Monitoring ferumoxide-labelled neural progenitor cells and lesion evolution by magnetic resonance imaging in a model

Efficacy of neural stem/progenitor cell (NPC) therapies after cerebral ischaemia could be better evaluated by monitoring   migration and in vivo distribution of cells post-engraftment in parallel with analysis of lesion volume and functional recovery. Magnetic resonance imaging (MRI) is ideally placed to achieve this, but still poses several challenges. We show that combining the ferumoxide MRI contrast agent Endorem with protamine sulphate (FePro) improves iron oxide uptake in cells compared to Endorem alone and is non-toxic. Hence FePro complex is a better contrast agent than Endorem for monitoring NPCs. FePro complex-labelled NPCs proliferated and differentiated normally  , and upon grafting into the in vitro brain 48 hours post-ischaemia they were detected   by MRI. Imaging in vivo over four weeks showed the development of a confounding endogenous hypointense contrast evolution at later timepoints within the lesioned tissue. This was at least partly due to accumulation within the lesion of macrophages and endogenous iron. Neither significant NPC migration, assessed by MRI and histologically, nor a reduction in the ischaemic lesion volume was observed in NPC-grafted brains.  Crucially, while MRI provides reliable information on engrafted cell location early after an ischaemic insult, pathophysiological changes to ischaemic lesions can interfere with cellular imaging at later timepoints.


Introduction
Stroke remains a common cause of mortality and disability worldwide, with few effective treatments available. Cerebral ischaemia, caused by the reduction of blood flow in a cerebral artery, is the most common type of stroke, accounting for 80% of strokes 1 . Cellular therapy is an emerging and promising avenue for treatment of cerebral ischaemia. Several experimental studies have demonstrated that treatment with neural stem cells after an ischaemic event improves neurological score and behavioural recovery, and can reduce infarct volume [2][3][4][5] . These findings have led to a small number of clinical trials of cellular therapy in stroke patients, with modest effects on neurological outcome and motor skill recovery [6][7][8][9] . It is clear that further optimisation of cell type and cell delivery, and a greater understanding of the behaviour of engrafted cells in vivo is required for improved therapeutic outcome.
Previous research has demonstrated that fetal telencephalic neural progenitor cells engrafted into the ischaemic brain are capable of migration towards the infarct, and of improving behavioural recovery 10 . A small percentage of engrafted neural stem/progenitor cells (NPCs) have been shown to differentiate into neurons in the peri-infarct area and integrate into the neural network [11][12][13][14][15][16][17][18] . Developmentally, NPCs give rise to the cortical neuron layers in the normal developing embryo, and may have the appropriate migratory capacity and neurogenic potential to give rise to neurons after engraftment into host injury tissue.
In vivo monitoring of engrafted cells is an important tool that could be developed to correlate cell behaviour and distribution with clinical outcome. Cellular imaging with MRI contrast agents can be used to monitor the distribution of transplanted cells in models of cerebral ischaemia. Hoehn et al. demonstrated that mouse embryonic stem (ES) cells could be labelled with an iron oxide-based agent 19 and subsequently migrate into the infarcted area from the contralateral hemisphere. Cell migration and functional improvement has been observed in rat models of stroke after intracisternal injection of superparamagnetic iron oxide (SPIO)-labelled neurospheres 20 . Engrafted cells labelled with SPIO have been shown to survive and differentiate into neural lineages in the ischaemic striatum 21 .
Further development of contrast agents with low toxicity, high labelling efficiency and retention is required for improved detectability and effective long-term cell tracking in vivo.
The ferumoxide Endorem is a dextran-coated superparamagnetic iron oxide agent that can form a complex with the polycationic protamine sulphate -the FePro complex [22][23][24] . Previous studies have suggested that the FePro complex has little effect on cell viability and behaviour 23-25 .
Therefore we wished to assess the effect of FePro on NPC behaviour, and establish whether labelling and in vivo imaging of NPCs could be enhanced by using this complex. We studied the effect of FePro labelling on NPC metabolism, proliferation and differentiation capacity. FePro-labelled NPCs were engrafted into the ipsilateral parenchyma at 48 hours post-ischaemia and monitored over 4 weeks using MRI. Assessment of NPC proliferation and differentiation capacity demonstrated no difference in the behaviour of FePro-labelled NPC compared to unlabelled NPC. In vivo, we observed the evolution of hypointense T 2 contrast in the ischaemic striatum over several weeks in both control and FePro-NPC engrafted animals.

