Recent studies have raised appealing possibilities of replacing damaged or lost neural cells by transplanting in vitro-expanded neural precursor cells (NPCs) and/or their progeny. Magnetic resonance (MR) tracking of superparamagnetic iron oxide (SPIO)-labeled cells is a noninvasive technique to track transplanted cells in longitudinal studies on living animals. Murine NPCs and human mesenchymal or hematopoietic stem cells can be efficiently labeled by SPIOs. However, the validation of SPIO-based protocols to label human neural precursor cells (hNPCs) has not been extensively addressed. Here, we report the development and validation of optimized protocols using two SPIOs (Sinerem and Endorem) to label human hNPCs that display bona fide stem cell features in vitro. A careful titration of both SPIOs was required to set the conditions resulting in efficient cell labeling without impairment of cell survival, proliferation, self-renewal, and multipotency. In vivo magnetic resonance imaging (MRI) combined with histology and confocal microscopy indicated that low numbers (5 × 103 to 1 × 104) of viable SPIO-labeled hNPCs could be efficiently detected in the short term after transplantation in the adult murine brain and could be tracked for at least 1 month in longitudinal studies. By using this approach, we also clarified the impact of donor cell death to the MR signal. This study describes a simple protocol to label NPCs of human origin using SPIOs at optimized low dosages and demonstrates the feasibility of noninvasive imaging of labeled cells after transplantation in the brain; it also evidentiates potential limitations of the technique that have to be considered, particularly in the perspective of neural cell-based clinical applications.
Disclosure of potential conflicts of interest is found at the end of this article.
The hypothesis that stem/precursor cell populations found in the mammalian central nervous system (CNS) could contribute to neural plasticity/repair has opened new and exciting areas of research in the fields of basic cell biology and regenerative medicine. The recognition that neural precursor cells (NPCs) displaying bona fide stem cell functional features could be propagated in long-term cultures [1, –3] and that, upon transplantation in the host brain, they could reintegrate appropriately and stably express foreign genes made them attractive for gene and cell therapy [4, 5] for local  and even global [7, 8] pathologies. Interestingly, transplanted NPCs home specifically in the areas of tissue damage, exerting beneficial effects either directly or through bystander effects [7, 9, , –12]. This functional plasticity is shared by human neural precursor cell (hNPC) lines that have been established by our group and others [13, , , –17].
Standard histopathological methods are currently used to evaluate engraftment, migration, and differentiation of exogenous transplanted NPCs or to track endogenous precursors in physiological conditions  or after mobilization and recruitment to site of damage [19, , –22]. Design and improvement of NPC-based therapies will be favored by the development of sensitive, noninvasive techniques for tracking donor cells to be used for brain repair. Recently, stem and progenitor cells of neural and non-neural origin have been magnetically labeled using dextran-coated superparamagnetic iron oxide (SPIO) particles to track them by magnetic resonance imaging (MRI) in cell therapy approaches for CNS disorders [23, , , , –28].
Although SPIO particles are easily internalized by macrophages, uptake by nonphagocytic and slow-dividing cells is poor , so that high (and often toxic) concentrations are needed for efficient cell labeling. Different modifications and coating of SPIO particles through linking to lectins , HIV-tat peptide [31, 32], internalizing monoclonal antibodies [33, 34], or dendrimers [35, 36] have been described to improve the internalization of the contrast agent, with indubitable advantages in terms of labeling efficiency but also with potential disadvantages in terms of technical complexity, poor availability, species specificity, and biosafety, issues that are critical for the potential clinical translation of these labeling procedures.
A convenient method to improve SPIO internalization has been developed using polycationic transfection agents (TA). Through electrostatic interactions, these macromolecules form complexes with SPIOs that are more easily transferred in the cells by endocytosis and pinocytosis [37, 38]. Among those TA, high molecular weight poly-l-lysine (PLL) has become the molecule of choice in most studies [29, 39], and different protocols have been established to label immortalized precursor cells [40, 41], oligodendrocyte progenitors, Schwann cells and olfactory ensheathing cells (OECs) [37, 42, 43], neural progenitors/stem cells [24, 44, 45], and embryonic stem cells  of rodent origin. Similar approaches have been translated to human-derived mesenchymal stem cells (MSCs) and hematopoietic stem cells (HSCs) [46, , –49], endothelial cells , and neural cells . However, a detailed documentation of protocol development to label hNPCs with SPIOs has not been extensively addressed.
