Mesenchymal Stem Cells DisplayTumor-Specific Tropism in anRCAS/Ntv-a Glioma Model1

Tiffany Doucette*, Ganesh Rao*, Yuhui Yang*,Joy Gumin*, Naoki Shinojima*, B. Nebiyou Bekele†,Wei Qiao†, Wei Zhang‡ and Frederick F. Lang*

*Department of Neurosurgery, The University of Texas,MD Anderson Cancer Center, Houston, TX, USA;†Department of Biostatistics, The University of Texas,MD Anderson Cancer Center, Houston, TX, USA;‡Department of Pathology, The University of Texas,MD Anderson Cancer Center, Houston, TX, USA

AbstractBone marrow–derived mesenchymal stem cells (MSCs) have been shown to localize to gliomas and deliver thera-peutic agents. However, the clinical translation of MSCs remains poorly defined because previous studies relied onglioma models with uncertain relevance to human disease, typically xenograft models in immunocompromisedmice. To address this shortcoming, we used the RCAS/Ntv-a system, in which endogenous gliomas that recapitulatethe tumor and stromal features of human gliomas develop in immunocompetent mice. MSCs were harvested frombonemarrow of Ntv-amice and injected into the carotid artery of Ntv-amice previously inoculated with RCAS–PDGF-Band RCAS-IGFBP2 to inducemalignant gliomas (n=9). MSCswere labeledwith luciferase for in vivo bioluminescenceimaging (BLI). After intra-arterial injection, BLI revealed MSCs in the right frontal lobe in seven of nine mice. At nec-ropsy, gliomas were detected within the right frontal lobe in all these mice, correlating with the location of the MSCs.In the twomicewithoutMSCs based on BLI, no tumor was found, indicating thatMSC localizationwas tumor specific.In another cohort of mice (n = 9), MSCs were labeled with SP-DiI, a fluorescent vital dye. After intra-arterial injection,fluorescence microscopy revealed SP-DiI–labeled MSCs throughout tumors 1 to 7 days after injection but not innontumoral areas of the brain. MSCs injected intravenously did not localize to tumors (n = 12). We conclude thatsyngeneic MSCs are capable of homing to endogenous gliomas in immunocompetent mice. These findings supportthe use of MSCs as tumor-specific delivery vehicles for treating gliomas.

Neoplasia (2011) 13, 716–725

IntroductionDespite multimodal therapy, patients with high-grade gliomas have apoor prognosis, with a median survival of 14 months for grade 4 and36 months for grade 3 gliomas according to the World Health Orga-nization [1,2]. This poor outcome is due at least in part to an inabilityto deliver therapeutic agents to the infiltrative tumor cells. Delivery ofmost conventional therapies is restricted by the blood-brain-tumorbarrier, and the efficacy of newer biologic therapies, such as geneand viral therapies, has been limited by the inability to deliver thetherapeutic genes to most of the tumor [3]. Consequently, significantefforts have been undertaken to develop new methods for deliveringtherapeutic agents to infiltrative gliomas [4].Recent evidence has suggested that stem cells may be effective de-

livery vehicles for a variety of solid tumors arising throughout thebody. Initial studies of brain tumors indicated that neural stem cells

Abbreviations: BLI, bioluminescence imaging; GSC, glioma stem cell; H&E, hema-toxylin and eosin; hMSC, human mesenchymal stem cell; IGFBP2, insulin-likegrowth factor–binding protein 2; MSC, mesenchymal stem cell; RCAS, replication-competent, avian leukosis virus, splice acceptorAddress all correspondence to: Frederick F. Lang, MD, Department of Neurosurgery,Unit 442, The University of Texas, MD Anderson Cancer Center, 1515 HolcombeBlvd, Houston, TX 77030. E-mail: flang@mdanderson.org1This study was supported by grants from the National Cancer Institute (R01CA115729 and P50 CA127001) to F.F.L.; the Elias Family Fund for Brain TumorResearch, the Gene Pennebaker Brain Cancer Fund, the Brian McCulloph ResearchFund, and the Run for the Roses Foundation to F.F.L.; and the MD Anderson Cen-ter for Targeted Therapy, and NINDS (NS070928) to G.R. This research is also sup-ported in part by the National Institutes of Health through MD Anderson’s CancerCenter Support Grant (CA016672).Received 5 December 2010; Revised 22 May 2011; Accepted 25 May 2011

Copyright © 2011 Neoplasia Press, Inc. All rights reserved 1522-8002/11/$25.00DOI 10.1593/neo.101680

www.neoplasia.com

Volume 13 Number 8 August 2011 pp. 716–725 716

have the capacity to migrate toward and deliver therapeutic agents tohuman gliomas [5,6]. More recently, mesenchymal stem cells(MSCs), acquired either from bone marrow or from adipose tissue,have been explored as potential delivery vehicles in the treatment ofhuman gliomas [7]. MSCs are defined by their ability to grow as ad-herent cells in culture, to express an array of positive and negativesurface markers, and to differentiate into adipocytes, chondrocytes,and osteocytes under appropriate conditions [8,9]. Compared withall other stem cells [10–12], MSCs have the clinical advantages thatthey can be easily acquired from patients, are readily expanded ex vivo,can be autotransfused without fear of immune rejection, and are freeof ethical concerns [13–15].We [16] and others [17,18] have demonstrated the feasibility of

using human MSCs (hMSCs) as delivery vehicles for glioma therapy.Specifically, hMSCs have been shown to be able to localize to malig-nant gliomas after systemic delivery and to track infiltrating tumor cellsafter local delivery [16]. The tropism of hMSCs for glioma seems to bemediated at least in part by specific tumor-derived growth factors, par-ticularly platelet-derived growth factor-BB [16,19]. Importantly, thehoming capacity of hMSCs can be translated into therapeutic benefitbecause hMSCs can be engineered to deliver biologic antiglioma agentsto gliomas, including interferon β [16], S-TRAIL [18,20], and oncolyticviruses [21], with demonstrable survival advantages [16,22].Despite these promising results, the clinical translation of MSC-

