The goal of this study was to combine high resolution and CE MRI at a high field to longitudinally evaluate the in vivo migration of 2 SC sources, hMSCs and fNSCs, toward and in GBM tumors. Two GBM tumor models were studied, 1 highly neoangiogenic and circumscribed (U87vIII) and 1 poorly neoangiogenic and highly invasive (GBM26), to recapitulate the range of GBM physiological characteristics.
SPIO and USPIO have been commonly used as MR labeling agents for cell tracking.32
Because of their biocompatibility, these particles have been approved by the Food and Drug Administration (FDA) for use in humans, thus facilitating the translation of cell tracking using SPIO/USPIO from preclinical studies to the clinic. Recently, MPIO have emerged as another promising MR labeling agent for tracking cell migration.33
Although these particles have not been FDA approved, biocompatible MPIO are currently being developed,34,35
and MPIO present other advantages. They enable cellular labeling with significantly more iron per unit volume than do SPIO or USPIO. As a result, the sensitivity of detection of MPIO-labeled cells using MRI is improved, and tracking of single cells labeled with MPIO has been reported in several studies.23,33,36
Furthermore, commercially available MPIO can be tagged with a fluorophore. The fluorescence enables easy detection of MPIO in histological sections and provides a quantitative measure of MPIO uptake in cells. In this study, we therefore chose to label the SC with MPIO. Using their fluorescent tag, we confirmed that both hMSCs and fNSCs were successfully labeled with the same quantity of MPIO. This point was essential to achieve a direct comparison of the 2 SC sources. We were also able to obtain histological confirmation of the presence of MPIO using Dragon Green fluorescence imaging of the MPIO-labeled cells. Finally, because of the high R2* relaxivity of the MPIO at 14.1Tesla (R2* = 1427 s−1
), both MPIO-labeled SC types were highly detectable in vivo, allowing the longitudinal evaluation of their pathotropism and biodistribution in both tumor models.
When injected in the opposite hemisphere of U87vIII tumors, MPIO-labeled hMSCs and MPIO-labeled fNSCs demonstrated a comparable tropism toward tumors. Only the U87vIII tumor model was chosen for the evaluation of pathotropism. These tumors are highly circumscribed and highly vascularized, thus making it possible to clearly assess the location, size, and boundaries of the tumor masses on CE-MR images after injection of a Gd-based contrast agent. On the other hand, such analysis would have been challenging in the GBM26 model, because this tumor model is poorly enhancing, and the location of tumor boundaries is sometimes unclear, especially when the tumor is small.
When injected in the contralateral hemisphere, MPIO-induced areas of hyposignal from hMSCs and fNSCs were found to localize around the rim of the U87vIII tumor masses starting 2 days after injection of SCs. The comparison of CE-MR images and high resolution T2*-w MR images confirmed that MPIO-induced areas of hyposignal were colocalized with the edges of the tumors (ie, in neovascularized regions of high vascular permeability). These observations are consistent with reports of SC tropism to brain tumors being induced by vascular endothelial growth factor (VEGF) and, in particular, with the report by Zhao et al. that SCs exhibit a higher tropism to the neovascularized rims of GBM xenografts.37,38
At later time points (day 7 after SC injection), MPIO-induced areas of hyposignal were also found inside the tumor masses, for both SC sources, in agreement with previous studies.10,12,37,38
Thus, our results confirm the expected high pathotropism of both hMSCs and fNSCs toward tumors and show that MPIO-labeled SCs are able to migrate from one hemisphere to another and to provide good coverage of the tumor over a 7-day period of observation.
When injected directly in the tumor masses, MPIO-labeled hMSCs and fNSCs appeared to behave in a similar fashion in the 2 tumors studied. In the circumscribed U87vIII tumor model, MPIO-induced areas of hyposignal were found to distribute both around the periphery of the tumor and inside the tumor masses, following a similar pattern to the one observed after contralateral injections. In contrast, in the slowly growing and highly invasive GBM26 tumor, MPIO-labeled hMSCs and fNSCs distributed in the central region of the tumor masses, but no clear rim of hyposignal could be detected. The difference in SC distribution observed between the 2 tumor models is likely attributable to the different physiological characteristics of these tumors. Unlike the U87vIII tumor model, the orthotopic GBM26 tumor model is highly invasive, infiltrative, and noncircumscribed. GBM26 tumors did not show any enhancement after injection of a Gd-based contrast agent, indicating vessel co-option behavior and low VEGF levels.39
The absence of areas with high VEGF is in line with the absence of an MPIO-induced rim of hyposignal on the T2*-w images of GBM26 tumors.
Although T2*-w imaging for cell tracking is commonly used, the absolute quantification of the number of MPIO-labeled cells in a region of interest on T2*-w images is extremely challenging.35,40
The size of the signal void in the T2*-w images is much larger than the size of the MPIO-labeled cells. This increases the sensitivity of detection of MPIO-labeled cells but makes it impossible to estimate how many cells are responsible for each area of hyposignal observed. Alternative MRI sequences producing positive rather than negative contrast, such as SWIFT,41
have recently been developed to overcome these quantification issues, and we are in the process of implementing this approach for future studies. In this study, an approach similar to the one reported recently by Song et al. to quantify SPIO-induced areas of hyposignal on T2*-w images was implemented.35
However, rather than performing a full histogram analysis of the number of voxels associated with each T2*-value, our approach consisted of performing a histogram analysis of the number of voxels associated with each SNR value in a brain slice and then calculating the variance, mean, and ID (variance-to-mean ratio) as a parameter reflecting both the spatial distribution of labeled cells in the slice and the overall amount of labeled cells. Although analysis of the index of dispersion does not provide an absolute quantification of the number of MPIO-labeled cells inside the tumor region, it can serve as a tool to perform a relative comparison of both SC sources in each tumor model. This analysis revealed that both MPIO-labeled fNSCs and hMSCs were highly spatially distributed throughout the tumor masses in both tumor models. An increase in the ID was observed for both SC sources and for both tumor types, demonstrating that both SC types distribute throughout the tumor masses during the period of observation. Of interest, in each tumor model, no statistically significant differences in the values of the ID were found between SC types at any time point, suggesting that both SC types reach a comparable coverage of the tumor mass.
Results from the quantitative analysis of the ID were confirmed by IHC and IF analysis. Quantification of the PB-positive SCs showed that both SC types distribute inside the tumor mass after intratumoral injection in U87VIII tumors. Furthermore, the spatial distribution of both SC types was not significantly different: both MPIO-labeled SCs distributed inside and at the edges of the tumor and were also found further away in the brain parenchyma. These results confirmed the expected pathotropism of both SC sources to tumors and are in line with the analysis of the ID.
This study presents for the first time, to our knowledge, a direct comparison of pathotropism and biodistribution of MPIO-labeled hMSCs and fNSCs in 2 GBM tumor models. Our results indicate that both SC sources exhibit strong pathotropism toward tumor and distribute comparably throughout the tumor masses, thus presenting comparably good candidates for the development of SC-based therapies.
Conflict of interest statement. None declared.