Materials and methods
Neural stem cell cultures For initial cell labelling assays and quantification of iron, the ST14A rat neural progenitor cell (NPC) line was used, and cultured as previously described (generous gift from E. Cattaneo) 26 . For subsequent labelling, viability, proliferation and differentiation assays, and for the in vivo study, primary embryonic forebrain-derived NPCs were used.
Cell labelling with contrast agent For neurosphere labelling, neurospheres were dissociated at 6 days after passage, by incubation in 1X trypsin with DNase for 7 minutes at 37°C. Cells were washed in trypsin inhibitor and EBSS, and then incubated with contrast agent (Endorem or the FePro complex) in proliferation medium for 24 hours to allow contrast agent internalisation. Cells were washed in EBSS, and dissociated for replating or use in intracerebral microinjections.
For adherent ST14A cell labelling, cells were incubated with contrast agent in culture medium for 24 hours to allow contrast agent internalisation, washed three times with EBSS, then passaged.
FePro complex: 2 mg/ml protamine sulphate was incubated with Endorem in a ratio of 9:5 for 10 minutes at room temperature to form the complex (FePro), making a final concentration of 100 µg/ml iron and 10 µg/ml protamine sulphate in the culture medium.

Cellular viability assays
Cells were incubated with Endorem, FePro, or no contrast agent (control) for 24 hours. Cells were washed and replated into 96 well plates at a density of 10 4 cells per well, 8 wells per treatment group, for cell metabolic activity assays as a measure of cell viability. The MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide; Sigma) and Alamar Blue (Invitrogen) assays were performed as previously described 27,28 . For the Trypan Blue assay, cells were resuspended 10:1 in 0.5% Trypan Blue (Sigma), and transferred to a haemocytometer for counting. Under a light microscope (Zeiss), non-viable blue-stained cells and total viable cells were counted. For each assay, two samples were counted for each treatment group, and the assay was repeated three times.

Cellular proliferation assays
The neurosphere-forming assays measured the number of neurospheres generated per 1000 cells and the average neurosphere diameter per sample, which are indicators of the number of stem and progenitor cells in the sample and their rate of proliferation, respectively. Cells were plated at a density of 10 4 cells per ml in neurosphere proliferation medium into 96 well plates, where each well contained 100 µl of proliferation medium and 10 3 cells. After 5 days in culture, the number of neurospheres generated per 10 3 cells was counted, and neurosphere diameter was measured using the imaging analysis software, ImageJ developed by U. S. National Institutes of Health, Bethesda, Maryland, USA (http://imagej.nih.gov/ij/, 1997-2012). For each assay, a minimum of 8 wells per treatment group was counted, and each assay was repeated twice.

Prussian blue stain for iron in neurospheres
NPCs cultured at a density of 10 4 cells per ml for 7 days were transferred to polyornithine-coated coverslips in 24 well plates for one hour to allow neurosphere adherence to coverslips. NPC were then fixed in 4% paraformaldehyde (PFA) in PBS for 10 minutes, incubated in 6% hydrochloric acid and 4% potassium ferrocyanide for 20 minutes, then washed in PBS and mounted.
Measurement of iron using a superconducting quantum interference device (SQUID) ST14A cells were labelled with Endorem or FePro using the labelling methods described above. Labelled cells were then washed and fixed in 4% PFA. Cells were transferred to gelatin capsules (size 0, Capsugel) containing cotton wool, at a density of 1.25 × 10 6 per gel cap. Iron content of samples was measured using a superconducting quantum interference device (SQUID) and the amount of iron per cell was calculated. Sample magnetization was measured over a range of magnetic field strengths at 300K (Kelvin; room temperature) and 10K, and hysteresis loops were plotted. To calculate the amount of iron in each sample, remnant magnetization and saturation magnetization of the samples were compared to the magnetization of a sample with Endorem alone, which had a known iron content against which cell magnetisation was calibrated.