based therapeutic approaches remains incompletely defined becausenearly all published studies analyzing the application of MSCs in glio-mas have relied on tumor models that fail to recapitulate the genotypicand/or phenotypic characteristics of human tumor cells and the micro-environment in which they arise. The shortcomings of xenograft mod-els of brain tumors for preclinical testing have been well described[23,24]. To date, most studies analyzing the tropism of MSCs forhuman gliomas have relied on commercially available human gliomacell lines (e.g., U87, U251) that poorly mimic the invasive growth ofgliomas in patients. In addition, because these cell lines are grown asxenografts in nude mice, the interaction of the tumor with the micro-environment or stroma is, by necessity, artificial, raising concerns aboutthe true attraction of exogenously delivered MSCs for human gliomas.Although a few studies have used newer human glioma stem cells(GSCs) [18,25], which are derived directly from patient tumors andgrow as highly infiltrative tumors, GSCs must also be grown as xeno-grafts in immunodeficient mice, which obviates the impact of the tu-mor microenvironment and the immune system on MSC engraftment.Syngeneic rodent models using commercially available murine gliomasimplanted in the brain have allowed for some assessment of immunefactors influencing MSC migration, but these models poorly recapitu-late the genetics of human tumors, and thus their relevance to the hu-man disease is questionable. These concerns call into question theutility of these model systems for evaluating novel therapeutic strategies[26,27]. Lastly, in all previous studies, MSCs have been injected withinonly a few days of tumor implantation, making it difficult to differen-tiate the ability of MSCs to localize to areas of injury resulting from theprocess of cell implantation compared with localization to tumors intheir natural setting [28].Since its initial description by Holland et al. [29], the RCAS/Ntv-a

system has been extensively used to model brain tumors (reviewed inBecher and Holland [30]). This system uses a modified avian leukosisvirus (the replication-competent\, avian leukosis virus, splice acceptor[RCAS] vector) to deliver genes of interest to brain cells in transgenicmice that express the receptor (TVA) for the virus. TVA expression is

driven by the Nestin promoter, making glioneuronal precursor cellsspecifically susceptible to infection and subsequent expression of aninserted gene. This method of somatic cell gene transfer has been usedto model a variety of brain tumors [29,31–33]. In particular, somatictransfer of RCAS–PDGF-B and RCAS-IGFBP2 results in murinegliomas with all the histologic features of grade 4 gliomas identicalto human gliomas [34]. Activation of the PDGF-B and insulin-likegrowth factor–binding protein 2 (IGFBP2) signaling pathways is rel-evant in glioma because up to 30% of human gliomas demonstrateevidence of aberrant PDGF expression [35,36], and most high-gradehuman gliomas overexpress IGFBP2. Importantly, tumor cells in thismodel have stem-like features and have been shown to reside withinthe perivascular niche, indicating that the model faithfully recapitu-lates the microenvironment and cellular interactions of human glio-mas [37,38]. Recent work by our group has shown the benefit ofpreclinical testing of agents with this model [39]. The RCAS/Ntv-asystem is ideal for studying the capacity of MSCs to home to gliomasnot only because the gliomas in this model recapitulate the phenotypeand genotype of human gliomas but also because the gliomas ariseendogenously within the native stroma of the brain in immunocom-petent mice and permit natural interactions of MSCs with the sur-rounding microenvironment [40,41].Given the advantages of the RCAS/Ntv-a system, we sought to de-

termine the extent to which MSCs isolated from Ntv-a mice are ca-pable of localizing to endogenous high-grade gliomas induced in thebrains of Ntv-a mice by somatic transfer of PDGF-B and IGFBP2.Our results demonstrate that MSCs obtained from Ntv-a mice canengraft endogenous brain tumors after systemic delivery, supportingthe evolving concept that MSCs have an intrinsic capacity to hometo gliomas.

Materials and Methods

Harvesting and Characterization of MSCs from Ntv-a MiceMSCs were harvested using the technique described by Peister et al.

[42]. Briefly, the cells from the long bones of Ntv-a mice were isolatedand centrifuged at 1500 rpm. The pellet was then resuspended in α-minimal essential medium (α-MEM) and 20% MSC supplement me-dium (StemCell Technologies, Vancouver, British Columbia, Canada).Cells were then plated in a T75 flask in α-MEM with 20%MSC sup-plement medium and incubated at 37°C and 5% CO2. After 24 hours,nonadherent cells were discarded; adherent cells were washed with PBS,and fresh complete isolation medium was added every 3 to 4 days for4 weeks. Cells were collected after trypsinization and replated in 30 mlof complete isolation medium in 175-cm3 flasks. After 1 to 2 weeks,passage 3 cells were either frozen or further expanded by plating at50 cells/cm2 and incubating in the complete expansion medium. Bypassage 4, cells were analyzed using FACS for the mesenchymal mark-ers Sca-1 and CD9, the hematopoietic markers CD45 and CD11b, thelymphocytic marker CD73, the endothelial markers CD31 and CD105,and the GSC marker CD133. Cells were plated in α-MEM with 10%fetal bovine serum (FBS) and supplemented with L-glutamine. Furtherverification of a mesenchymal lineage was performed by plating the cellsin adipogenic, chondrogenic, and osteogenic media.

Differentiation Methods

Adipogenesis. Cells were seeded at 105 cells/μl in a six-well plate. At100% confluence, cell differentiation was induced with supplemented

Neoplasia Vol. 13, No. 8, 2011 Tropism of MSCs for Endogenous Gliomas in Mice Doucette et al. 717

adipogenesis induction medium (Lonza, Walkersville, MD) for 3 days,followed by 3 days of culture in supplemented adipogenesis mainte-nance medium (Lonza). After four cycles of induction/maintenance,cells were rinsed with PBS, fixed with 10% buffered formalin, andstained with Oil Red O.

Osteogenesis. Cells were seeded at 5 × 104 cells/μl in a six-wellplate. After 24 hours, cell differentiation was induced with the osteo-genesis induction medium (Lonza). Cells were fed every 3 to 4 daysby completely replacing the medium with a fresh osteogenesis induc-tion medium. After 3 to 4 weeks, cells were rinsed in PBS, fixed with70% ethanol, and stained with Alizarin Red.

Chondrogenesis. Cell pellets were prepared by spinning down 3 ×105 cells in 15-ml polypropylene tubes and growing the cells in thecomplete chondrogenic medium (Lonza). Cell pellets were fed every2 to 3 days by completely replacing the medium with a freshly pre-pared complete chondrogenic medium. After 3 to 4 weeks, pelletswere fixed in buffered 10% formalin and embedded in paraffin. Then,5-μm sections were slide-mounted and stained for glycosaminoglycanswith Safranin O.

Generation of Gliomas in Ntv-a Mice

Vector constructs. RCAS–PDGF-B was constructed with a hemag-glutinin epitope tag as described previously [43]. Details of the cre-ation of RCAS-IGFBP2 are described elsewhere [34]. Briefly, theRCAS-IGFBP2 vector was constructed by subcloning the 1.4-kb com-plementary DNA fragment of IGFBP2 into a Yap vector, which wasthen transferred into the RCAS-X vector using NotI and ClaI restric-tion enzymes. RCAS-GFP was provided by Dr Yi Li (Baylor Collegeof Medicine).