Here, we report the development and validation of an efficient in vitro protocol to label hNPCs with two different human-grade SPIOs (Sinerem and Endorem). We optimized both the concentration of SPIOs and the incubation time, achieving robust labeling of hNPCs in vitro without impairment of their functional properties. In vivo MRI combined with histology and confocal microscopy indicated that low numbers (5 × 103 to 1 × 104) of viable SPIO-labeled hNPCs could be efficiently tracked for at least 1 month in longitudinal studies. By using this approach, we also clarified the impact of donor cell death on the magnetic resonance (MR) signal. This study describes a simple protocol to label NPCs of human origin using SPIOs at optimized low dosages and demonstrates the feasibility of noninvasive imaging of labeled cells after transplantation in the brain; it also evidentiates potential limitations of the technique that have to be considered, particularly in the perspective of neural cell-based clinical applications.
Multipotential hNPC lines were originally established from the diencephalon of a 10.5-week-postconception human brain, as previously described . Details are available in the supplemental online data.
Labeling of hNPCs with SPIOs
Two different magnetic resonance contrast agents were compared: (a) Sinerem (Ferumoxtran-10; a generous gift of C. Corot, Guerbet, Roissy CdG Cedex France, http://www.guerbet.com; particle size, 20–40 nm; stock solution, 20 mg Fe/ml from lyophilized powder), phase III clinical trial completed ; and (b) Endorem (AMI-25; Guerbet; particle size, 80–150 nm; stock solution, 11.2 mg Fe/ml), U.S. Food and Drug Administration-approved. Both SPIOs were used alone or in the presence of PLL hydrobromide (molecular weight, 388 kDa; P-1524; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com; a stock solution at 0.5 mg/ml was prepared in distilled water and kept frozen at −20°C until required). Details are available in the supplemental online data.
Staining of hNPCs with Prussian Blue
After incubation with SPIOs, the Prussian blue (PB) method (Perls' acid ferrocyanide) was used for the detection of iron within the cell cultures. This induces a reduction of ferric iron to the ferrous state with formation of a blue precipitate. Details are available in the supplemental online data.
Assessment of Cell Viability, Growth Rate, Multipotency, and Apoptosis
Cell viability, growth rate, and multipotency were evaluated as previously described [53, 54]. Apoptosis was evaluated in SPIO-labeled cells and in untreated controls by quantifying the number of cells expressing the activated form of caspase-3 or the number of positive cells in a terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assay. Details are available in the supplemental online data.
After treatments with SPIO, cells were replated in SPIO-free growth medium. After 7–10 days, neurospheres formed by SPIO-labeled and unlabeled control cells were processed for electron microscopy analysis. Details are available in the supplemental online data.
Transplantation of SPIO-Labeled hNPCs
After the incubation in SPIO-containing medium, cells were washed twice in phosphate-buffered saline (PBS), and viable cells were counted by trypan blue exclusion using a hemocytometer. Cells were resuspended in the appropriate volume of ice-cold PBS containing 0.1% DNase to produce the final concentration (from 7.5 × 103 to 5 × 104 SPIO-labeled or unlabeled cells in 2 μl). In one experiment, the final cell suspension was exposed to four cycles of freezing (dry ice) and thawing to kill cells. Cell suspension was slowly injected (0.2 μl/minute) unilaterally or bilaterally in the striatum or corpus callosum of adult CD1 or SCID mice using a 5-μl Hamilton syringe mounted on a stereotactic apparatus (David Kopf Instruments, Tujunga, CA, http://www.kopfinstruments.com). Details are available in the supplemental online data.
Histology and Immunofluorescence
After the last time point of analysis, animals were killed by perfusion via the descending aorta under deep Avertin (2,2,2-Tribromoethanol, 500 mg/kg, Sigma-Aldrich) anesthesia with 0.9% NaCl followed by 4% paraformaldehyde in PBS. Brains were equilibrated for 24 hours in 30% sucrose in PBS and were either quick-frozen in optimal cutting temperature compound or cut at the vibratome. Serial coronal cryostatic (20 μm-thick) or vibratom (40 μm-thick) sections were processed for PB staining, and adjacent sections were processed for immunofluorescence. Details are available in the supplemental online data.
MRI and Relaxometry Studies on SPIO-Labeled hNPCs and Isolated Brains
Images of cells and isolated brains were acquired using a Biospec Tomograph (Bruker Medizintechnik GmbH, Karlsruhe, Germany, http://www.bruker.com) equipped with a 4.7-tesla (T) horizontal magnet. Phantom relaxometry studies were performed on a 1.5-T human MR scanner with a circular surface coil of 23 mm in diameter. Details are available in the supplemental online data.