Middle cerebral artery (MCA) occlusion and intracerebral microinjections
All procedures were carried out under the Animals Scientific Procedures Act 1986 (project license: PPL 70/05617). Eleven wild type male Sprague-Dawley rats (250 g, 8 weeks old, Charles Rivers, UK; n=11) were anaesthetised with 2% isofluorane. Temperature was monitored with a rectal probe. The middle cerebral artery occlusion was performed as previously described 32 , to occlude the MCA for 30 minutes. Briefly, the right common carotid artery was exposed. A sterile 4-0 suture with an epoxy resin tip (Araldite ® , Huntsman Advanced Materials) was inserted into the carotid artery and advanced 17 mm into the brain to occlude the MCA for 30 minutes. Following occlusion, the MCA was reperfused by slowly removing the suture, and animals were then recovered on a heated mat. Animal weight was monitored. Following MCAO surgery, animals were assigned a number and alternately allocated to control or treatment group. After surgery, animals were housed individually, provided with soft tissue bedding and had unrestricted access to water and softened food pellets.
In six animals, FePro-labelled NPCs (FePro-NPCs) were injected into the ipsilateral corpus callosum 48 hours following cerebral ischaemia. Animals were anaesthetised with 2% isofluorane, and secured in a stereotactic frame (Kopf, Germany). Animals were injected with 1.25 × 10 5 FePro-labelled cells in a maximum of 3 µl, as described above, at 0.2 l/min into the ipsilateral corpus callosum. The coordinates used were AP +1.0, ML -0.3; DV -0.2 to Bregma. Five animals received no intracerebral injection (control uninjected group).
Animals were imaged to determine whether injected cells could be identified and monitored over time with MRI, relative to the control uninjected group.

MRI
For in vitro MRI, contrast agent-labelled cells were washed and fixed in 4% PFA. Cells were transferred to 250 µl Eppendorf tubes and centrifuged to produce a cell pellet. Eppendorfs were placed in a custom-made probe and imaged using a 2.35T horizontal bore magnet (Oxford Instruments, UK) interfaced to an SMIS console (Guildford, UK). A 2D spin echo sequence was used with the following parameters: TR=1500 ms; TE=80 ms; FOV 25 mm; slice thickness 1 mm.
For in vivo MRI, animals were anaesthetised with 2% isofluorane, and secured on a stereotactic probe. Animals were imaged using the above 2.35T system. A 2D gradient echo (T 2 *-weighted) sequence was used with the following parameters: TR=500 ms; TE=30 ms; FOV 30 mm; 128 × 128 voxels; 11 slices; 1 mm slice thickness; and 16 averages. A 2D spin echo (T 2 -weighted) sequence was also used: TR=1500 ms; TE=120 ms; FOV 30 mm; 128 × 128 voxels; 7 slices; slice thickness 1 mm; 16 averages. Animals were imaged at 3, 7, 14, 21, and 28 days post MCA occlusion surgery. For lesion volume measurements, hyperintense lesion areas from T 2 -weighted images were segmented manually using ImageJ software available at http://rsb.info.nih.gov/ij (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD). Lesion areas per slice were used to determine the total brain lesion volume, taking into account the thickness of the 2D slices (total brain lesions covered a mean of 7.33 (1.44 SD) slices in FePro-NPC animals and 7.50 (1.27 SD) slices in control animals). The lesion volume of the central slice covering the central MCA territory was measured and compared between groups. Homologous regions of interest were defined in the contralateral hemisphere.

