In the current work we demonstrate a procedure to develop therapeutic strategies against a human U87 glioblastomas and an effective therapy model in SCID mice. The procedure is based on the use of hAMSCs genetically modified to express the herpes Simplex tTK as vehicles to deliver bystander toxicity to tumors. To analyze the behavior of tumor and therapeutic cells, we labeled the former with a CMV promoter regulated Pluc-eGFP chimerical reporter, and the latter with a different trifunctional chimerical reporter comprising Rluc-RFP and tTK activities 
. This approach allowed us to monitor the location and number of tumor and therapeutic cells in live mice during treatment with GCV, and optimize cell and prodrug treatment.
We established that both, tTK activity in genetically modified hAMSCs, and administration of GCV were both required for bystander killing of tumor cells co-cultivated with hAMSCs.
In vivo BLI of predetermined numbers of hAMSC and U87 cells, stereotactically implanted at a specific site of the mouse brain, allowed us to correlate the number of photon events recorded by BLI with that of implanted cells, for their quantification during therapy.
An optimum therapeutic ratio of 4 hAMSCs to 1 U87 cell was empirically determined by implanting predetermined proportions of both cell types in mouse brains and administering GCV i.p. after a 6 day period, allowed for tumor establishment. Analysis of hAMSC behavior during the 6 day pre-treatment period indicated that a large proportion of the therapeutic cells disappeared, while tumor cells proliferated. This lead to a 50 fold, reduction in the proportion of therapeutic to tumor cells, that was 1
13 by the time the GCV treatment began. This result not only emphasizes the effectiveness of hAMSCs as therapeutic vehicles, but also suggests that procedures or scaffold materials for their protection would likely improve even more the therapeutic outcome. The disappearance of a large proportion of hAMSCs when injected in immunosupressed mice has been previously observed 
. However, the reduction in hAMSC number should not be attributed to CMV promoter silencing. In a previous study by Vilalta et al. 
, hAMSCs expressing Pluc regulated by the constitutively active SV40 promoter also lost between 75–90% of cells within the first 10 days post inoculation in SCID mice. Moreover, since in our cell differentiation experiment all the therapeutic cells are initially red fluorescent and become also green fluorescent upon differentiation to the endothelial lineage, silencing of the CMV promoter should result in the appearance of green fluorescent cells that show no red fluorescence. Thus CMV silencing can be evaluated by determining the proportion of therapeutic cells expressing GFP (regulated by the inducible PECAM/CD31 promoter) that do not express RFP (regulated by the CMV promoter) in histological sections of tumors. While extensive silencing of the CMV promoter would be required to produce the 92 fold increase in PECAM/CD31-PLuc expression observed in the BLI experiments, our assessment found less than 10% of GFP positive cells that were not RFP positive. This result excludes CMV promoter silencing, and pointing to cell death or diffusion from the tumor sites as the likely causes for the drop in RLuc production following cell implantation.
The therapeutic effectiveness of hAMSCs was evaluated by comparing their capacity to inhibit U87 tumor growth by implanting the optimal 4
1 (hAMSC to U87) proportion in mice brains, and treating with GCV or PBS. Such experiments showed that treatment with GCV of tumors containing hAMSCs reduced the number of tumor cells by a factor of 104
, relative to tumors without hAMSCs, also treated with GCV, or tumors with hAMSCs treated with PBS.
For the same tumor, by the end of the experiment, treatment with hAMSCs plus GCV had reduced tumor cell number to 0,12% of that at T
0. However, in tumors without hAMSCs but also treated with GCV tumor cell number increased by 319 fold in the same period.
Macroscopic and laser confocal microscope examination of brain slices at the end of experiments corroborated BLI imaging results. While brains from mice with untreated mixed-cell tumors had large tumor masses containing green and red fluorescent cells, no macroscopic tumors and few green fluorescent cells were found in the brains of mice implanted with tumors plus therapeutic cells and treated with GCV.
In experiments better modeling a clinical situation, we show that implantation of therapeutic hAMSCs on preestablished glioblastoma U87 tumors and treatment with GCV also results in tumor growth inhibition that lasts as long as there were surviving therapeutic hAMSCs. Moreover, repeated inoculations of therapeutic hAMSCs resulted in a progressive reduction of tumor size and a significant extension of mice survival, relative to untreated controls.
Insight on the therapeutic mechanism mediated by hAMSCs was gained using a double label strategy that allows monitoring of changes in gene expression on a “per cell" basis. This was achieved by labeling of the same therapeutic cell that already expressed Rluc-RFP reporter constitutively, with a different Pluc-eGFP reporter regulated by the inducible human PECAM promoter. By co-implanting these cells with unlabelled U87 tumors and monitoring by BLI we could observe that within the first 7 days post implantation, a large proportion of the hAMSCs disappeared from the tumor site, either by death or dispersion. However, a subpopulation of the cells that expressed PECAM-promoter regulated luciferase remained in close association with the tumor. Moreover, in the tumor associated cell population, the ratio of PECAM-regulated to CMV-regulated luciferase activity increased with time reaching by day 7 a value 92 fold higher than that at implantation time. This results indicated that tumor associated hAMSCs were actively differentiating to the endothelial lineage. Independent analysis by fluorescence laser confocal microscopy of tumor slices revealed RFP and eGFP expressing hAMSCs with endothelial morphology in close association with tumor vascular structures. This latter result was further corroborated by positive staining of microvessel associated red fluorescent hAMSCs with an anti human PECAM antibody.
Association of bone marrow derived MSCs with vascular structures in tumors has been reported previously 
. However, the vascular associated cells were shown to express pericyte specific antigens while PECAM/CD31 is considered an endothelial specific marker. Both findings are not mutually exclusive, and in the absence of additional data, could be reconciled by considering that differentiation of mesenchymal stromal cells may depend on their tissue of origin as well as on the specific tumor environment in which they are implanted 
. Our results suggest that GCV induced suicide of hAMSCs not only kills neighboring tumor cells by bystander effect but also by eliminating the associated vascular system that supplies oxygen and nutrients.
In conclusion, we demonstrate a general and versatile strategy to develop cell based tumor therapies and evaluate their effectiveness based on the use of chimeric bioluminescent and fluorescent reporters introduced in tumor and therapeutic cells, followed by analysis of their behavior by non invasive real time BLI and confocal microscopy.
With this strategy we were able to show that tTK expressing hAMSCs are very effective vehicles to deliver localized cytotoxic therapy to U87 glioblastomas and that hAMSCs within tumors differentiate to endothelial lineage cells that associate with vascular structures. Implantation of genetically modified hAMSCs may provide an effective procedure to eliminate residual tumor cells in surgical borders after tumor removal.
In future research we will use the above strategy to improve the preclinical significance of the therapy model. We will focus on evaluating procedures for therapeutic cell delivery, including the use of a biomaterial to protect and control the delivery of therapeutic cells implanted in tumors. To evaluate the relevance of the immune environment on the therapeutic effectiveness, we will use immune-competent mice implanted with singeneic glioma tumor model that will be treated with autologous therapeutic hAMSCs. Finally, we will seek to validate the generality of the therapeutic procedure using glioma tumors from human biopsies implanted in immunedefficient mice.