In Vivo MRI
Details are available in the supplemental online data.
Two-tailed unpaired t test and one-way analysis of variance followed by Tukey's multiple comparison test were used to analyze the data. Details are available in the supplemental online data.
Incorporation of SPIOs by hNPCs Is Dose- and Time-Dependent
Three days after the last subculturing passage, hNPCs (passages 25–27) were incubated in growth medium in the presence of increasing concentrations of SPIO particles (up to 800 and 100 μg Fe/ml for Sinerem and Endorem, respectively) and for different incubation periods (4, 24, 48, and 72 hours), with or without the polycationic TA PLL.
We observed a linear correlation between Sinerem uptake, as assessed by quantifying the number of cells that displayed intracellular iron storage (visualized by PB staining) and both the incubation time and the iron concentration in the culture medium (Fig. 1A, 1C). Blue cytoplasmatic inclusions with typical perinuclear localization were present only in cells incubated with the SPIO (Fig. 1C). At the lowest doses and shorter incubation times, we could recognize different intensities of staining among the cell population. As the concentration of Sinerem and the incubation time increased, the amount of intracellular iron increased, resulting in heavy labeling in almost all the cells.
Treatment with doses up to 800 μg Fe/ml and incubation times up to 72 hours, which resulted in 100% labeling efficiency (Fig. 1A), did not significantly affect hNPC survival, as evaluated by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium (MTT) assay (Fig. 1B). Moreover, labeled cells proliferated normally (as demonstrated by time-dependent increase in MTT absorbance) and formed new neurospheres (not shown). This suggested that the levels of intracellular iron accumulation resulting from the labeling procedure did not induce acute cell toxicity.
To test whether the iron load might result in delayed toxicity, neurospheres formed from hNPCs exposed to the different labeling treatments were washed free of the iron-containing medium, dissociated to single cells, and plated in standard growth conditions to monitor the formation of new neurospheres. We observed that hNPCs exposed to the highest dose (800 μg Fe/ml) and the longest incubation times (48–72 hours) displayed a severe decrease in cell survival (>80% cell death, as assessed by trypan blue incorporation) and a reduced neurosphere-forming capacity (Fig. 1D). Thus, for subsequent experiments, we chose to use Sinerem at a dose of 400 μg Fe/ml for a 48-hour incubation time, conditions that allowed efficient labeling (80% of PB-labeled cells) without delayed impairment of cell survival, self-renewal, and proliferation capacity.
It has previously been reported that cationic TA might improve the efficiency of SPIO intracellular uptake. When hNPCs were incubated in the simultaneous presence of Sinerem and PLL we observed, at the highest dose, increased numbers of detectably labeled cells (Fig. 2A) and of cells displaying high density of blue cytoplasmic inclusions (Fig. 2B), reflecting higher levels of intracellular iron accumulation. The presence of PLL did not impair cell survival, which, surprisingly, increased at the highest doses (Fig. 2C).
To compare the efficiency of different SPIOs in hNPC labeling, we repeated the dose-response experiments (with or without PLL) using Endorem, a contrast agent currently used in clinical MRI (Fig. 2D–2F). Our results showed that 10-fold lower doses of Endorem and shorter incubation times were required, compared with Sinerem, to obtain high levels of iron accumulation, resulting in efficient labeling and detection in vitro. We observed a linear correlation between the number of PB-labeled cells and Endorem doses in the range 12.5–50 μg Fe/ml, then cells reached saturation (Fig. 2D). Incubation of hNPCs in the presence of 25 μg Fe/ml of Endorem for 24 hours resulted in intense PB staining of more than 80% of the cells, with no short-term impairment in cell survival (Fig. 2F) and neurosphere-forming capacity (not shown). The combination of PLL and Endorem at any dose tested produced only a modest increase in the absolute number of PB-labeled cells (Fig. 2D) but resulted in a striking increase of the intracellular blue granule density (Fig. 2E, Prussian blue). The higher amounts of iron deposits were clearly visualized by electron microscope analysis as multiple dark electron-dense cytoplasmic granular inclusions (Fig. 2E, electron microscopy). Iron deposits were not seen in untreated control cells (Fig. 2E, untreated [UT]). On the basis of these results, we decided to use Endorem at 25 μg Fe/ml for a 24-hour incubation time.