Transfection of DF-1 cells. DF-1–immortalized chicken fibro-blasts were grown in Dulbecco modified Eagle medium with 10%FBS (GIBCO, Carlsbad, CA) in a 5% CO2 humidified incubatorat 37°C. To produce live virus, plasmid versions of RCAS vectors weretransfected into DF-1 cells using FuGene6 (Roche, Nutley, NJ) andallowed to replicate in culture.

In vivo somatic cell transfer in transgenic mice. Creation of thetransgenic Ntv-a mouse has been previously described [29]; the miceare mixtures of the following strains: C57BL/6, BALB/C, FVB/N,and CD1. To transfer genes through the RCAS vectors, DF-1 pro-ducer cells transfected with a particular RCAS vector (105 cells in 1-2 μl of PBS) were injected into the right frontal lobe of Ntv-a mice froman entry point just anterior to the coronal suture of the skull using a10-μl gas-tight Hamilton syringe. Equal numbers of DF-1 producercells for each vector (RCAS–PDGF-B or RCAS-IGFBP2) were coin-jected to induce gliomas. Control mice were injected with RCAS-GFP. To determine whether MSCs were themselves tumorigenic, wecoinjected RCAS–PDGF-B with bone marrow–derived MSCs fromNtv-a mice. Escalating concentrations of MSCs (1 × 103, 1 × 104,and 1 × 105) in 1 μl of α-MEM plus 10% FBS were coinjected withRCAS–PDGF-B–infected producer cells (1 × 105 cells in 1 μl of PBS)into cohorts of 30 mice each. A control group of 30 mice was injectedwith RCAS–PDGF-B alone. We injected mice within 48 to 72 hoursafter birth because the population of Nestin+ cells producing TVA re-ceptors diminishes progressively with time. To identify the presence or

absence of tumor, we killed the mice by CO2 inhalation. Their brainswere removed, fixed in formalin, embedded in paraffin, stained withhematoxylin and eosin (H&E), and analyzed for tumor formationusing light microscopy.

Systemic Delivery of MSCs to Ntv-a MiceBetween 6 and 10 weeks after RCAS vector injection, MSCs were

delivered into Ntv-a mice systemically either intra-arterially or intra-venously. Intra-arterial delivery was performed by injecting MSCsinto the right common carotid artery using a technique described pre-viously [16]. Mice were appropriately anesthetized (in accordance withInstitutional Animal Care andUse Committee guidelines) with ketamine/xylazine, and the common carotid artery was visualized microscopically.MSCs were trypsinized and suspended in α-MEM plus 10% FBS. Thecells were counted, and 106 MSCs suspended in 100 μl of mediumwere injected into the carotid artery using a 30-gauge needle attachedto a tuberculin syringe. Mice were monitored until awake. For intrave-nous delivery, MSCs were prepared as above, and 106 cells were injectedinto the tail vein of mice.

In Vivo Visualization of MSC EngraftmentTo visualize and track MSCs in live mice, Ntv-a MSCs were trans-

duced with an adenoviral vector containing the complementary DNAof the firefly luciferase gene (Ad-Luc; 1000 virus particles per cell) be-fore injection. Bioluminescence imaging (BLI) was used to detect theMSCs. On the day of imaging, animals were treated with luciferin(150 mg/kg, intraperitoneally) and imaged with the IVIS ImagingSystem, 200 Series (Xenogen, Alameda, CA). Luciferase in cells con-verts luciferin to oxyluciferin, generating a light signal that is detectedby a CCD camera. Bioluminescence color images were overlaid ongrayscale photographic images of the mice to allow for localizationof the light source within the animal using the Living Image version2.11 software overlay (Xenogen) and IGOR image analysis software(version 4.02A; WaveMetrics, Portland, OR). Mice were imaged 1, 4,7, and 10 days after Ad-Luc MSC injection.

Histologic Confirmation of MSC Migration to TumorHistologic verification of MSC delivery to tumor-bearing brains

was performed by labeling MSCs with a fluorescent dye and analyzingbrains after MSC injection with fluorescence microscopy. SP-DiI(Molecular Probes, Eugene, OR), a vital fluorescent dye that incorpo-rates into the cell membrane, was dissolved in dimethylformamide at21.5 mg/ml. The resulting solution was added to the culture mediumat 10 μg/ml. MSCs were incubated with 25 ml of medium with SP-DiIin 175-cm3 flasks for 48 hours. A cohort of Ntv-a mice underwentsomatic cell transfer with RCAS–PDGF-B and RCAS-IGFBP2 to gen-erate gliomas and were injected with SP-DiI–labeled MSCs (106 cells)through the carotid artery at least 6 weeks after RCAS vector injectionand presumed tumor formation. Animals were killed 1, 4, 7, and 10 daysafter MSC injection (n = 3 mice per group). Their brains were removedand sectioned in the coronal plane. Half of the brain was fixed, dehy-drated, and frozen with cryoembedding medium at −80°C. The otherhalf was fixed in formalin and stained with H&E. The sections stainedwith H&E were analyzed to determine the presence of tumor. Frozensections were analyzed by fluorescence microscopy, and five 5-μm sec-tions were obtained every 150 μm. In each of the five sections, the col-onies of SP-DiI–labeled MSCs present throughout the tumor mass at100× magnification were counted. Because it was difficult to distin-guish single cells versus groups of cells, high-signal areas were counted

718 Tropism of MSCs for Endogenous Gliomas in Mice Doucette et al. Neoplasia Vol. 13, No. 8, 2011

as one colony. The median colony count for the five sections was calcu-lated for each mouse at days 1, 4, 7, and 10 after intra-arterial injection.To determine whether MSCs were present in other organs, we per-formed intracarotid injections on mice and killed them at days 1, 4, 7,and 10 after injection (n = 3 mice at each time point). The lungs, livers,and spleens were removed and analyzed by fluorescence microscopy.

Statistical MethodsThe median colony count of SP-DiI–labeled MSCs detected in the

tumor sections at different time points was compared using the Kruskal-Wallis test. P < .05 was considered significant. The Kaplan-Meiermethod was used to estimate the time to symptomatic tumor develop-ment in mice. The log-rank test was used to compare the distributions,and P < .05 was considered significant.

