plated 200 U87Mg/DsRed per well in a microtiter plate. 24 h later, we added to each well 50 neurospheres derived from 3-day-old cultures of NSC/CYP2B6-eGFP or NSC/eGFP, and 750 μm CPA. After 48 h, a clear cytotoxic effect of CPA was observed on both neurospheres and U87Mg/DsRed in the presence of NSC/CYP2B6, but not in the presence of NSC/eGFP (Figure 11).

Figure 10.

The effects of CPA on neural stem/progenitor cells (NSCs)/CYP2B6 and NSC/eGFP cells measured by an MTS/PMS assay and by a neurosphere assay. Primary cultures of murine NSCs were transduced with SF91-CYP2B6-eGFP (●) or with SF91-eGFP (■). The transduction efficiency, measured by flow cytometry as percentage of eGFP⁺ cells, was 60 and 75%, respectively. (Upper left panel) Cells were plated at 2000 cells per well. CPA was added 24 h after plating, and the incubation with the drug was continued for 6 additional days. The sensitivity of cells to CPA was measured by the MTS/PMS assay and expressed as percentage of the absorbance value of cells treated with the solvent alone (ranging from 1.0 to 1.5 ΔA_{490 nm}). Data are the means of two separate experiments, with a variability of less than 10%. (Upper middle and right panels) The sensitivity of NSC/CYP2B6-eGFP (middle panel) and NSC/eGFP cells (right) to CPA was measured by count of total (●) and eGFP⁺ (■) neurospheres and expressed as percentage of neurospheres formed by cells treated with the solvent alone. Only neurospheres with a maximum diameter above 100 μm were counted. Data are the means of two separate experiments; (lower panels) fluorescence micrographs of CPA-treated NSC/CYP2B6-eGFP (left) and NSC/eGFP (right).

Figure 11.

Cytotoxicity of CPA on U87Mg/DsRed in the presence of neural stem/progenitor cells (NSCs)/CYP2B6-eGFP. (Upper left panel) U87Mg/DsRed cells alone; (upper right and lower left panels) U87Mg/DsRed cells cultured in the presence of NSC/CYP2B6-eGFP and exposed to 750 μm CPA for 48 h; (lower right panel) U87Mg/DsRed cells cultured in the presence of control eGFP-transduced NSCs and exposed to 750 μm CPA for 48 h.

We wished to ascertain whether treatment of glioblastoma with NSC/endostatin-eGFP would translate into tumor-growth inhibition in vivo. To this aim, we studied the effects of a single intracerebral inoculum of NSC/endostatin-eGFP in a syngeneic glioblastoma model: 3–4 days after an orthotopic implant of 2.5 × 10⁵ mouse GL-261 cells, we injected 3 × 10⁶ NSC/endostatin-eGFP cells or the same number of NSC/eGFP cells into C57Bl/6 mice at the same brain coordinates as control. The ratio between tumor volume and total brain volume was determined from glioblastoma-bearing animals harvested from mice euthanized 8 days after NSC inoculation. Although no difference in tumor size was observed between animals implanted with tumor cells alone or with tumor cells followed by NSC/eGFP (not shown), we observed a 65% decrease in tumor size of animals inoculated with NSC/endostatin-eGFP (Figure 12). This decrease in tumor size associated with endostatin-secreting NSCs was highly significant (Wilcoxon test, P = 0.02).

Figure 12.

Treatment of intracranial GL-261 gliomas with neural stem/progenitor cells (NSCs)/endo-eGFP results in significant inhibition of tumor growth. C57Bl/6 mice were stereotactically inoculated with 2.5 × 10⁵ mouse GL-261 glioblastoma cells. At day 4 post-implantation, the animals were divided into three groups and treated using the same burr hole and stereotactic coordinates with a single intracerebral inoculum of 3 × 10⁶ NSC/endo-eGFP cells (3 μl; n = 8), or 3 × 10⁶ NSC/eGFP cells (3 μl; n = 6) as control. Eight days after NSC inoculations, mice were euthanized, and tumor size determined as described in ‘Materials and methods’. Data are expressed as the ratio between tumor volume and total brain volume × 100, determined on the sections bearing tumor with the maximum visible diameter. NSC/endo-eGFP therapy resulted in significantly decreased intracranial tumor size (Wilcoxon test, P = 0.02) compared with NSC/eGFP. Columns, means of 6–8 animals (bars represent s.d.).

Discussion

Using previously published techniques^9,36 and a newly developed fluorescent glioblastoma model, we found that (1) primary murine gene-engineered NSCs displayed in vivo tropism for glioblastoma cells; (2) 10 days after injection into the brain, many NSCs continued to express the transgene and the NSC marker, nestin and (3) endostatin-expressing NSCs significantly decreased GL-261 glioblastoma growth in a syngeneic mouse model. Throughout all the experiments, we have employed NSCs prepared from neonatal forebrains of C57BL/6 mice, expanded, transduced on rFN and injected within 10 days from the brain dissection, to prevent changes in phenotype.⁹ Under our culture conditions, they do not lose their stem cell nature or multipotentiality, and the retroviral-mediated transgene expression is sustained with time in culture and upon differentiation into neurons and astrocytes.⁹ In view of the potential clinical translation of our findings, we have developed and employed in the present study high-efficiency conditions of gene transduction into NSCs that allow us to work with low MOIs. Previously, to achieve relatively high (>10%) transduction rates, it has often been necessary to employ high MOIs that generally result in multiple integrated copies per cell and a higher risk of insertional mutagenesis. The importance of the mutagenesis risk has been confirmed by recent observations of retroviral vector-mediated insertional oncogene activation in the clinics, leading to a rare form of leukemia.^37,38 The patients who developed leukemia received a particularly high number of genetically modified stem cells with multiple retroviral insertions.³⁹