Our data allowed us to define labeling protocols based on optimized SPIO doses and incubation times that did not affect short-term hNPC survival/proliferation (as assessed by MTT assay) and self-renewal (as assessed by the formation of new neurospheres derived from SPIO-labeled cells). To exclude that treatment of hNPCs with these optimized SPIO dosages might result in sublethal damages that could manifest when cells are exposed to hostile conditions, we measured apoptotic activity in Sinerem-and Endorem-treated hNPCs 2 and 7 days after plating them in the absence of growth factors, a condition that is known to trigger either cell death or differentiation. We found only small fractions of caspase-3-positive and TUNEL-positive cells (1%–3%) in cultures analyzed at both time points (Fig. 3A, 3B) without significant differences in the numbers of apoptotic cells in SPIO-labeled cells with respect to unlabeled controls.
Finally, we assessed that the optimized labeling protocols did not affect multipotency of hNPCs. We could not observe significant differences in the proportions of neurons (TUJ1+) and astrocytes (glial fibrillary acidic protein [GFAP]+) in Sinerem-treated hNPCs compared with untreated controls in differentiated cultures obtained by removal of mitogens and addition of 2% fetal calf serum (FCS) (10.00 ± 6.90 and 17.66 ± 6.64 TUJ1+ cells; 65.83 ± 7.83 and 62.00 ± 12.09 GFAP+ cells; p = .2415 and .7788 for neurons and astrocytes, respectively; two-tailed unpaired t test; n = 5 replicates in two independent experiments; similar results were obtained using Endorem-labeled cells [not shown]). Oligodendrocytes represented only a minor fraction (<2%) in FCS-treated cultures (not shown). To increase the number of oligodendrocytes in the differentiated cultures and to assess the potential impact of SPIO labeling also on this cell type, we applied an alternative differentiation protocol in which FCS was replaced by insulin-like growth factor 1. In this way, we obtained 6%–7% of GalCer+ oligodendrocytes, 30%–50% of TUJ1+ neurons, and 50%–70% of GFAP+ astrocytes. Prussian blue staining revealed that the majority of the cells in hNPC differentiated cultures retained SPIO labeling (Fig. 3C, representative picture of Sinerem-treated and UT differentiated cells). No significant differences were observed in the proportions and in the morphology of neurons, astrocytes, and oligodendroglial cells in SPIO-treated cultures with respect to untreated controls (Fig. 3C). Our results indicated that labeling hNPCs with optimized SPIO dosages did not sensitize them to apoptotic stimuli or affect their multipotency.
Persistence of SPIO Labeling over Time
We tested the potential long-term effect of SPIO labeling on the proliferation ability of hNPCs by serially subculturing SPIO-labeled hNPCs (treatment: 400 μg Fe/ml for 48 hours and 25 μg Fe/ml for 24 hours, for Sinerem and Endorem, respectively) and monitoring the growth rate (Fig. 4A, 4B) and the persistence of PB staining over time (Fig. 4C, 4D). Over the first 3–4 subculturing passages postincubation (approximately 20–25 days in culture), we observed slight or no changes in the growth rate of Sinerem-labeled (Fig. 4A), and Endorem-labeled (Fig. 4B) hNPCs; this trend was unchanged when hNPCs were treated with SPIO plus PLL (not shown). At each subculturing passage postincubation, we plated a fraction of hNPCs and evaluated the percentage of cells retaining PB labeling. The progressive drop in the number of PB-labeled cells indicated that both intracellular Sinerem (Fig. 4C) and Endorem (Fig. 4D) were progressively diluted over time, likely because of cell proliferation. The temporal profile of this decrease correlated to the dose and to the incubation time used for hNPC labeling: the higher the dose and incubation time (resulting in higher intracellular SPIO accumulation), the slower the decrease of SPIO labeling (Fig. 4C, Sinerem). In all conditions tested, after the third subculturing passage the number of Prussian blue-labeled cells was decreased by more than 80% with respect to the number of labeled cells present soon after the incubation with SPIOs.
In Vitro MRI of SPIO-Labeled hNPCs
To test the detectability in MRI, increasing numbers of hNPCs (from 12.5 × 103 to 1 × 105) labeled with Sinerem (400 μg Fe/ml for 48 hours) were immobilized in gelatin (4% wt/wt). T2*-weighted images of cross-sections through the gelatin-containing tube revealed the presence of a hypointense signal (Fig. 5A). A good correlation between the number of labeled cells and the number of hypointense pixels in the MR images was observed (Fig. 5B; in each section, pixels were considered hypointense when their signal was below 50% of the maximum signal). Similar results were obtained using increasing numbers of Endorem-labeled hNPCs (not shown). Unlabeled cells were not detectable in MR images (not shown). These results showed that MRI can efficiently detect low numbers of SPIO-labeled hNPCs and that the intensity of the signal is proportional to the number of labeled cells.