Results

Isolation and Characterization of Ntv-a MSCsMSCs were harvested from the long bones of Ntv-a mice. In the

initial passages, the cells grew slowly, but by passage 7, the doublingtime was 24 to 36 hours and the cells had a spindle shape consistentwith the morphology described for human and other murine MSCs(Figure 1A) [44,45]. Successful selection of MSCs was verified by theirdifferentiation into adipocytes or mineralizing osteocytes or chondro-cytes when placed in differentiation medium (Figure 1, B-D). FACSwas performed on MSCs at passages 4, 6, 10, and 15 (Figure 2). Bypassage 6, Sca-1 and CD9 were detected on 98.0% and 99.0% of cells,respectively, and this high expression persisted so that by passage 15,Sca-1 and CD9 were detected on more than 99.6% and 99.5% of cells,

respectively. CD105 was detected on 75.0% of cells at passage 6 butonly 10.0% of cells at passage 10. At passage 6, the hematopoieticmarkers CD45 and CD11b were detected on 2.9% and 0.9% of thecells, respectively, but by passage 10, these markers were not detectedon the cells. The endothelial marker CD31 and the GSC markerCD133 were not detected at any passages. Thus, based on their mor-phology, pattern of surface marker expression, and differentiationcapacity, the cells isolated from the bone marrow of the Ntv-a micewere consistent with MSCs.

BLI Evaluation of MSCs after Intra-arterial DeliveryAnimals were analyzed by BLI 1, 4, 7 and 10 days after MSC injec-

tion. BLI signal was detected over the right frontal lobe of seven mice ateach time point, whereas no signal was seen in two mice (Figure 3, Aand B). In the seven mice in which BLI signal was detected afterMSC injection, the median BLI signal intensity at day 1 was 6.3 ×105 photons/sec per squared centimeter (range, 2.9 × 104 to 2.6 ×106 photons/sec per squared centimeter). To verify that the signal arosedirectly from the brain, brains were removed at day 4 or day 7 afterMSC injection and were reimaged. In all seven animals with positiveBLI signal after live imaging, the BLI signal was verified as originatingfrom the brain (Figure 3, A and B, right image). The median BLI in-tensity at day 4 was 5.6 × 105 photons/sec per squared centimeter(range, 2.9 × 105 to 1.5 × 107 photons/sec per squared centimeter)and that at day 7 was 2.5 × 105 photons/sec per squared centimeter(range, 2.5 × 104 to 9.8 × 105 photons/sec per squared centimeter).In the mice imaged at day 10 after injection, we did not observeany BLI signal. Two of the three mice in this group demonstratedtumor formation.

Figure 1. Differentiation of Ntv-a–derived MSCs to mesenchymal tissue subtypes. (A) MSCs in stem cell medium after six passages.(B) Bone formation fromMSCs after differentiation to osteoblastic cells after culture in an osteogenic medium. (C) Lipid droplet formationfromMSCs after differentiation to adipocytes after culture in an adipogenic medium. (D) Cartilage formation fromMSCs after differentiationto chondrocytes after culture in a chondroblastic medium.

Neoplasia Vol. 13, No. 8, 2011 Tropism of MSCs for Endogenous Gliomas in Mice Doucette et al. 719

To determine which of the nine mice harbored tumors, the brainswere removed and stained with H&E. Because somatic cell transfer ofgenes through RCAS vectors does not result in tumors in all mice, wewere able to use non–tumor-bearing mice as a control for MSC de-livery. All seven animals with positive BLI signal (i.e., MSCs were de-

tected) demonstrated tumor formation within the right frontal lobe,whereas in the brains of the two animals in which no signal was de-tected (i.e., MSCs not present), tumors were not identified (Figure 3,C and D). These results demonstrate that Ntv-a MSCs specificallylocalized to gliomas generated endogenously in Ntv-a transgenic mice.They also provide evidence that the MSCs that reside within thetumor are viable because generation of a signal requires that the cellsexpress functional luciferase.As a further control to show that Ntv-a MSCs localize specifically

to tumors, we injected a cohort of mice (n = 4) with RCAS-GFP andperformed intra-arterial injections of Ad-Luc–transfected MSCs6 weeks later. We did not visualize any BLI signal in the brains ofthese mice, and histological evaluation showed no evidence of tumor.We also evaluated the ability of intra-arterially injected MSCs to

migrate to extracranial organs. A cohort of mice (n = 9) was injectedwith Ad-Luc–transfected MSCs. At days 1, 4, 7, and 10 after injec-tion, mice (n = 3 at each time point) were analyzed for BLI signal. Themice were also killed, and the internal organs (including lungs, liver,spleen, and abdominal viscera) were again analyzed for BLI signal. Atnone of these time points did we observe BLI consistent with success-ful migration of Ntv-a MSCs to these extracranial organs.

Histologic Evaluation of Ntv-a MSCs in Endogenous GliomasAnother cohort of Ntv-a mice (n = 9) underwent somatic cell

transfer with RCAS–PDGF-B and RCAS-IGFBP2 and after 6 weeksreceived intracarotid injection of Ntv-a MSCs stained with SP-DiI.High-grade gliomas developed in all nine of the mice in this cohort.SP-DiI–labeled MSCs were detected throughout the tumors but notin nontumoral areas (Figure 4). At day 1, we identified a median of7 colonies (range, 3-18 colonies) in the five frozen sections obtainedfrom each of the three mice killed at day 1. The median number ofMSC colonies was 9 (range, 3 to 43) at day 4 and also 9 (range, 0-21)at day 7. No colonies were identified in any of the tumor-bearingbrains obtained from mice 10 days after injection. When the individ-ual time points were compared, the difference between days 1 and 4was significant (P = .02), but the difference between days 4 and 7 wasnot significant (P = .1). The difference among the colony counts at alltime points was not significant (P = .07). The extent of tumor engraft-ment by SP-DiI–labeled MSCs seemed to be associated with BLI sig-nal intensity, with both peaking at around day 4 and disappearing byday 10 (Figure 5).We also determined whether SP-DiI–labeled MSCs migrated to

extracranial organs by injecting mice and visualizing the lungs, livers,and spleens harvested from injected mice at days 1, 4, 7, and 10 (n = 3at each time point). In none of these organs did we observe fluores-cence suggesting the presence of SP-DiI–labeled MSCs.

