Mesenchymal Stem Cells Effectively Deliver an Oncolytic Adenovirus to Intracranial Glioma

Authors


Abstract

Gene therapy represents a promising treatment alternative for patients with malignant gliomas. Nevertheless, in the setting of these highly infiltrative tumors, transgene delivery remains a challenge. Indeed, viral vehicles tested in clinical trials often target only those tumor cells that are adjacent to the injection site. In this study, we examined the feasibility of using human mesenchymal stem cells (hMSC) to deliver a replication-competent oncolytic adenovirus (CRAd) in a model of intracranial malignant glioma. To do so, CRAds with a chimeric 5/3 fiber or RGD backbone with or without CXCR4 promoter driving E1A were examined with respect to replication and toxicity in hMSC, human astrocytes, and the human glioma cell line U87MG by quantitative polymerase chain reaction and membrane integrity assay. CRAd delivery by virus-loaded hMSC was then evaluated in vitro and in an in vivo model of mice bearing intracranial U87MG xenografts. Our results show that hMSC are effectively infected by CRAds that use the CXCR4 promoter. CRAd-CXCR4-RGD had the highest replication, followed by CRAd-CXCR4–5/3, in hMSC, with comparable levels of toxicity. In U87MG tumor cells, CRAd-CXCR4–5/3 showed the highest replication and toxicity. Virus-loaded hMSC effectively migrated in vitro and released CRAds that infected U87MG glioma cells. When injected away from the tumor site in vivo, hMSC migrated to the tumor and delivered 46-fold more viral copies than injection of CRAd-CXCR4–5/3 alone. Taken together, these results indicate that hMSC migrate and deliver CRAd to distant glioma cells. This delivery strategy should be explored further, as it could improve the outcome of oncolytic virotherapy for glioma.

Disclosure of potential conflicts of interest is found at the end of this article.

Introduction

Malignant gliomas represent the most common primary intracranial malignancy, with glioblastoma multiforme (GBM) being most frequent and carrying the worst prognosis [1, 2]. The overall poor outcome of patients with GBM is related to the fact that these tumors remain resistant to current therapy. With respect to surgical treatment, the complete resection of high-grade gliomas is not possible since these tumors infiltrate surrounding brain parenchyma [3]. To overcome these challenges, one of the novel treatments being explored is gene therapy. In this context, to achieve a robust transgene expression and a limited toxicity, various delivery systems are being developed and tested. In particular, the idea of using replication-competent oncolytic adenovirus (CRAd) vectors that can selectively replicate and kill tumor cells has evolved into the field of virotherapy. The therapeutic effect of these vectors is related to their intrinsic capacity to kill tumor cells, as well as to their ability to allow an effective and prolonged transgene expression [4, [5], [6], [7]–8].

Various adenoviral vectors have been proven safe for intracranial injection in early phase clinical trials [5, 9, [10]–11]. However, in most cases, a significant therapeutic effect has not been found when tested in patients. Apparently, after intracranial injection, these vectors have a limited reach and often do not infect glioma cells that are distant from the injection site [11]. Since high-grade gliomas are infiltrative, it is believed that tumor recurrence will remain inevitable unless remote tumor pockets are effectively targeted.

Stem cells have been explored as vehicles for gene therapy in brain tumors since they migrate toward tumor cells. For instance, neural stem cells are being tested as vehicles for targeting brain tumors [12, [13], [14], [15]–16]. In addition, human mesenchymal stem cells (hMSC) are attractive candidates as well. This is based on the fact that these cells have tropism for gliomas [17, 18]. hMSC migrate and localize glioma cells via platelet-derived growth factor-mediated, epidermal growth factor-mediated, and stromal-derived factor-1-mediated processes [17]. hMSC pose considerable advantages as vehicles for oncolytic virotherapy for brain tumors. First, hMSC can be isolated from patients and grown in culture relatively easily [19, 20] in comparison with the ethical and technical difficulties associated with neural stem cells. In addition, if isolated from the same patient that will need them, autologous transplantation overcomes the difficulties related to immune rejection of the transplanted cells. With regard to the delivery of oncolytic adenoviruses, hMSC have been shown to be infected by adenoviruses [17, 21] and have been used to deliver CRAd into lung metastases of breast carcinoma [22] and ovarian cancer [23].

In this study, we tested the ability of hMSC to deliver a CRAd to intracranial glioma. To do so, we have investigated transcriptional and transductional targeting of CRAd for glioma and hMSC. With regard to transcriptional specificity, we used the C-X-C chemokine receptor 4 (CXCR4) promoter to target adenoviral vectors to hMSC and gliomas. The rationale for using CXCR4 is based on the finding that hMSC express CXCR4 [24, 25] and the fact that the CXCR4 promoter is active in human gliomas [26]. Thus, the fact that such promoter is shared among cell carriers and glioma cells renders hMSC ideal for delivering a CRAd that uses the CXCR4 promoter. We hypothesized that this promoter could allow replication of an oncolytic virus first in carrier cells and then tumor tissue. We therefore examined the capacity of a CRAd that uses CXCR4 and contains fiber modifications previously shown by our group to enhance adenoviral glioma targeting [6, 27, [28], [29]–30] with respect to transduction, infectivity, and toxicity in hMSC, glioma cells, and normal human astrocytes, focusing on the effects of CRAd infection on hMSC migration and tumor infectivity. In this study, we show for the first time the ability of hMSC to deliver CRAd to distant glioma cells and offer a proof-of-principle for the ability of hMSC to act as carriers for oncolytic adenoviral vectors in the setting of malignant glioma.