Relaxometry studies revealed higher R2 values in Sinerem-labeled cells with respect to Endorem-labeled cells, and within each group, in SPIO plus PLL-treated cells with respect to SPIO minus PLL-treated cells (Fig. 5C), confirming the results obtained with PB staining and electron microscopy.
SPIO-Labeled hNPCs Can Be Efficiently Detected by MRI After Transplantation into the Mouse Brain: Short-Term Analysis
Having established a protocol allowing efficient labeling of hNPCs (in terms of number of labeled cells and intensity of labeling) using low dosages of SPIO without overall short-term or delayed toxicity and functional cell impairment, we sought to address whether SPIO-labeled cells could be localized by MRI after transplantation into the adult brain. A set of mice receiving transplants of different numbers of Sinerem-labeled or unlabeled hNPCs were killed 24 hours after injection, and the localization of SPIO-labeled cells was determined by Prussian blue staining on brain sections. MRI on isolated brains showed typical regions of hypointense signal attributable to SPIO-labeled cells (Fig. 5D), as further confirmed by PB staining on corresponding serial rostrocaudal sections cut from the same brains after the MRI procedure (Fig. 5D, Prussian blue 1× and 10×). Low numbers of Sinerem-labeled hNPCs (Fig. 5D, right [R], 1.5 × 104 cells) were detected with high efficiency. As predicted by in vitro experiments, the signal intensity was proportional to the number of injected cells and increased when cells were labeled in the presence of PLL (Fig. 5D, left [L], 1.5 × 104 cells; R, 3 × 104 cells). No MRI signal or PB staining were detected when similar numbers of unlabeled cells were injected (Fig. 5D, L, 1.5 × 104 cells). Similar results were obtained from isolated brains of mice transplanted with Endorem-labeled cells (not shown).
Transplanted SPIO-Labeled hNPCs Can Be Efficiently Tracked over Time: A Longitudinal Study
Having established an efficient protocol to localize SPIO-labeled hNPCs by MRI after transplantation in the brain, we next sought to verify whether transplanted cells could be localized over time, in longitudinal studies using anesthetized animals that were subsequently allowed to recover after each imaging session. hNPCs were labeled with Endorem (25 μg Fe/ml × 24 hours) or Sinerem (400 μg Fe/ml × 48 hours) plus or minus PLL. Adult immunodeficient SCID mice were stereotactically implanted unilaterally or bilaterally with different numbers of SPIO-labeled cells (plus or minus PLL), untreated control cells, or vehicle alone. MR images were taken 1, 7–9, and 28–30 days postinjection. Representative images of mice transplanted bilaterally with the same number of Endorem-labeled cells (plus or minus PLL) are shown in Figure 6. Similar results were obtained using Sinerem-labeled cells (not shown). Twenty-four hours after implantation, cells were detectable at the injection sites as hypointense spots in T2 images. The MR signal remained stable after 7 and even after 28 days postinjection (Fig. 6A). This confirmed that hNPCs transplanted in non-neurogenic areas of the normal adult brain integrate in the host parenchyma without extensive migration from the site of injection. Only when cells were injected in the corpus callosum or reached this area as a consequence of needle retraction could we observe SPIO-labeled cells migrating along the white matter tract, with no apparent loss of MR signal (Fig. 6A, 28 days, arrows).
After the last imaging procedure, animals were killed to determine the abundance and distribution of SPIO-labeled cells within the brain. PB labeling was present around the site of injections in all animals that received Endorem-labeled cells, in close correspondence to hypointense regions shown in MRI (Fig. 6B, PB 1× and 10×). Because of the T2 sequences used (thickness, 0.75 mm; gap, 0.075 mm), two or even three serial 20-μm-thick coronal sections contained PB labeling matching the hypointense regions (distance between the first sections in two consecutive series, 100 μm) (Fig. 6B, series 1 and 2). The identity of SPIO-labeled hNPCs was confirmed by immunofluorescence analysis using an antibody that specifically recognizes a human nuclear protein (anti-human nuclear protein [hNu]; Fig. 6B).