Histology
After the final MRI time point, animals were euthanized by transcardial perfusion with 0.9% saline then perfusion fixed with 4% PFA. Brains were removed from the skull and transferred to 30% sucrose in PBS for 48 hours, then frozen and stored at -80°C for tissue processing. Coronal brain sections of 30 µm thickness were taken on a cryostat (Bright Instruments, UK). For Prussian blue staining, sections were incubated in 6% hydrochloric acid and 4% potassium ferrocyanide for 20 minutes, then washed in PBS. Brain sections were counterstained with nuclear fast red or hematoxylin-eosin, and mounted with DPX (distyrene, plasticizer, xylene) mountant.

Neurosphere migration assay
The MCA occlusion surgery was performed and animals were imaged at 24 hours post-ischaemia to visualize the infarcted area. Four animals underwent MCA occlusion surgery and were euthanised using Schedule 1 method at 24 hours (n=2) or 48 hours (n=2) postischaemia. The area of infarct, as determined by MRI at 24 hours post-ischaemia, was isolated. The contralateral hemisphere was used as non-ischaemic control tissue. Tissue was homogenized and filtered through a 0.4 µm sterile filter. Protein content of the homogenate was calculated against an albumin standard using the Pierce BCA Protein Assay Kit.
Single neurospheres were transferred into polyornithine-coated wells of 96 well plates, in 100 µl of culture medium per well. Individual neurospheres were photographed on a Zeiss AxioVert inverted microscope with a Hamamatsu digital camera and using OpenLab imaging software, then tissue homogenate or growth factors were added to the control medium. Treatment groups were as follows: 1400, 700, 300 µg/ml MCAO homogenate from 24 hours time point; 1400, 700, 300 µg/ml MCAO homogenate from 48 hours time point. Control groups were as follows: DMEM:F12 with N2 medium; DMEM:F12 with control tissue homogenate. Neurospheres were photographed 72 hours following treatment and the area covered by the cells measured to quantify the extent of migration. Assays were repeated three times, with 8 wells per treatment group.

Data analysis
Data are expressed as mean +/-standard error. Data analysis was performed using ImageJ for MRI image analysis and neurosphere diameter measurements; Adobe Photoshop for microscopy image analysis, and SPSS software for statistical analysis. A one-way ANOVA was used for statistical comparisons, with P < 0.05 defined as significant. A two-way mixed ANOVA was used for statistical analysis of the change in lesion volume over time in control and FePro-NPC-treated animals.

Neurosphere response to cerebral ischaemia
We assessed whether the NPCs expanded as neurospheres from embryonic mouse forebrain were able to respond to cues from cerebral ischaemia. We investigated their migratory response to protein extracted from cerebral ischaemia infarct regions at two different timepoints -24 and 48 hours post-ischaemia and from the corresponding contralateral control (Figure 1). The protein extracts were added directly to the wells containing the individual neurospheres; hence migratory cues were evenly distributed.
As shown in Figure 1A-E, extensive migration in all directions had occurred by 72 hours from neurospheres exposed to ischaemia protein extracts ( Figure 1A-B) but not in the presence of control proteins ( Figure 1C-D). Figure 1E demonstrates the migratory response of NPC to 24 hr and 48 hours ischaemic tissue protein extracts at different protein concentrations. Migration in both 24 hours and 48 hours post-ischaemia groups was significantly different from control migration (F=6.680; p<0.001). Post hoc analysis showed that at 24 hr, the 1400 µg/ml concentration, and at 48 hours the 300 and 700 µg/ml MCAO concentrations, were significantly different from control non-ischaemic tissue protein extract. Hence NPCs are able to respond to cerebral ischaemia cues in vitro.
Rat neurosphere response to cerebral ischaemia