Materials and Methods

Cells and Culture Conditions

hMSC cells of male origin were obtained from Cambrex (Walkersville, MD, http://www.cambrex.com; Lonza, Walkersville, MD, http://www.lonza.com) and characterized using CD105-, CD166-, CD29-, and CD44-positive cellular markers. Isolated cells were then expanded in MesenPro RS growth medium (12746-012; Gibco, Grand Island, NY, http://www.invitrogen.com) supplemented with 20% fetal bovine serum (FBS) (Atlanta Biologicals, Lawrenceville, GA, http://www.atlantabio.com), 1% penicillin/streptomycin (Gibco), 4 ng/ml basic fibroblast growth factor (R&D Systems Inc., Minneapolis, http://www.rndsystems.com), and 0.25% Fungizone (Gibco) in a 37°C, 5% CO2 incubator. After three passages, hMSC were used in this study. U87MG cells were grown in minimal essential medium (Eagle's) with 2 mM l-glutamine and Earle's BSS adjusted to contain 1.5 g/l sodium bicarbonate, 0.1 mM nonessential amino acids, and 1.0 mM sodium pyruvate, 90%; fetal bovine serum, 10%. Normal human astrocytes (NHA) were maintained in the AGM (Lonza) medium containing reagents from BulletKits (Lonza). Human embryonic kidney cells (HEK293, CRL-1573) and lung carcinoma (A549, CCL-185) were purchased from American Type Culture Collection (Manassas, VA, http://www.atcc.org) and propagated in Dulbecco's modified Eagle's medium supplemented with 10% FBS.

Recombinant Adenoviruses and Adenoviral Constructions

The replication-defective vectors AdWT-Luc, AdWT-RGD-Luc, and Ad5/3-Luc were described previously [6]. These vectors were created on the basis of the adenoviral wild-type 5 backbone and contain a luciferase reporter gene driven by cytomegalovirus enhancer/promoter region in the E1A region. The replication-competent vectors AdWT, Ad5/3, and AdRGD have also been described previously [31, 32], as have CRAd-CXCR4-RGD [33] and CRAd-CXCR4–5/3 [26] (Table 1). Briefly, CXCR4-driven CRAd backbones were obtained from clones on the basis of recombination of pShuttle adenoviral vectors containing a CXCR4 promoter driving E1A region and adenoviral PVK700-based backbone (the latter backbone contains a modification at fiber region-HI loop [PVK700-RGD] or full replacement of wild-type 5 knob for analogous domain from adenovirus type 3 [PVK700–5/3]). Recombination was performed in the BJ5841 Escherichia coli strain. The resultant vectors were linearized and transfected into A549 cells for rescuing viruses. The replication-deficient and -competent vectors were propagated in HEK293 and A549 cells, respectively; purified by CsCl gradient; and titrated in HEK293 cells.

Table Table 1.. Description of wild-type and CXCR4-modified vectors used in the study
original image

Infection of hMSC and U87MG with Recombinant Adenoviruses

Cells were plated 24 hours before infection in 24-well plates with a density of 5 × 104 cells per well. Then, the cells were incubated with 1,000 viral particles (vp) per cell of AdWT-Luc, AdWT-RGD-Luc, or Ad5/3-Luc vectors using 2% FBS medium. After 4 hours of adsorption, virus-containing medium was replaced with fresh medium containing 10% FBS. Forty-eight hours later, the cells were analyzed by luciferase assay. All experiments were done in triplicate. For replication experiments, 5 × 104 cells (hMSC or U87MG) were seeded 24 hours before infection with competent vector AdWT, Ad5/3, AdRGD, CRAd-CXCR4-RGD, or CRAd-CXCR4–5/3 at concentrations of 10 vp per cell [34]. After 4 hours of the adsorption period, the medium was replaced with fresh medium containing 10% FBS. After 48 hours, cells were rinsed with phosphate-buffered saline (PBS), and total DNA was isolated for quantitative polymerase chain reaction (qPCR) analysis.

Transient Transfection Reporter Assays

Efficiency of adenovirus transduction was monitored by luciferase activity with the Luciferase Assay System (Promega, Madison, WI, http://www.promega.com) in accordance with the manufacturer's protocol. Each experiment was repeated at least twice. Error bars, present in Figures 1, 2, 3, 5, and 6, indicate ±SD from the average of triplicate samples from one experiment.

Figure Figure 1..

Adenoviral transduction of human mesenchymal stem cells (hMSC). (A): Expression of putative receptors implicated in transduction of hMSC by AdWT, Ad5/3, and AdWT-RGD was detected by flow cytometry. For AdWT, the CAR receptor was tested. For Ad5/3, receptors CD46, CD80, and CD86 were tested. Finally, for AdWT-RGD, we tested αvβ3 and αvβ5 integrins. Results are expressed as percentage of cells for a given receptor. For CAR expression, the difference in gating reflects a different batch of female derived hMSC that were used for this part of the experiment. (B): Efficiency of adenovirus transduction in hMSC was monitored by luciferase activity 2 days after infection by replication-deficient vectors caring luciferase transgene (AdWT-Luc, AdWT-RGD-Luc, and Ad5/3-Luc). Transduction of hMSC by Ad5/3-Luc (*) was significantly higher than that of AdWT-Luc or AdWT-RGD-Luc (p < .05). Each experiment was done in triplicate; results were repeated two times, independently. Abbreviations: FITC, fluorescein isothiocyanate; SSC, side scatter.

Figure Figure 2..

Detection of CXCR4 transcriptional activity and toxicity. (A): Total RNA was extracted from hMSC, NHA, and U87MG. The expression levels of CXCR4 were determined by quantitative reverse transcription-polymerase chain reaction. CXCR4 expression was significantly higher in U87MG (**) and hMSC (*) in comparison with NHA (p < .05). Results were normalized by expression of glyceraldehyde-3-phosphate dehydrogenase. All samples were determined in triplicate in two independent experiments. (B): Toxicity of Ad5/3, AdRGD, CRAd-CXCR4–5/3, and CRAd-CXCR4-RGD vectors in NHA was determined 2 days after infection by membrane integrity assay based on the detection of LDH release. Results are expressed on a relative scale as percentage of toxicity. Results were obtained from six replicates per experimental condition, in two independent experiments. Abbreviations: hMSC, human mesenchymal stem cells; NHA, normal human astrocytes.