In the majority (16 of 23; >70%) of mice transplanted with different numbers of SPIO-labeled hNPCs (7.5 × 103 to 5 × 104 cells), we found excellent correspondence among MRI signal, PB staining, and human nuclei immunoreactivity performed on adjacent sections. This suggested that the hypointense regions revealed by MRI were closely related to the presence of SPIO-labeled transplanted cells. However, we wanted to investigate in detail the potential contribution of leaked or released SPIO from donor cells into host cells. Thus, we assessed the presence of cells of the macrophagic-microglial lineage (by CD68, F4/80, and Iba-1 staining) and of reactive astroglia (by GFAP staining) within and around the site of transplant of mice injected with SPIO-labeled hNPCs, unlabeled hNPCs, or vehicle and analyzed 9 and 30 days after the transplant, comparing the distribution of these cell types with hNu immunoreactivity, PB staining, and MRI signal. Results of this qualitative analysis are available in supplemental online Figures 1, 2, and 4.
At both time points, we occasionally observed activated ameboid Iba-1+ microglia around the donor cells compared with Iba-1 immunoreactivity in areas far from the graft, 2F), but never in close correspondence with hNu immunoreactivity (supplemental online Figs. 1B, 2D′, 2E′), which, on the contrary, mostly overlapped with PB staining and MR signal (supplemental online Figs. 1A, 2A, 2B). We observed moderate astrogliosis (supplemental online Fig. 2C′; compare with GFAP immunoreactivity in areas far from the graft [Fig. 2C″]); however, few or no CD68+ macrophages or F4/80+ microglia were detected at the site of transplant (supplemental online Fig. 2B). Notably, a modest recruitment of activated microglia around the site of injection was also detected in mice injected with both unlabeled cells (not shown) and with vehicle alone (supplemental online Fig. 2E, 2E′). We did not observe hNu+ cells expressing markers of macrophages or microglia. Clusters of hNu+ donor cells were detected 28 days after injection of 2.5 × 103 unlabeled control hNPCs (supplemental online Fig. 3D) but were not detected in MRI (supplemental online Fig. 3A) or after PB staining of brain tissue (supplemental online Fig. 3B).
In 7 of 23 mice transplanted with SPIO-labeled hNPCs (five of seven mice transplanted with Sinerem-labeled cells), we could detect few if any hNu+ donor cells at the site of injection, at both 7 and 30 days after transplant (supplemental online Fig. 4B). In the same region, we observed CD68+ and F4/80+ cells, likely representing SPIO-laden macrophages (as assessed by PB staining; supplemental online Fig. 4A), that accounted for the hypointense signal observed in MRI (supplemental online Fig. 4A). Overlapping between hypointense MR signal and PB staining was also observed 7 days after transplanting dead SPIO-labeled hNPCs. Also, in this case, CD68+ macrophages were infiltrating the area in which dead cells were present (supplemental online Fig. 4C). These results indicated that dead SPIO-donor cells can directly (likely by releasing iron in the host tissue) and indirectly (by recruiting host-derived phagocytic cells) contribute to the MR signal, even at late time points (30 days) after cell transplantation.
This study describes the development of a simple in vitro protocol to label NPCs of human origin using SPIOs at optimized low dosages that allows maintaining stable functional properties without acute or delayed toxicity. Most important, it demonstrates the feasibility of reliable, noninvasive imaging of labeled cells after transplantation in the brain. It also evidentiates and discusses potential limitations of the technique that are important to be aware of in the perspective of future neural cell-based clinical applications.
We tested two different dextran-coated iron oxide particles: Endorem, which is currently clinically used in MRI for metastatic lesion detection in liver, and Sinerem (Ferumoxtran-10), which presents smaller particles and longer blood resistance time  with respect to Endorem and has completed phase III clinical development for metastatic detection in lymph nodes [52, 56, 57]. Our study shows a dose- and time-dependent uptake of both SPIOs in hNPCs, in line with previous studies on several rodent (OECs, Schwann cells, and oligodendrocytes)  and human non-neural cell types (macrophages and HSCs) . In our labeling protocol, 10-fold lower doses of Endorem and shorter incubation times were required, compared with Sinerem, to obtain efficient labeling and MR detection of hNPCs. Similar differences in uptake efficiency have been described in non-neural human cells [38, 55]. These are likely dependent on different mechanisms of SPIO uptake because of their different particle sizes (20–40 and 80–150 nm for Sinerem and Endorem, respectively) [37, 38]. Also, we observed a higher sensitivity of hNPCs to SPIO uptake with respect to murine NSCs (5-fold lower doses to obtain >80% efficiency of labeling with both Sinerem and Endorem; ; A. Gritti, personal observations). All these data indicated that the issues of particle size, cell type, and species specificity should be carefully taken into account when developing preclinical strategies relying on SPIO-based cell tracking techniques.