BLI Evaluation of MSCs after Intravenous DeliveryYang et al. [46] have shown that MSCs localize to U87 xenografts

after intravenous delivery. Ntv-a MSCs transduced with Ad-Luc wereinjected into the tail vein of mice 6 weeks after they were injectedwith RCAS–PDGF-B and RCAS-IGFBP2. BLI signal was not ob-served in any of the 12 mice at any of the imaging time points (1, 4,7, and 10 days after MSC injection). Of the subgroups of animals killedon days 1, 4, 7, and 10 after MSC injection, none demonstrated BLIsignal in the brain (Figure 3, E and F). Histologic analysis of the brains(H&E) showed that 10 of the 12 mice demonstrated formation of in-tracranial tumors (Figure 3G ). The visceral organs of three of the mice

Figure 2. Flow cytometric analysis for Ntv-a–derived MSCs. FACSresults forMSCs at passages 6 and 15 for themesenchymalmarkersSca-1 and CD9, the endothelial markers CD105 and CD31, the hema-topoietic markers CD45 and CD11b, and the GSC marker CD133.

720 Tropism of MSCs for Endogenous Gliomas in Mice Doucette et al. Neoplasia Vol. 13, No. 8, 2011

were examined with BLI, and a positive signal was identified in thelungs of each of these mice (data not shown).

MSCs Derived from Ntv-a Mice Do NotEnhance TumorigenicityTo determine whether MSCs derived from Ntv-a mice could en-

hance tumor formation in the PDGF-B–dependent glioma model,we coinjected RCAS–PDGF-B and MSCs in newborn Ntv-a mice.Escalating doses of MSCs were coinjected with RCAS–PDGF-B.The median survival in the cohort of mice injected with RCAS–PDGF-B alone was 77 days (range, 23-90 days). There was no statis-tical difference in survival between the cohorts of mice coinjected with

RCAS–PDGF-B and Ntv-a–derived MSCs compared with the controlgroup injected with RCAS–PDGF-B alone. In the cohort coinjectedwith 1 × 103 MSCs, the median survival was 90 days (range, 35-90 days;log-rank test, P = .13). In the cohort coinjected with 1 × 104 MSCs, themedian survival was 64 days (range, 33-90 days; log-rank test, P = .16).In the cohort injected with 1 × 105 MSCs, the median survival was67 days (range, 41-90 days; log-rank test, P = .4).

DiscussionIn this report, we exploit the advantages of the RCAS/Ntv-a mousemodel system to show for the first time that harvested MSCs are capa-ble of localizing to endogenous high-grade gliomas after intra-arterial

Figure 3. Ad-Luc–transfected MSCs derived from Ntv-a mice track to endogenous brain tumors formed by coexpression of RCAS–PDGF-Band RCAS-IGFBP2 in Ntv-a mice. (A) Whole mouse and brain of mouse at day 1 after intra-arterial MSC injection demonstrating BLI signalover the right frontal region only. (B) Whole mouse and brain of mouse at day 1 after MSC injection. No BLI signal was observed in either thewhole mount or the harvested brain. A small, localized area of positive BLI signal was detected over the right ear, a nonspecific findingprobably indicating MSCs circulating in the external carotid artery. (C) Coronal section of the brain of the mouse in A, demonstrating tumorin the right frontal lobe of the brain (indicated by arrow). (D) Coronal section (H&E) of the mouse brain in B, without any evidence of tumorformation. (E) Whole mouse at day 1 after intravenous MSC injection, demonstrating absent BLI signal over the right frontal region. (F) Brainofmouse at day 1 after intravenousMSC injection. No BLI signal was observed in the brain. (G) Coronal section (H&E) of themouse brain in F,demonstrating significant tumor formation (arrow).

Neoplasia Vol. 13, No. 8, 2011 Tropism of MSCs for Endogenous Gliomas in Mice Doucette et al. 721

delivery. BLI confirmed that the MSCs in the tumors were viable, andhistologic studies showed that MSCs display tumor-specific tropism be-cause we did not observe MSCs outside the tumor in the normal brainin this model. The demonstration of the tumor-tropic capacity ofMSCs for endogenous gliomas that develop in their native micro-environment provides crucial support for the clinical translation ofthe strategy of using MSCs to deliver therapeutic agents to human pa-tients with malignant gliomas.Because previous studies have used engrafted tumors, the impact of

the stroma on systemically delivered MSCs, particularly from a differ-ent species, has been unclear. Our results, however, definitively dem-onstrate that bone marrow–derived MSCs can be autotransfused into

the donor and that these MSCs localize to tumors that arise endoge-nously within the natural milieu of the brain; thus, the tropism ofMSCs for gliomas is not a result of an artificial interaction betweenthe MSCs, the tumor, and the microenvironment.We used a model of high-grade glioma induced by overexpression of

PDGF-B and IGFBP2 to study tumor engraftment by MSCs. Tumorsarising from this combination of genes recapitulate the phenotypic fea-tures of high-grade glioma, particularly microvascular proliferation. Thedevelopment of abnormal blood vessels may be critical for MSCs tolocalize to gliomas, as previous work from our group has suggested thatareas of neovascularization in tumors may be necessary for MSCs togain entry into the tumor after intravascular delivery [22]. The use of

Figure 4. SP-DiI–labeled MSCs distribute throughout endogenously forming gliomas. (A) Whole mount of tumor-bearing mouse brain atday 4 after injection of SP-DiI–labeled MSCs. (B) Photomicrograph (H&E; original magnification, ×200) of glioma induced by RCAS–PDGF-B and RCAS-IGFBP2. (C) Fluorescence photomicrograph (original magnification, ×200) of tumor section, demonstrating SP-DiI–labeled MSCs (red cells). (D) Merged fluorescence photomicrograph (original magnification, ×200) showing distribution of MSCs (redcells) throughout the tumor mass (counterstained blue with DAPI). Scale bars, 100 μm.

Figure 5. Tumor engraftment by MSCs declines after day 4. (A) Median BLI signal detected in the brains of tumor-bearing Ntv-a mice atdays 1, 4, 7, and 10 after intra-arterial Ad-Luc MSC injection. (B) Median SP-DiI–labeled MSC colony count in the brains of tumor-bearingNtv-a mice at days 1, 4, 7, and 10 after intra-arterial SP-DiI–labeled MSC injection.