Figure Figure 3..

Replication and toxicity of CRAd in human mesenchymal stem cells (hMSC) and human glioma cell line U87MG. Replication-competent vectors AdWT, Ad5/3, AdRGD, CRAd-CXCR4–5/3, and CRAd-CXCR4-RGD were used for infection of hMSC or U87MG. Replication in hMSC (A) and U87MG (C) was determined by detection of E1A copy numbers via quantitative polymerase chain reaction. Results are expressed as fold increase from viral copy number exhibited by AdWT, normalized to human glyceraldehyde-3-phosphate dehydrogenase. Results were obtained in triplicate; two independent experiments were performed. CRAd-mediated cytotoxicity of hMSC (B) and U87MG (D) was assessed by membrane integrity assay based on the detection of LDH release. Results are expressed on a relative scale as percentage of toxicity. Results were obtained from six replicates per experimental condition, in two independent experiments. For (A) and (C), * indicates statistically significant difference over AdWT (p < .05). For (D), * indicates statistically significant difference over Ad5/3 (p < .05). ** indicates a statistically significant difference between representative groups (p < .05).

Isolation of RNA and Real-Time Quantitative Reverse Transcription-Polymerase Chain Reaction

Total cellular RNA was extracted from cells with the RNeasy mini prep kit (Qiagen, Valencia, CA, http://www1.qiagen.com) and treated with DNase I (Gibco/Life Technologies, Rockville, MD, http://www.lifetech.com) for 30 minutes. Polymerase chain reaction (PCR) products from CXCR4 were used for standard curve. RNA (1 μg) isolated from samples was used in a one-step reverse transcription (RT)-PCR. The reaction was done using the Gene Amp RNA PCR core kit (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com). All primers and probe set were designed on Primer Express software and synthesized by Applied Biosystems. The primer and probe sequences used to amplify and detect the transcript were as follows (5′–3′): CXCR4 sense, CTTCCCTTCTGGGCAGTTGA; CXCR4 antisense, ACATGGACTGCCTTGCATAGG; and probe, 6FAM-CCGTGGCAAACTGGTACTTTGGGAACT-TAMRA. A standard curve was generated using serial 10-fold dilutions for quantification by RT-PCR. The PCR included denaturation (94°C, 5 minutes) followed by 29 cycles, each consisting of denaturation (94°C, 1 minute), annealing (60°C, 1 minute), and extension (72°C, 2 minutes), with a final extension phase (10 minutes). The PCR was performed on a LightCycler system (model 2.0; Roche Diagnostic Corporation, Indianapolis, http://www.roche-diagnostics.us). Reactions were incubated at 50°C for 2 minutes, 60°C for 30 minutes, and 95°C for 5 minutes, and then subjected to 40 cycles under the following conditions: 94°C for 20 seconds and 60°C for 1 minute. Data were analyzed with LightCycler software. Each data point was repeated two times. Quantitative values were obtained from the threshold PCR cycle number (CT), where the increase in signal associated with an exponential growth of PCR product became detectable. The relative mRNA levels in each sample were normalized to human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) content.

Quantitative Real-Time PCR

Adenoviral E1A replications were evaluated for relative changes of E1A cDNA expression by quantitative PCR. Total DNA was extracted using the DNeasy Tissue kit (Qiagen). The DNA was subjected to qPCR using the primers targeting E1A and human GAPDH genes for assessment of viral replication in vitro and normalized to nanograms of DNA isolated from tumor xenograft and mouse brain tissue samples. The sequences were as follows: E1A primers, forward, 5′-AACCAGTTGCCGTGAGAGTTG-3′, and reverse, 5′-CTCGTTAAGCAAGTCCTCGATACAT-3′; and GAPDH, forward, 5′-GGTTTACATGTTCCAATA-3′, and reverse, 5′-ATGGGATTTCCATTGATGACAAG-3′. Reactions ran under the following conditions: 2 minutes at 95°C, followed by 32 cycles each consisting of 20 seconds at 94°C, 20 seconds at 56°C, and 20 seconds at 72°C. Quantification using SYBR Green PCR Master Mix (Applied Biosystems) was performed according to the manufacturer's instructions. For each qPCR, a no-template reaction was included as the negative control. The threshold cycle values for E1A and GAPDH were obtained from qPCRs and converted to the gene copy number from the standard curves. The data are presented as ratio of E1A per nanogram of total DNA.

For detection of hMSC DNA in tumor xenograft or mouse brain tissue samples, quantification of the human Y chromosome DYS14 region was done with the following primers: DYS14 forward, 5′-GGG CCA ATG TTG TAT CCT TCTC-3′, and DYS14 reverse, 5′-GCC CAT CGG TCA CTT ACA CTTC-3′. Reactions ran under the following conditions: 2 minutes at 50°C and 10 minutes at 95°C for initial denaturation of the DNA and polymerase activation, followed by 50 cycles of 1 minute at 60°C and 15 seconds at 95°C, as described previously [35].