In the perspective of using SPIO-labeled cells for brain repair, it is mandatory to establish protocols yielding high efficiency of cell labeling (in terms of both number of labeled cells and intensity of labeling) while preserving the biological features of the labeled cells. Our data showed that neither survival nor differentiation potential was severely impaired in hNPCs after the end of the incubation treatment, even in treatments resulting in 100% labeling efficiency. However, this short-term analysis was not sufficient to exclude a potential impairment in self-renewal and multipotency, the main functional features that identifies NSCs. Our hNPC lines  have been isolated by means of a peculiar culture system that supports the proliferation of a small subset of cells (including both primary precursors and progenitor cells) that respond to selected mitogens forming clonal aggregates called primary neurospheres . These primary clones can be long-term subcultured, maintaining stable proliferation and multipotency over time. Most important, the formation of new neurospheres at each subculturing passage returns an index of the self-renewal capacity of the original stem cell population . Interestingly, we observed important cell death and impairment in neurosphere-forming ability in the heavy-labeled hNPCs soon after the first postlabeling subculturing passage, suggesting that iron overload might be responsible for delayed toxicity and functional impairment.
Iron is an essential element in cellular metabolism, and its accumulation produces enhanced cell proliferation [62, , –65]. Indeed, an iron-induced increase in cell proliferation might explain the unexpected higher numbers of hNPCs observed in the short-term analysis with the highest doses of SPIO and for the longest incubation times (Fig. 2C, 2F). However, intracellular free iron overload eventually results in functional impairment and cell death . In our culture conditions, this delayed toxicity was forced to manifest as a result of cell harvesting, dissociation, and replating of SPIO-labeled hNPCs in our stringent culture conditions. Since similar challenging conditions might be found after transplantation in a host environment, careful titration and optimization of SPIO labeling doses are critical issues to be addressed in the perspective of using these agents to label NPCs or other cell types for therapeutic approaches.
We could overcome this problem by reducing both the SPIO concentration and the incubation time. Doses of 25 and 400 μg Fe/ml and incubation times of 24 and 48 hours for Endorem and Sinerem, respectively, resulted in efficient (>80%) cell labeling, without impairment of cell survival, sphere-forming ability, and multipotency. Due to the rapid dilution of SPIO labeling as a result of cell proliferation (described below), measurement of these functional parameters on hNPCs labeled with optimized SPIO doses is significant only if performed at the first postlabeling subculturing passage, when at least 50%–75% of the cells still retain cytoplasmic SPIOs.
Incubation of iron nanoparticles with PLL prior to cell treatment resulted in a moderate increase in the numbers of labeled cells but in significantly more efficient accumulation and detection of SPIO particles in the cytoplasm, with no increase in toxicity. Unlike what was described for other human non-neural (MSCs, HSCs, and endothelial progenitor cells) and neural cell types [46, 50, 51, 67], the presence of a transfection agent was not an absolute requirement to obtain excellent and reliable levels of hNPC labeling and detectability in vitro, even when using relatively low doses of SPIOs. This was further confirmed by the low number of SPIO-labeled cells that could be efficiently detected by MRI and by the good correlation between the numbers of labeled cells and intensity of MRI signal.
Intracellular SPIOs were progressively diluted and almost completely cleared from hNPCs after three subculturing passages. Loss of SPIO labeling due to cell proliferation has previously been reported for different cell types and different SPIO-based protocols [55, 68, 69]. This might hamper a reliable long-term cell tracking in the perspective of using MRI in neural cell-based approaches . An advantage of using NPCs could be that following transplantation in the CNS [4, 5, 11, 45], the majority of them either differentiate in neurons and glial cells [5, 71] or maintain features of immature, slow-cycling cells [11, 12], thus increasing the chances to obtain a stable MR signal over time. The development of MR reporter genes could allow the overcoming of potential limitations due to labeling methods or to different cell types used .
Our in vitro data defined optimal conditions to label hNPC using Sinerem and Endorem, resulting in robust SPIO uptake and excellent detectability by MRI. These data were supported by short-term MRI studies performed either on isolated brains or in living animals, demonstrating that low numbers of SPIO-labeled hNPCs (<1.5 × 104 cells) could be efficiently detected short term following transplantation. Also, longitudinal MRI studies demonstrated that they could be tracked noninvasively for as long as 1 month after transplantation, further supporting the efficiency of the labeling protocol.