We have found that NSC/endo-eGFP released endostatin into the culture medium, achieving concentrations sufficient to inhibit the growth of mouse PMVECs. We propose that NSC-mediated release of endostatin in the local brain tumor milieu will deviate the overall net balance toward angiostatic factors, resulting in the inhibition of angiogenesis and subsequent tumor regression. Prolonged local delivery of endostatin mediated by engineered NSCs should obviate the problems related to the short biological half-life and the insufficient blood flow. To improve delivery of endostatin in vivo, local microinfusion and several gene transfer approaches have been evaluated, including direct injection of the endostatin gene,^40,41 recombinant adenoviral vectors,^42-46 or microencapsulated endostatin producer cells.^14,47 The results of these different strategies have been varied, with some studies reporting significant antitumor effect, and others reporting negative or partially positive results.³⁴ Our finding of an antiendothelial effect over the range of endostatin concentrations achieved in NSC-conditioned medium is especially important in light of recent reports of a biphasic dose–reponse curve for the therapeutic efficacy of endostatin.^48,49 The concentrations of endostatin released into the medium correlate with the systemic concentrations of about 200 ng ml⁻¹ endostatin reportedly sufficient to induce tumor suppression in mice.⁵⁰

In NSC/CYP2B6-eGFP, the forced expression of CYP2B6 resulted in intracellular activation of CPA and subsequent cell death. In the presence of CYP2B6-transduced NSCs, we observed CPA cytotoxicity to DsRed-expressing U87Mg glioma cells. A well-studied prodrug-activating system, for which a very effective bystander activity has been demonstrated, is the combination of herpes simplex virus-thymidine kinase (HSV-Tk) and the antiviral nucleoside analog, ganciclovir.⁵¹ However, the cell cycle-dependent HSV-Tk-mediated cytotoxicity potentially limits its effectiveness in the treatment of malignant gliomas, in which only a small fraction of tumor cells undergo active division at any point in time.⁵² Thus, transfer of genes that sensitize tumor cells to chemotherapeutic agents that are not cellcycle dependent, such as CPA, may be a more effective strategy for these tumors. An additional advantage of our protocol is that the bystander effect mediated by CYP2B6/CPA, unlike the HSV-Tk/ganciclovir system, does not require cell–cell contact.^53,54 Thus, the intracellular activation of CPA should result in efficient killing of NSCs as well as of neighboring brain tumor cells, even in the absence of cell–cell contact. The recent report by Danks et al.⁷ of antitumor efficacy of NSC-mediated carboxylesterase activation of CPT-11, supports the hypothesis that NSC-mediated production of high concentrations of active metabolites of CPA selectively at tumor sites in the brain can be effective against glioblastoma.

The recent demonstration that glioblastomas originate from tumor-initiating stem cells whose phenotype resembles that of NSCs,⁵⁵ associated with the exceptional migratory ability of NSCs,⁵⁶ strongly suggests a great potential clinical role of NSCs as cellular therapeutic delivery system, to track glioblastoma cells and deliver anticancer molecules. However, most data in the literature have been obtained on immortalized NSC cell lines. Although they would easily provide large amounts of cells for gene transduction and transplantation, the risk for immunological rejection would be dangerously high.⁵⁷ Also, the recent findings that (1) a stem cell line can undergo spontaneous malignant transformation⁵⁸ and that (2) the most used NSC line, C17.2 differs significantly from other NSCs and does not maintain karyotypic stability⁸ cautions against the use of cell lines for human therapy. Our report of antitumor activity of primary murine NSCs transduced with an endostatin-expressing retroviral vector is encouraging for the future clinical application of primary somatic stem cells for personalized therapy of brain tumors. Although brain-derived NSCs have proven difficult to isolate from the adult CNS, recent reports^57,59 have pointed to autologous bone marrow-derived stem cells as an option to deliver therapeutic genes to the brain for treatment of glioblastoma: human bone marrow-derived neural stem-like cells behave similarly to the murine NSCs and show tropism to neural pathology.⁶⁰

Further in vivo investigation of the antitumor potential of NSC/endo-eGFP and NSC/CYP2B6-eGFP is warranted. Also, we are currently assembling a bicistronic vector co-expressing endostatin and CYP2B6: incorporation of CYP2B6 as suicide gene into endostatin-expressing NSCs may, upon administration of CPA, avoid uncontrolled growth of engineered NSCs and block production of endostatin, if needed; it would also allow, through a bystander effect, to eliminate any residual brain tumor cells remaining after the antiangiogenic effect of endostatin.