Flow Cytometric Analysis for Receptor Surface Expression

hMSC cells were detached using Versene (Gibco). After being washed, 2 × 105 cells were used for incubation with 1 μg of mouse anti-human αvβ3 (CBL 544; Chemicon, Temecula, CA, http://www.chemicon.com), mouse anti-human αvβ5 (MAb1961z; Chemicon), CD46 (catalog number 35948; BD Biosciences, San Diego, http://www.bdbiosciences.com), CD80 (catalog number 555681; BD Pharmingen, San Diego, http://www.bdbiosciences.com/index_us.shtml), CD86 (catalog number 555663; BD Pharmingen), or CAR (catalog number ab9891; Abcam, Cambridge, MA, http://www.abcam.com) monoclonal antibody for 30 minutes at 4°C. Subsequently, cells were washed with fluorescence-activated cell sorting buffer (PBS supplemented with 1% FBS) two times and incubated with fluorescein isothiocyanate-conjugated goat anti-mouse antibody (554001; BD Pharmingen) for 30 minutes at 4°C. After incubation with the secondary antibody, cells were washed two times with washing buffer, and 104 cells were analyzed in triplicate by flow cytometry. Cell samples were resuspended in 1 ml of DPBS and then analyzed on a FACSCalibur (Becton, Dickinson and Company, Erembodegem-Aalst, Belgium, http://www.bd.com). A cytometric analysis of 10,000 events per sample was conducted using FlowJo software, version 6.3 (Tree Star, Ashland, OR, http://www.treestar.com).

3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Assay

hMSC, NHA, or U87MG cells were seeded in a 96-well plate (1.0 × 103 cells per well) overnight before being treated with or without 1,000 vp per cell of vector AdWT, Ad5/3, AdRGD, CRAd-CXCR4-RGD, or CRAd-CXCR4–5/3 for 4 hours. Then, medium was replaced with fresh medium containing 10% FBS. Forty-eight hours postinfection, 20 μl of MTS reagent (Promega) was added to each well, cells were incubated in the dark at 37°C for another 2 hours, and then optical density at 490 nanometers wavelength was measured by a microplate reader (Bio-Rad, Hercules, CA, http://www.bio-rad.com).

Labeling of hMSC Cells

A Vybrant carboxyfluorescein diacetate (CFDA-SE) succinimidyl ester (5 μM) cell tracer kit (Invitrogen, Carlsbad, CA, http://www.invitrogen.com; Molecular Probes, Eugene, OR, http://probes.invitrogen.com) was used to label hMSC according to manufacturer's instructions. Thus, for in vitro migration assays, CFDA-SE was used as a marker for hMSC cells infected or mock-infected with CRAds. One hundred thousand hMSC were seeded in a six-well plate the day before infection. The next day, cells were mock-infected or infected with 1,000 vp per cell of either CXCR4-RGD or CXCR4–5/3 for 4 hours and then incubated in 10% FBS medium for 16 hours or directly prepared for migration experiments (in vitro) without further incubation. For in vivo experiments, cells were infected for 4 hours, followed by intracranial injection without labeling or further incubation.

Cell Migration

For evaluation of migration of virus-loaded hMSC, BD Biocoat Tumor Invasion System (catalog number 354165; BD Biosciences, Bedford, MA, http://www.bdbiosciences.com) containing BD Falcon Fluoroblock 24-Multiwell inserts (8-μm pore size; PET membrane) was used in accordance with the manufacturers' instructions. Twenty thousand hMSC were plated using in 500 μl of serum-free RPMI medium in the top chamber. U87MG cells (5 × 104 cells per well) were plated in the lower chamber the night before using 10% FBS, and then medium was changed to 2% FBS medium (in the bottom chamber) before the experiment. Cell migration was analyzed 2 and 19 hours after plating hMSC. Migration was determined by counting fluorescent cells at the bottom of the upper chamber of the Fluoroblock filter using a Olympus IX81 (×2 objective), MetaMorph software (Olympus, Tokyo, http://www.olympus-global.com). For the supplemental online movies, a Zeiss Axiovert 100tv microscope (Carl Zeiss, Jena, Germany, http://www.zeiss.com) with a microscope incubator (Solent Scientific, Segensworth, U.K., http://www.solentsci.com) at 37°C, 5% CO2 was used. For counting, ×10 fields were captured and analyzed by ImageJ software. Experiments were repeated two times, each in triplicate for each condition.

To test whether migrant hMSC can release viral particles that can subsequently infect human glioma cells (U87MG), DNA from the latter cell population (from lower chambers) was analyzed for the presence of viral genomic copies. U87MG cells were harvested 24 hours after the initiation of migration experiments, and DNA was isolated for qPCR and analyzed using E1A and housekeeping gene GAPDH primers. The experiment was performed in triplicate and repeated twice. Numbers of migrating cells were calculated by the average of triplicate counted from ×10-power fields using ImageJ software. Data are presented as mean ± SD, and comparisons were made using the two-tailed Student's t test.

Animal Study

Nu/nu mice (Charles River Laboratories, Wilmington, MA, http://www.criver.com) were anesthetized with an intraperitoneal injection of 0.1 ml of a stock solution containing 25 mg/ml ketamine hydrochloride, 2.5 mg/ml xylazine, and 14.25% ethyl alcohol diluted 1:3 in 0.9% NaCl. The surgical site was shaved and prepared with 70% ethyl alcohol and Prepodyne (West Agro, Inc., Kansas City, MO, http://www.westagro.com) solution. After a midline incision, a 1-mm right parietal burr hole centered 2 mm posterior to the coronal suture and 2 mm lateral to the sagittal suture was made. Animals were then placed in a stereotactic frame, and 1 × 105 U87MG cells were injected using a 26-gauge needle to a depth of 3 mm over a period of 3 minutes. The total volume of injected cells was 5 μl. The needle was removed, the site was irrigated with sterile 0.9% NaCl, and the skin was sutured with 4-0 nylon. Seven days after tumor implantation, CRAd-CXCR4–5/3-infected hMSC or CRAd-CXCR4–5/3 were injected 5 mm anterior to the site of tumor cell injection using a similar stereotactic technique. A total of 2 × 104 hMSC infected with CRAd-CXCR4–5/3 at 100 vp per cell were injected in a volume of 5 μl. CRAd-CXCR4–5/3 (2 × 106 vp) were also injected in a volume of 5 μl. Experiments were done with three mice per group. Mice were then euthanized 1 week after injection of virus-loaded hMSC.