722 Tropism of MSCs for Endogenous Gliomas in Mice Doucette et al. Neoplasia Vol. 13, No. 8, 2011

PDGF-B to induce tumors is also of interest because we have recentlyshown that PDGF-BB may be a critical mediator of MSC tropism togliomas [20,21]. Because the RCAS/Ntv-a glioma model is dependenton PDGF-B to initiate tumor formation, it is possible that recruitmentof MSCs to the tumor was facilitated by PDGF-B expression in thetumor. Further studies are needed to decipher the specific mechanismunderlying the tropism of MSCs for gliomas. Nevertheless, the poten-tial for PDGF-B to mediate the tropism ofMSCs for gliomas is relevantto the human disease because approximately a third of gliomas are de-fined by the PDGF pathway activation [35]. Thus, the model describedherein to study the migration of MSCs to brain tumors is relevant to asignificant subset of human glioma patients.Another strength of our study is that the time between RCAS vector

injection and the introduction of MSCs was at least 6 weeks. Becausethese tumors were formed well in advance of MSC injection, it is un-likely that the MSC migration observed in this study is a consequenceof the known ability of MSCs to track to areas of injury [47]. Becauseall other studies to date have relied on intracranial injection of tumorcell lines, it has always been uncertain whether the injury induced bythe injection was at least partly responsible for the localization of MSCsto glioma. Because of the long latency in this study between injection ofthe RCAS vector and injection of the MSCs, this is the first study todemonstrate MSC migration to gliomas that is independent of tissueinjury in the brain. Thus, our study shows that the tropism of MSCsfor gliomas is a direct response to the tumor and not to brain injury.Successful harvesting of functional MSCs from Ntv-a mice is also

significant because these mice are a mixture of different genetic back-grounds and supports the translational application of this therapeuticstrategy to humans. In an extensive report describing the isolation ofMSCs from five independent strains of mice, Peister et al. [42] notedthat each strain demonstrated variable expansion in different types ofculture media. Ntv-a mice are derived from four strains: C57BL/6,BALB/C, FVB/N, and CD1. We determined that the optimal growthmedium was α-MEM, which resulted in robust cell growth. Althoughprevious studies have shown that hMSCs migrate to xenograft tumorsin mice, we are the first, to our knowledge, to show that it is possible toobtain MSCs from a mouse, expand them in culture, and deliver theseMSCs into the same mouse bearing an endogenously formed intra-cranial glioma. This strategy exactly mimics the approach that will beused in humans and, therefore, is clinically important.The intra-arterial route of delivery was required for a successful sys-

temic delivery of MSCs to brain tumors in our model. We observedBLI signal in the region of the tumor (i.e., the right frontal lobe) fromAd-Luc MSCs in 100% of mice harboring tumors. As we have dem-onstrated previously, BLI can be used to track murine MSCs in braintumors in vivo [16]. We were able to demonstrate BLI signal in tumor-bearing mice through day 7 after injection but not at day 10. We sus-pect that the cells lose their viability over time, and this observation issupported by our work showing only rare visualization of MSCs be-yond 14 days after injection of human MSCs into nude mice bearingU87 xenografts [unpublished data]. Because the decline in the MSCactivity over time has been observed in both the immunocompetentRCAS/Ntv-a and immunodeficient systems, the loss of viability appearsto be independent of the immune status of the host. This may be anadvantage to usingMSCs as delivery vehicles because they may deliver atherapeutic agent and then gradually disappear.Although a positive BLI signal verified delivery of MSCs to the re-

gion of the tumor (i.e., right frontal area), we relied on SP-DiI–labeledMSCs to confirm engraftment of MSCs throughout the tumor mass.

Mice injected with SP-DiI–labeled MSCs after RCAS vector injectionhad labeled MSCs distributed throughout the tumor, but these did notappear in non–tumor-bearing areas. The number of MSCs engraftedinto the tumor declined after day 4, consistent with our observationfrom the mice injected with Ad-Luc MSCs that these cells have de-creased viability over time. This result confirms the tumor-specific tro-pism of MSCs and their ability to insinuate themselves throughout thetumor mass. In contrast, intravenous delivery of MSCs did not result insuccessful engraftment of Ad-LucMSCs.We did detect Ad-LucMSCs inthe thoracic cavity of the animals, and we interpret this result to mean thatMSCs were trapped in the lung parenchyma. It is possible that if we hadimaged at later time points, MSCs would have been able to engraft thetumor after passing into the arterial circulation.We suspect, however, thatthis would be a much smaller population of cells compared with intra-arterial delivery and that their viability would be questionable. Otherinvestigators have had success with intravenous delivery of MSCs. For ex-ample, Yang et al. [46] recently demonstrated that hMSCs migrate tobrainstem glioma xenografts in nude mice after intravenous injection.However, in other studies using human neural stem cells to track to in-tracranial xenografts in nude mice, migration to tumors was observed afterintravenous delivery but at a much lower efficiency compared with di-rect intratumoral injection [10]. Although our group has had success withintra-arterial delivery of hMSCs, we have been unable to demonstrateeffective delivery of hMSCs through intravenous injection to intracranialxenografts in nude mice [16].The ability of MSCs to be engineered to deliver therapeutic agents

has been well described.MSCs have been engineered to secrete a varietyof therapeutic proteins, including interferon β, S-TRAIL, and others[16,22,46,48–50]. These studies have demonstrated that delivery ofthese agents through MSCs confers a survival benefit compared withcontrol animals. Virtually all of these studies have relied on xenograftmodels in immunocompromised mice, however. Limitations to usinggene-based therapies through MSC delivery remain, however, par-ticularly with respect to the successful expression of a potentially ther-apeutic gene from a short-lived MSC. Although our study was notdesigned to evaluate a therapeutic benefit fromMSC-delivered therapy,the results suggest that such a strategy would be viable in endogenoustumors in an immunocompetent host. Studies evaluating the thera-peutic benefit of MSCs in the RCAS/Ntv-a glioma model are cur-rently underway.

ConclusionsIn this study, MSCs from Ntv-a mice were successfully derived froma heterogeneous population of bone marrow cells. These cells werecarried from harvest, and characterization through injection and en-graftment of tumors generated endogenously in an immunocompe-tent host. These results lend credence to the concept of using MSCsas delivery vehicles for glial neoplasms in humans.

AcknowledgmentsThe authors thank Karen Muller and David M. Wildrick PhD, forediting the article.

References[1] Cairncross G, Berkey B, Shaw E, Jenkins R, Scheithauer B, Brachman D, Buckner

J, Fink K, Souhami L, Laperierre N, et al. (2006). Phase III trial of chemotherapyplus radiotherapy compared with radiotherapy alone for pure and mixed anaplasticoligodendroglioma: Intergroup Radiation Therapy Oncology Group Trial 9402.J Clin Oncol 24, 2707–2714.

Neoplasia Vol. 13, No. 8, 2011 Tropism of MSCs for Endogenous Gliomas in Mice Doucette et al. 723

[2] Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ,Belanger K, Brandes AA, Marosi C, Bogdahn U, et al. (2005). Radiotherapy plusconcomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 352,987–996.