It has been demonstrated that SPIO-labeled OECs and neural progenitors maintain a physiological differentiation program upon transplantation in pathological contexts [37, 51]. We transplanted hNPCs in the brain parenchyma of healthy adult animals; therefore, we did not expect them to migrate far from the site of injection or to undergo extensive differentiation into neuronal/glial cells. The majority of the cells transplanted in the striatum were found clustered at the site of injection 30 days after transplantation, although scattered cells were observed in the adjacent parenchyma and in periventricular areas (supplemental online Fig. 5). A detailed assessment of the fate potential of the transplanted cells was beyond the scope of our study. Additional studies using animal models of neurodegenerative diseases will be needed to confirm that the cell differentiation program is not compromised in transplanted SPIO-labeled hNPCs.
The high efficiency of in vitro cell labeling was consistent with the good R2 values obtained in relaxometry studies and allowed achieving excellent detection of less than 1 × 104 transplanted SPIO-labeled hNPCs up to 1 month after injection. Most important, the presence of SPIO/PLL complexes could improve cell labeling but was not necessary to achieve robust MR signal, which, in addition, remained stable over time.
In optimized conditions allowing excellent cell integration and survival, iron leaking was probably low, and the contribution of iron-loaded host cells (i.e., macrophages and microglia) to MR signal was not relevant. This resulted in optimal correlation between the MR signal, the extent and localization of PB staining, and the presence of grafted hNPCs, as detected by anti-human specific antibody, at least up to 1 month after injection. However, in the presence of massive cell death of SPIO-labeled donor cells, which occurred in approximately 25% of the injected animals, iron-laden macrophages/microglia and likely dead cells themselves  could indeed contribute to the MR signal as early as 7–9 days after injection (supplemental online Fig. 4). Interestingly, cell death was predominant in mice injected with Sinerem-labeled cells. Our studies showed that Sinerem displays higher R2 values (as measured by in vitro relaxometry) with respect to Endorem at the optimized labeling concentrations, and this accounts for the excellent detection of Sinerem-labeled cells in MRI, even without PLL. Our results seemed to exclude apoptotic activation in SPIO-labeled cells after growth factor deprivation. However, we should consider the possibility that this in vitro model was not stringent enough to trigger cell death and that optimized dosages in vitro might still result in cell toxicity when cells are exposed to a really challenging in vivo environment. Notably, only a few caspase-3+ cells were present in the dead grafts even at early time points (9 days) after hNPC injection (not shown), suggesting that cell necrosis, rather than apoptosis, was the prevalent mechanism of cell death. These observations indicate that careful titration of this compounds in stress models in vitro is absolutely necessary and has to be further validated in in vivo settings.
A loss of MR signal during the first 35 days after transplantation of dead SPIO-labeled human neural cells in the striatum of unlesioned animals has been reported . In this study, the macrophage/microglial involvement was not carefully assessed. Although we did not perform in vivo relaxometry and quantification of MR signal, we did not observe a detectable loss of signal in a 30-day time window after transplanting viable cell. Long-term monitoring (at least 8–12 weeks) would be needed to verify whether complete clearance of dead cells and released iron ultimately results in definitive loss of MR signal.
In summary, these studies describe a simple protocol for efficient SPIO labeling of hNPCs and demonstrate the feasibility of noninvasive imaging of low numbers of SPIO-labeled cells after transplantation in the brain; however, they also point out potential limitations of the technique that have to be carefully considered, particularly in the perspective of a clinical context. Combining MRI with other imaging techniques that could evaluate some functional/metabolic parameters (e.g., positron emission tomography) and development of pertinent hardware and software tools allowing integration of in vivo imaging results with high-resolution ex vivo imaging techniques (such as histology and immunohistochemistry) could greatly improve the reliability and efficiency of both clinical and small-animal imaging approaches.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
We thank Dr. C. Panzeri for electron microscopy images; Drs. M. Bacigaluppi, S. Pluchino, and L. Muzio for sharing reagents, protocols, and ideas; and Drs. B.D. Brown and R. Galli for critically reviewing the manuscript. We thank Dr. Corot (Guerbet) for providing Sinerem. This work was supported in part by Fondazione Italiana Sclerosi Multipla (FISM 2002/R/33), by the Italian Ministry of Health (RF2003), by the Italian Telethon Foundation (grant TGT06B02) (to A.G.) and by Fondazione Cariplo “bando chiuso 2005” (to G.S.). M.N., C.M., C.C., and A.G. are currently affiliated with the San Raffaele Scientific Institute, San Raffaele Telethon Institute for Gene Therapy, Milano, Italy.