Analysis of Mouse Brains

Murine brains were fixed in 10% formalin and embedded in paraffin blocks, and 5-μm sections were obtained for H&E staining and subsequent analysis. Immunohistochemical analysis for detection of adenovirus was performed with goat anti-hexon antibodies (1:500 dilution; ViroStat, Portland, ME, http://www.virostat-inc.com), whereas detection of hMSC was done using anti-human CD29 antibodies (1:100 dilution; Chemicon) and detection was processed with Histostain-SP Kit (Invitrogen) and the EnVision+ system (Dako, Glostrup, Denmark, http://www.dako.com) according to the manufacturers' instructions. For viral and hMSC detection, tumor and brain parenchyma samples were separated by laser microdissection (Leica, Heerbrugg, Switzerland, http://www.leica.com) for subsequent DNA isolation and qPCR. Results are expressed as viral genomic copies per nanogram of isolated DNA. For hMSC DNA detection, a standard curve with a different ratio of U87MG/male human DNA was prepared and used to assess the ratio of hMSC DNA/total DNA.

Statistical Analysis

All data were analyzed using independent sample t tests and are expressed as mean ± SD. p < .05 was considered significant.

Results

Adenoviral Transduction of hMSC

Adenoviral transduction of hMSC is necessary to use these cells for delivering CRAds to gliomas. To study transduction, we began by evaluating the expression of putative receptors responsible for the binding of different adenoviral backbones. Expression of CAR, implicated in AdWT-Luc transduction [36]; CD46, CD80, and CD86, implicated in Ad5/3-Luc transduction [29, 37]; and αvβ3 and αvβ5 integrins, implicated in AdWT-RGD-Luc transduction [6] was assessed on hMSC by flow cytometry (Fig. 1A). The expression, presented as percentage of positive hMSC for a given receptor, was found to be as follows: CAR, 2%; CD46, 66%; CD80, 6%; CD86, 2.7%; αvβ3, 3.4%; and αvβ5, 21.6%. We then tested the efficiency of adenoviral transduction using replication-deficient vectors AdWT-Luc, Ad5/3-Luc, and AdWT-RGD-Luc. These vectors contain a luciferase transgene, and thus, transduction was assessed by monitoring luciferase activity 2 days after infection of hMSC (Fig. 1B). Ad5/3-Luc showed significantly higher transduction (444,225.16 ± 16,328 RLU/mg of protein) than AdWT-RGD-Luc (46,733.55 ± 3,533 RLU/mg of protein) or AdWT-Luc (26,697.5 ± 9,795.4 RLU/mg of protein) (p < .05).

Selectivity of CXCR4 Promoter Activity and Toxicity of CXCR4-Driven CRAd

CXCR4 has been shown to be expressed in hMSC [24, 38], and the CXCR4 promoter is highly active in gliomas [26]. Therefore, we hypothesized that the use of this promoter could limit replication of CRAds to hMSC and glioma cells. To this end, we compared the activity of CXCR4 promoter in hMSC, NHA, and U87MG glioma cell line by quantitative RT-PCR (Fig. 2A). The CXCR4 promoter was found to be significantly more active in U87MG (99.1 ± 19 copies per nanogram of RNA), followed by hMSC (49.69 ± 26.79 copies per nanogram of RNA), and negligible in NHA (3.67 ± 0.3 copies per nanogram of RNA) (p < .05). To evaluate the potential toxicity of CRAds with the CXCR4 promoter, we tested the cytotoxicity of these vectors in NHA. Toxicity of Ad5/3, AdRGD, CRAd-CXCR4–5/3, and CRAd-CXCR4-RGD in NHA was determined 2 days after infection by membrane integrity assay based on the detection of LDH release (Fig. 2B). Results are expressed on a relative scale as percentage of toxicity: CRAd-CXCR4-RGD, 7.66% ± 0.46%; AdRGD, 7.34% ± 0.31%; CRAd-CXCR4–5/3, 6.47% ± 0.42%; and Ad5/3, 3.37% ± 0.07%. Although the toxicity in NHA varied among the CRAds tested, all of the values obtained were relatively low (less than 8%).