[3] Lang FF, Bruner JM, Fuller GN, Aldape K, Prados MD, Chang S, Berger MS,McDermottMW, Kunwar SM, Junck LR, et al. (2003). Phase I trial of adenovirus-mediated p53 gene therapy for recurrent glioma: biological and clinical results.J Clin Oncol 21, 2508–2518.

[4] Ahmed AU, Alexiades NG, and Lesniak MS (2010). The use of neural stem cellsin cancer gene therapy: predicting the path to the clinic. Curr Opin Mol Ther12, 546–552.

[5] Studeny M, Marini FC, Champlin RE, Zompetta C, Fidler IJ, and Andreeff M(2002). Bone marrow–derived mesenchymal stem cells as vehicles for interferon-βdelivery into tumors. Cancer Res 62, 3603–3608.

[6] Studeny M, Marini FC, Dembinski JL, Zompetta C, Cabreira-Hansen M,Bekele BN, Champlin RE, and Andreeff M (2004). Mesenchymal stem cells:potential precursors for tumor stroma and targeted-delivery vehicles for anti-cancer agents. J Natl Cancer Inst 96, 1593–1603.

[7] Kosztowski T, Zaidi HA, and Quinones-Hinojosa A (2009). Applications ofneural and mesenchymal stem cells in the treatment of gliomas. Expert RevAnticancer Ther 9, 597–612.

[8] PittengerMF,MackayAM, Beck SC, Jaiswal RK,Douglas R,Mosca JD,MoormanMA, Simonetti DW, Craig S, and Marshak DR (1999). Multilineage potential ofadult human mesenchymal stem cells. Science 284, 143–147.

[9] Dominici M, Le Blanc K, Mueller I, Slaper-Cortenbach I, Marini F, Krause D,Deans R, Keating A, Prockop D, and Horwitz E (2006). Minimal criteria fordefining multipotent mesenchymal stromal cells. The International Society forCellular Therapy position statement. Cytotherapy 8, 315–317.

[10] Aboody KS, Brown A, Rainov NG, Bower KA, Liu S, Yang W, Small JE,Herrlinger U, Ourednik V, Black PM, et al. (2000). Neural stem cells displayextensive tropism for pathology in adult brain: evidence from intracranial glio-mas. Proc Natl Acad Sci USA 97, 12846–12851.

[11] Ehtesham M, Kabos P, Gutierrez MA, Chung NH, Griffith TS, Black KL, andYu JS (2002). Induction of glioblastoma apoptosis using neural stem cell–mediated delivery of tumor necrosis factor–related apoptosis-inducing ligand.Cancer Res 62, 7170–7174.

[12] Ehtesham M, Kabos P, Kabosova A, Neuman T, Black KL, and Yu JS (2002).The use of interleukin 12–secreting neural stem cells for the treatment of intra-cranial glioma. Cancer Res 62, 5657–5663.

[13] Bianco P, Riminucci M, Gronthos S, and Robey PG (2001). Bone marrow stromalstem cells: nature, biology, and potential applications. Stem Cells 19, 180–192.

[14] Caplan AI and Bruder SP (2001). Mesenchymal stem cells: building blocks formolecular medicine in the 21st century. Trends Mol Med 7, 259–264.

[15] Tocci A and Forte L (2003). Mesenchymal stem cell: use and perspectives.Hematol J 4, 92–96.

[16] Nakamizo A, Marini F, Amano T, Khan A, Studeny M, Gumin J, Chen J,Hentschel S, Vecil G, Dembinski J, et al. (2005). Human bone marrow–derivedmesenchymal stem cells in the treatment of gliomas. Cancer Res 65, 3307–3318.

[17] Nakamura K, Ito Y, Kawano Y, Kurozumi K, Kobune M, Tsuda H, Bizen A,Honmou O, Niitsu Y, and Hamada H (2004). Antitumor effect of genetically en-gineered mesenchymal stem cells in a rat glioma model.Gene Ther 11, 1155–1164.

[18] Sasportas LS, Kasmieh R,Wakimoto H, Hingtgen S, van deWater JA, MohapatraG, Figueiredo JL, Martuza RL, Weissleder R, and Shah K (2009). Assessment oftherapeutic efficacy and fate of engineered human mesenchymal stem cells forcancer therapy. Proc Natl Acad Sci USA 106, 4822–4827.

[19] Cassiede P, Dennis JE, Ma F, and Caplan AI (1996). Osteochondrogenicpotential of marrow mesenchymal progenitor cells exposed to TGF-beta 1 orPDGF-BB as assayed in vivo and in vitro. J Bone Miner Res 11, 1264–1273.

[20] Cheng P, Gao ZQ, Liu YH, and Xue YX (2009). Platelet-derived growth factorBB promotes the migration of bone marrow–derived mesenchymal stem cellstowards C6 glioma and up-regulates the expression of intracellular adhesionmolecule-1. Neurosci Lett 451, 52–56.

[21] Hata N, Shinojima N, Gumin J, Yong R, Marini F, Andreeff M, and Lang FF(2010). Platelet-derived growth factor BB mediates the tropism of human mes-enchymal stem cells for malignant gliomas. Neurosurgery 66, 144–156, discus-sion 156–157.

[22] Yong RL, Shinojima N, Fueyo J, Gumin J, Vecil GG, Marini FC, Bogler O,Andreeff M, and Lang FF (2009). Human bone marrow–derived mesenchymalstem cells for intravascular delivery of oncolytic adenovirus Delta24-RGD to hu-man gliomas. Cancer Res 69, 8932–8940.

[23] de Vries NA, Beijnen JH, and van Tellingen O (2009). High-grade gliomamouse models and their applicability for preclinical testing. Cancer Treat Rev35, 714–723.

[24] Hu X and Holland EC (2005). Applications of mouse glioma models in pre-clinical trials. Mutat Res 576, 54–65.

[25] Wakimoto H, Kesari S, Farrell CJ, Curry WT Jr., Zaupa C, Aghi M, Kuroda T,Stemmer-Rachamimov A, Shah K, Liu TC, et al. (2009). Human glioblastoma–derived cancer stem cells: establishment of invasive glioma models and treat-ment with oncolytic herpes simplex virus vectors. Cancer Res 69, 3472–3481.

[26] Garber K (2006). Realistic rodents? Debate grows over new mouse models ofcancer. J Natl Cancer Inst 98, 1176–1178.

[27] Garber K (2009). From human to mouse and back: “tumorgraft” models surgein popularity. J Natl Cancer Inst 101, 6–8.

[28] Burns TC, Verfaillie CM, and Low WC (2009). Stem cells for ischemic braininjury: a critical review. J Comp Neurol 515, 125–144.