Replication and Toxicity of CRAds in hMSC and Human Glioma Cell Line U87MG

To assess the replication and lytic potential of candidate CXCR4-driven CRAd in cell carriers and target cells, vectors AdWT, Ad5/3, AdRGD, CRAd-CXCR4–5/3, and CRAd-CXCR4-RGD were used for infection of hMSC and U87MG. Replication of these viruses was determined 2 days after infection by qPCR for viral gene E1A. Results from viral replication are expressed as fold increase from AdWT (E1A copy number). In hMSC, replication was found to be as follows: Ad5/3, 2.95 ± 0.49; CRAd-CXCR4–5/3, 38.6 ± 2.38; AdRGD, 17.6 ± 1.24; and CRAd-CXCR4-RGD, 242.3 ± 23.1 (Fig. 3A). Consistent with CXCR4 promoter activity in these cells, CXCR4-driven CRAds replicated multiple fold higher than their controls in hMSC (p < .05), and CRAd-CXCR4-RGD had a significantly higher replication than CRAd-CXCR4–5/3 (p < .05). To evaluate the lytic potential of these vectors in hMSC, a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium (MTT) assay was performed 2 days after infection, and the results are expressed as percentage of toxicity (Fig. 3B). Cytotoxicity of hMSC was found to be as follows: Ad5/3, 54.59% ± 3.44%; CRAd-CXCR4–5/3, 54.49% ± 4.4%; AdRGD, 62.38% ± 3.39%; and CRAd-CXCR4-RGD, 62.83% ± 3.39%. These CRAds elicited a considerable toxicity in hMSC after 2 days. It is important to stress, however, that although replication of these vectors in hMSC was enhanced by CXCR4 promoter, the presence of this transcriptional regulator did not augment toxicity of such vectors in hMSC. To evaluate the replication and oncolysis that CXCR4-driven CRAds might elicit in glioma cells, these processes were investigated in U87MG cells 2 days after infection by these vectors. As shown in Figure 3C, replication in U87MG was found to be as follows: Ad5/3, 2.52 ± 0.15; CRAd-CXCR4–5/3, 1,783.53 ± 10.72; AdRGD, 8.18 ± 0.91; and CRAd-CXCR4-RGD, 273.81 ± 31.36. CRAd-CXCR4–5/3 exhibited significantly higher replication in U87MG cells than any other vector tested, and the presence of CXCR4 promoter enhanced replicative capacity of both of the vectors. Finally, the oncolytic effect of CRAd in U87MG was determined by MTT assay 2 days after infection. As shown in Figure 3D, the following results were obtained: Ad5/3, 12.87% ± 1.56%; AdRGD, 11.56% ± 0.8%; CRAd-CXCR4–5/3, 55.36% ± 3.26%; and CRAd-CXCR4-RGD, 7.16% ± 0.55%. Consistent with the results of replication in U87MG cells, CXCR4–5/3 was significantly more oncolytic than the other vectors tested (p < .05). Similarly, results obtained 6 days after infection of U87MG cells showed that the highest oncolytic effect was elicited by CRAd-CXCR4–5/3 vector (data not shown).

hMSC Deliver CRAd to Glioma Cells In Vitro

To investigate the capability of hMSC to deliver CRAd to glioma cells, we first tested the migration capacity of these cell carriers in vitro. For these experiments, hMSC were labeled with CFDA. To document migration of hMSC, consecutive fluorescent micrographs were taken, and representative pictures are presented in Figure 4. A movie showing the migration of CRAd-loaded hMSC is available for download from the supplemental online data. As clearly documented, CRAd-loaded hMSC effectively migrate in vitro over a relatively short period of time.

Figure Figure 4..

Assessment of migration of human mesenchymal stem cells (hMSC). To document migration of hMSC, consecutive fluorescent micrographs were taken every min for 15 min using a ×10 objective. Movie is available for download from the supplemental online data. For this assay, cells were plated on a Tumor Invasion System containing Fluoroblock. Abbreviation: min, minutes.

As shown in Figure 3B, infection by CXCR4-driven CRAd results in considerable toxicity in hMSC. This toxicity might reflect the release of viral particles by these cell carriers, a desired process for the end of CRAd delivery. On the other hand, such toxicity might limit the migratory capacity of hMSC. To explore the latter possibility, we tested the migration of CRAd-infected hMSC over time in a Tumor Invasion System containing Fluoroblock. First, migration of CRAd-infected hMSC was assessed during the first cycle of adenoviral replication (Fig. 5A, top). Interestingly, in this setting, migration of CRAd-infected hMSC was similar to that shown by uninfected hMSC at two and 19 hours after the initiation of the experiment (Fig. 5A, middle) (p > .05), and in all conditions, the number of hMSC found on the lower end of the insert increased over time (p < .05). However, if the initiation of migration experiments was preceded by incubation of hMSC for 16 hours after infection (the time required for completion of one adenoviral replication cycle) (Fig. 5B, top), the migration of CRAd-infected hMSC was impaired in comparison with the migration shown by uninfected hMSC. This is suggested by the observation that there were fewer hMSC on the lower end of the insert in the case of CRAd-infected hMSC versus uninfected hMSC (Fig. 5B, middle) (p < .05).

Figure Figure 5..

hMSC deliver CRAd to glioma cells in vitro. hMSC were infected by CRAd during a 4-hr adsorption period. Then, hMSC were stained with carboxyfluorescein diacetate and used immediately for migration experiments ([A], top) or incubated for 16 hrs prior to migration experiments ([B], top). The cells that migrated to the lower end of the inserts were photographed and counted using ImageJ software ([A, B], middle). In addition, the ability of hMSC-infected cells to release viral progeny and infect U87MG cells grown in bottom chambers was evaluated via quantitative polymerase chain reaction (qPCR) ([A, B], bottom). Replication was determined by qPCR for viral gene E1A for estimation of viral genomic copies. Results were normalized by human glyceraldehyde-3-phosphate dehydrogenase. Abbreviations: FBS, fetal bovine serum; hMSC, human mesenchymal stem cells; hrs, hours; PF, power field; Rep., replication.

To test the ability of hMSC to deliver CRAd to glioma cells, U87MG cells were plated in the bottom chambers of the Tumor Invasion System. Twenty-four hours after the beginning of the migration experiments, U87MG cells were collected and DNA was obtained to assess CRAd replication in glioma cells by qPCR. Viral DNA was encountered and reflected infection of U87MG cells by CRAd delivered via hMSC while undergoing the first cycle of viral replication (CRAd-CXCR4–5/3, 30.83 ± 10.73, and CRAd-CXCR4-RGD, 57.46 ± 6.8 E1A copies per nanogram of DNA) (Fig. 5A, bottom), as well as hMSC that migrated after the first cycle of viral replication (CRAd-CXCR4–5/3, 58.13 ± 6.91, and CRAd-CXCR4-RGD, 1,062.12 ± 55.1 E1A copies per nanogram of DNA) (Fig. 5B, bottom). In both cases, there were more genomic copies of CRAd-CXCR4-RGD than of CXCR4–5/3 (p < .05). This finding is consistent with the higher replication shown by CRAd-CXCR4-RGD in hMSC (Fig. 3A) and clearly shows that CRAd-infected hMSC can deliver the oncolytic virus to target cells in vitro. However, given the finding that CRAd-CXCR4–5/3 has a higher oncolytic effect in glioma cells (Fig. 3D) than CRAd-CXCR4-RGD, the former vector likely represents an optimal vector for glioma virotherapy and was therefore chosen by us for additional in vivo studies.

hMSC Deliver CRAds to Human Glioma Cells Growing in Murine Brains

To assess the feasibility of CRAd delivery by hMSC to glioma cells in vivo, we tested this strategy in brains of mice bearing intracranial U87MG tumor xenografts. After 7 days of tumor growth, hMSC, CRAd-CXCR4–5/3, or CRAd-CXCR4–5/3-loaded hMSC were injected 5 mm anterior to the U87MG injection site. Mice were euthanized 7 days later (14 days after U87MG tumor establishment), and brains were collected for further analysis. The presence of tumors was corroborated by H&E staining (Fig. 6A). To evaluate the infection of tumor cells by CRAd-CXCR4–5/3, 5-μm slices were analyzed by immunohistochemistry with adenoviral anti-hexon antibody. Adenoviral antigens were uniformly and abundantly detected in tumors from mice treated with CRAd-CXCR4–5/3-loaded hMSC, and a weaker anti-hexon staining was observed in tumors of mice treated with CRAd-CXCR4–5/3 alone (Fig. 6A). In the brains of mice injected with CRAd-CXCR4–5/3-loaded hMSC, anti-hexon staining was present at the site of CRAd-CXCR4–5/3-loaded hMSC injection (Fig. 6B, white arrow) and at the tumor (Fig. 6B, black arrow), with only a weak signal present along the hMSC path of migration. This finding confirms that the use of hMSC to deliver CRAd-CXCR4–5/3 in murine brains leads to a higher adenoviral infection of distant glioma cells.

Figure Figure 6..

hMSC deliver CRAd-CXCR4–5/3 to glioma cells in murine brains. (A): Nu/nu mice were injected intracranially with U87MG cells. After 7 days, hMSC (2 × 104 cells), CRAd-CXCR4–5/3 (2 × 106 vp), or CRAd-CXCR4–5/3 loaded-hMSC (2 × 104 cells infected with 100 vp per cell) were injected 5 mm anterior to the U87 injection site. Mice were euthanized 7 days later, and brains were collected. (A, B): The presence of tumors was corroborated by H&E staining. Infection of tumor cells by CRAd-CXCR4–5/3 was assessed by IHC with adenoviral anti-hexon antibody. In the brains of mice injected with CRAd-CXCR4–5/3-loaded hMSC, anti-hexon staining was present at the site of CRAd-CXCR4–5/3-loaded hMSC injection ([B], white arrow) and at the tumor ([B], black arrow), while remaining absent elsewhere in the brain (B). (C): Quantitative assessment of CRAd-CXCR4–5/3 delivery by hMSC into glioma cells was performed on tumor and brain parenchyma samples separated by 5-μm slices by laser microdissection, and subsequently, DNA was isolated and viral replication was analyzed by quantitative polymerase chain reaction (qPCR). (D): Immunohistochemical staining of the tumor slices was performed using an anti-human CD29 antibody, an integrin that is known to be expressed by hMSC. Tumors of mice (black arrow) treated with hMSC infected with CRAd-CXCR4–5/3 stained positive for CD29 (yellow arrow), whereas tumors of mice treated with CRAd-CXCR4–5/3 did not show any staining for this antigen. Finally, since the hMSC used were obtained from a male donor and the U87MG glioma cell line was isolated from a female patient, hMSC DNA was quantified by amplifying DYS14 region of the human Y chromosome by qPCR. (E): hMSC DNA was encountered in a ratio of 9.17 × 10−3 ± 1.36 × 10−3 hMSC DNA/total DNA in the tumors and 6.35 × 10−4 ± 2.76 × 10−4 hMSC DNA/total DNA in normal brain samples of mice treated with CRAd-CXCR4–5/3-infected hMSC, and it was not detected in mice treated with CRAd-CXCR4–5/3 alone. * indicates a statistically significant difference compared with all other groups (p < .05). Results are expressed as viral genomic copies per nanogram of isolated DNA. Abbreviations: hMSC, human mesenchymal stem cells; IHC, immunohistochemistry; NB, normal brain; ND, none detected.

To obtain a quantitative assessment of CRAd-CXCR4–5/3 delivery by hMSC into glioma cells, 5-μm slices were used to separate tumor and brain parenchyma samples by laser microdissection; subsequently, DNA was isolated, and viral replication was analyzed by qPCR. CRAd-CXCR4–5/3 delivery by hMSC led to 245.4 ± 49 viral genomic copies per nanogram of glioma DNA versus 5.3 ± 0.022 viral genomic copies per nanogram of glioma DNA achieved by distant injection of CRAd-CXCR4–5/3 virus alone (Fig. 6C). This difference represents a 46-fold increase in viral copies in distant glioma tissue when the vector was delivered by hMSC (p < .05).

The presence of hMSC at the site of the tumor was investigated via two independent experiments. First, immunohistochemical staining of the tumor slices was performed using an anti-human CD29 antibody, an integrin that is known to be expressed by hMSC [39]. We observed that tumors of mice treated with hMSC infected with CRAd-CXCR4–5/3 stained positive for CD29, whereas tumors of mice treated with CRAd-CXCR4–5/3 did not show any staining for this antigen (Fig. 6D). Moreover, since the hMSC used for this experiment were obtained from a male donor, whereas the U87MG glioma cell line was isolated from a female patient (American Type Culture Collection), we quantified the DNA isolated from tumors or brain samples after separating them by laser microdissection. To quantify hMSC DNA, we amplified the DYS14 region of the human Y chromosome using a qPCR protocol described earlier [35]. hMSC Y-chromosome DNA was encountered at a ratio of 9.17 × 10−3 ± 1.36 × 10−3 hMSC DNA/total DNA in the tumors and 6.35 × 10−4 ± 2.76 × 10−4 hMSC DNA/total DNA in normal brain samples of mice treated with CRAd-CXCR4–5/3-infected hMSC, and it was not detected in mice treated with CRAd-CXCR4–5/3 alone (p < .05) (Fig. 6E).