[29] Holland EC, Hively WP, DePinho RA, and Varmus HE (1998). A constitu-tively active epidermal growth factor receptor cooperates with disruption of G1cell-cycle arrest pathways to induce glioma-like lesions in mice. Genes Dev 12,3675–3685.

[30] Becher OJ and Holland EC (2006). Genetically engineered models have advan-tages over xenografts for preclinical studies. Cancer Res 66, 3355–3358, discus-sion 3358–3359.

[31] Holland EC, Li Y, Celestino J, Dai CK, Schaefer L, Sawaya RA, and Fuller GN(2000). Astrocytes give rise to oligodendrogliomas and astrocytomas after genetransfer of polyoma virus middle T antigen in vivo. Am J Pathol 157, 1031–1037.

[32] Rao G, Pedone CA, Coffin CM, Holland EC, and Fults DW (2003). c-Mycenhances sonic hedgehog–induced medulloblastoma formation from nestin-expressing neural progenitors in mice. Neoplasia 5, 198–204.

[33] Rao G, Pedone CA, Valle LD, Reiss K, Holland EC, and Fults DW (2004).Sonic hedgehog and insulin-like growth factor signaling synergize to inducemedulloblastoma formation from nestin-expressing neural progenitors in mice.Oncogene 23, 6156–6162.

[34] Dunlap SM, Celestino J, Wang H, Jiang R, Holland EC, Fuller GN, and ZhangW (2007). Insulin-like growth factor binding protein 2 promotes glioma devel-opment and progression. Proc Natl Acad Sci USA 104, 11736–11741.

[35] Brennan C, Momota H, Hambardzumyan D, Ozawa T, Tandon A, PedrazaA, and Holland E (2009). Glioblastoma subclasses can be defined by activityamong signal transduction pathways and associated genomic alterations. PLoSOne 4, e7752.

[36] Phillips HS, Kharbanda S, Chen R, Forrest WF, Soriano RH, Wu TD, Misra A,Nigro JM, Colman H, Soroceanu L, et al. (2006). Molecular subclasses of high-grade glioma predict prognosis, delineate a pattern of disease progression, andresemble stages in neurogenesis. Cancer Cell 9, 157–173.

[37] Bleau AM, Hambardzumyan D, Ozawa T, Fomchenko EI, Huse JT, BrennanCW, and Holland EC (2009). PTEN/PI3K/Akt pathway regulates the side pop-ulation phenotype and ABCG2 activity in glioma tumor stem–like cells. CellStem Cell 4, 226–235.

[38] Charles N, Ozawa T, Squatrito M, Bleau AM, Brennan CW, HambardzumyanD, and Holland EC (2010). Perivascular nitric oxide activates notch signaling andpromotes stem-like character in PDGF-induced glioma cells. Cell Stem Cell 6,141–152.

[39] Kong LY, Wu AS, Doucette T, Wei J, Priebe W, Fuller GN, Qiao W, Sawaya R,Rao G, and Heimberger AB (2011). Intratumoral mediated immunosuppressionis prognostic in genetically engineered murine models of glioma and correlates toimmunotherapeutic responses. Clin Cancer Res 16, 5722–5733.

[40] Binning MJ, Niazi T, Pedone CA, Lal B, Eberhart CG, Kim KJ, Laterra J, andFults DW (2008). Hepatocyte growth factor and sonic Hedgehog expression incerebellar neural progenitor cells costimulate medulloblastoma initiation andgrowth. Cancer Res 68, 7838–7845.

[41] Hu X, Pandolfi PP, Li Y, Koutcher JA, Rosenblum M, and Holland EC (2005).mTOR promotes survival and astrocytic characteristics induced by Pten/AKTsignaling in glioblastoma. Neoplasia 7, 356–368.

[42] Peister A, Mellad JA, Larson BL, Hall BM, Gibson LF, and Prockop DJ (2004).Adult stem cells from bone marrow (MSCs) isolated from different strains ofinbred mice vary in surface epitopes, rates of proliferation, and differentiationpotential. Blood 103, 1662–1668.

[43] Dai C, Celestino JC, Okada Y, Louis DN, Fuller GN, and Holland EC (2001).PDGF autocrine stimulation dedifferentiates cultured astrocytes and inducesoligodendrogliomas and oligoastrocytomas from neural progenitors and astro-cytes in vivo. Genes Dev 15, 1913–1925.

724 Tropism of MSCs for Endogenous Gliomas in Mice Doucette et al. Neoplasia Vol. 13, No. 8, 2011

[44] Pittenger MF and Marshak DR (2001). Mesenchymal stem cells of humanadult bone marrow. In Stem Cell Biology. DR Marshak, RL Gardner, andD Gottlieb (Eds). Cold Spring Harbor Laboratory Press, Cold Spring Harbor,NY. pp. 349–373.

[45] Azizi SA, Stokes D, Augelli BJ, DiGirolamo C, and Prockop DJ (1998). Engraft-ment andmigration of human bonemarrow stromal cells implanted in the brains ofalbino rats—similarities to astrocyte grafts. Proc Natl Acad Sci USA 95, 3908–3913.

[46] Yang B, Wu X, Mao Y, Bao W, Gao L, Zhou P, Xie R, Zhou L, and Zhu J(2009). Dual-targeted antitumor effects against brainstem glioma by intravenousdelivery of tumor necrosis factor–related, apoptosis-inducing, ligand-engineeredhuman mesenchymal stem cells. Neurosurgery 65, 610–624, discussion 624.

[47] Karp JM and Leng Teo GS (2009). Mesenchymal stem cell homing: the devil isin the details. Cell Stem Cell 4, 206–216.

[48] Hong X, Miller C, Savant-Bhonsale S, and Kalkanis SN (2009). Antitumortreatment using interleukin-12–secreting marrow stromal cells in an invasiveglioma model. Neurosurgery 64, 1139–1146, discussion 1146–1147.

[49] Menon LG, Kelly K, Yang HW, Kim SK, Black PM, and Carroll RS (2009).Human bone marrow–derived mesenchymal stromal cells expressing S-TRAILas a cellular delivery vehicle for human glioma therapy. Stem Cells 27, 2320–2330.

[50] Sonabend AM, Ulasov IV, Tyler MA, Rivera AA, Mathis JM, and Lesniak MS(2008). Mesenchymal stem cells effectively deliver an oncolytic adenovirus tointracranial glioma. Stem Cells 26, 831–841.

Neoplasia Vol. 13, No. 8, 2011 Tropism of MSCs for Endogenous Gliomas in Mice Doucette et al. 725