Discussion

In this work, we provide a proof-of-principle that hMSC are capable of delivering CRAds into distant glioma cells in the brain. To our knowledge, this is the first report describing the use of any kind of stem cells for delivering conditionally replicative adenoviral vectors in the setting of intracranial glioma. This novel strategy therefore represents a potential means of improving the effects of CRAds for the treatment of brain tumors.

To achieve this end, we followed a systematic approach. To be delivered by a cellular vehicle, a specific CRAd should be capable of transducing both cell types, the cellular vehicle as well as the tumor cell to be targeted. We present evidence that adenoviral vectors bearing 5/3 or RGD modifications on their fiber knobs are capable of effectively transducing hMSC. This finding is supported by Knaän-Shanzer et al., who have shown the infection of hMSC by different species of group B human adenovirus [40]. The scarce expression levels of CAR, CD80, and CD86 are consistent with the previously described absence of these receptors on hMSC [40].

Next, we tested CXCR4-driven CRAds, since we hypothesized that with this promoter, it is possible to achieve transcriptional targeting of neoplastic cells while maintaining appropriate replication within hMSC. As our group has described before, the CXCR4 promoter is upregulated in both passaged glioma cell lines and primary glioma specimens [26]. At the same time, CXCR4 is expressed in hMSC, and its presence is closely related to the functional phenotype as it has been shown to be involved in hMSC migration [41]. The incorporation of CXCR4 promoter enhances CRAd replication and toxicity in hMSC. This is suggested by comparison of Ad5/3 and AdRGD with CRAd-CXCR4–5/3 and CRAd-CXCR4-RGD. Thus, for the goal of delivery of replication-competent viruses by hMSC, the use of this promoter leads to a greater viral yield. On the other hand, as described before [22], we observed that replication-competent viruses elicit toxicity in hMSC, and the toxicity correlates with replication. In the present study, we showed that this toxicity limits the in vitro migration of hMSC to some extent. Nevertheless, since this toxicity increases over time as viral replication takes place, injection of hMSC right after adsorption might allow migration of a considerable number of cells.

Furthermore, we hypothesize that as hMSC approach glioma cells, these vehicles become subject to CRAd-induced toxicity and release viral progeny capable of targeting tumor cells. To overcome suboptimal delivery secondary to impaired hMSC migration, two possibilities exist: (a) decreasing the number of viral particles used for the infection of hMSC or (b) searching for CRAds with a higher ratio of target cell/vehicle cell toxicity. For our in vivo study, we have decreased our virus/cell ratio to 100 vp per hMSC, which led to a significant increase in CRAd-CXCR4–5/3 delivery to distant glioma cells in comparison with the injection of a similar number of viral particles. This delivery was achieved in spite of the toxicity that the CRAd-CXCR4–5/3 vector might have elicited in hMSC. Since the nature of our oncolytic vectors is to replicate and kill their host cells, we assume that the delivery of only a few survivor hMSC carrying these vectors could potentially lead to an enhanced killing of distant tumor cells and achieve a significant therapeutic effect.

With regard to the route of administration, we tested intracranial administration of hMSC at a site distant from that of tumor cells for a series of reasons. First, we believe that this route of administration resembles a possible clinical scenario where CRAd-loaded hMSC could be injected into the walls of tumor resection cavity after tumor debulking. In addition, we think that intracranial administration might provide advantages over systemic delivery of hMSC. Although intra-arterial injection of hMSC has been shown to lead to the localization of these cells in glioma xenografts, this effect is achieved only following disruption of the blood-brain barrier [17]. Moreover, the ability of hMSC to localize remote tumor pockets with few cells has not been demonstrated after systemic delivery. Since the presence of a few tumor cells might not disrupt the blood-brain barrier, the possibility that intracranial administration of hMSC might enhance the reach of these tumor pockets is likely and should be investigated further.

Finally, we readily acknowledge the potential limitations of our study. First, all of the experiments were performed in a murine model. The degree to which these results can be translated to humans remains to be seen and established. Nevertheless, the established capacity of hMSC to migrate through the brain parenchyma suggests that these cells can survive within the CNS for a prolonged period of time, a prerequisite condition for using hMSC within a much larger human brain. Second, as previously pointed out, hMSC are susceptible to cytotoxicity secondary to infection with an oncolytic virus. Although this can clearly affect the number of hMSC migrating to the tumor site, the delivery of an oncolytic virus by the remaining hMSC appears superior to injection of the virus alone.

Conclusion

This work is an exploration of a novel way to deliver oncolytic adenoviruses to brain tumors. We show that hMSC provide CRAd delivery to distant glioma cells and that this delivery significantly enhances the infection of tumor cells in comparison with the injection of distant CRAd alone. Since hMSC can reach distant glioma cells, the latter consequently become accessible to infection and toxicity elicited by these viruses and thus might pose a therapeutic advantage.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.

Acknowledgements

We thank Yu Han (Brain Tumor Center, University of Chicago) for help with flow cytometry analysis and propagation of CRAds, U87MG, and hMSC. We thank Dr. Vytas Bandokas (Immunofluorescence Core Facility, University of Chicago) for technical support. This work was supported by the National Institute of Neurological Disorders and Stroke Grant K08-NS046430, the Alliance for Cancer Gene Therapy Young Investigator Award, and the American Cancer Society Grant RSG-07-276-01-MGO.

Ancillary