FePro cell labelling and viability
The effects of iron oxide-based MRI contrast agents on cell viability were first assessed in a fetal neural stem cell line and on primary NPCs. Incubation of ST14A cells with Endorem or FePro for 24 hours did not affect cell viability as assessed using metabolic activity assays (Supplementary Figure 1). Iron content was measured using a SQUID. Mean iron uptake was 3.93 pg Fe/cell for Endorem-labelled cells, and 14.5 pg Fe/cell for FePro-labelled cells. Therefore the iron content was 3.7-fold greater in FeProlabelled cells than in Endorem-labelled cells.
We then investigated the effect of cell labelling with FePro and Endorem on NPC viability and proliferative capacity (Figure 2A-D). Neither contrast agent affected NPC viability as shown by the Trypan Blue exclusion assay and Alamar blue assay (Figure 2A-B). Furthermore, neither neurosphere-forming ability nor neurosphere growth were negatively affected by FePro or Endorem ( Figure 2C-D). Serial MRI of cerebral ischaemia following FePro-NPC injection In vivo MRI of injected FePro-labelled NPC was carried out 3, 7, 14, 21 and 28 days after cerebral ischaemia, to determine whether cells could be identified and monitored over time, relative to a control MCAO group without NPCs (Figure 4). The T 2 -weighted imag- The iron within NPC neurospheres was visualized after 7 days in culture by Prussian blue. As shown in Figure 2E-F, the contrast agent is retained within cells across cell divisions.
To further characterise the effects of Endorem and FePro, the differentiation capacity of contrast agent-labelled NPCs was assessed. Figure 3 shows that FePro and Endorem-labelled NPCs can differentiate into neurons, oligodendrocytes, and astrocytes ( Figure 3A   es revealed extensive heterogeneous ischaemic lesions in both groups, with hypointense regions developing in the striatum and middle cortical layers, from day 7 and persisting until day 28 ( Figure 4A-B). This lesion heterogeneity was present in all control animals as well as in the FePro-NPC group. The FePro-labelled cells at the injection site were clearly detected in T 2 *-weighted images ( Figure 4C-D).
We compared T 2 -and T 2 *-weighted images at 28 days in the FePro-NPC and control group. Figure 5A-D shows the profile of a FePro-NPC-treated animal and control animal at the 28 day time period. Lesion heterogeneity was observed in all animals in both T 2 and T 2 *-weighted images. Signal intensity in hypointense ischaemic striatum regions was significantly different from the intensity in corresponding contralateral regions ( Figure 5E; F=3.154, p=0.05).
We investigated the origin of the hypointense signal on T 2 -weighted MRI by examining histological sections. Iron can be a source of T 2 -weighted hypointensity, and we assessed its distribution in both groups. In both groups, iron was detected in the ipsilateral (ischaemic) striatum at the ischaemic border, and in the ischaemic cortex in some animals (2 of 5 control, and 3 of 6 FePro-NPC animals) ( Figure 6C-F). Iron accumulation was detected in the mid-striatum at the lesion border, which approximates the area of hypointensity on MRI. Additionally in the FePro-NSC group, iron was detected at the injection site ( Figure 6G). No iron was detected in the contralateral cortex or striatum in the FePro-NSC group ( Figure 6H).
We investigated whether the MRI signal hypointensity was associated with the distribution of macrophage/microglia in the brain.    control, unlabelled cells, and that labelled NPC can differentiate into neurons, oligodendrocytes and astrocytes. The FePro label is retained throughout multiple cell divisions during the formation of new neurospheres. In addition, data from the ST14A cell line (Supplementary Figure 1) also confirm that FePro does not affect cell viability, and has greater labelling efficiency than Endorem. Imaging NPC in vitro demonstrated that FePro uptake under the described labelling conditions was sufficient to generate MRI contrast. SQUID measurement of iron oxide uptake also confirmed and quantified cellular uptake of FePro and suggest that FePro may be a superior MRI contrast agent for NPC labelling compared with Endorem 24 . Our results indicate that there are no cytotoxic effects associated with FePro labelling, which is consistent with other studies that have similarly shown limited cytotoxicity of FePro in other cell types, and that FePro uptake is sufficient to generate contrast on T 2 *-weighted images 20,21,33,34 . Taken together with our previous work demonstrating that endogenous neural stem cells could be labelled with FePro and migrate from the subventricular zone (SVZ) to the olfactory bulb 22 , NPCs are ideally suited for labelling with FePro for in vivo detection of cell migration.

Evolution of MRI T 2 contrast and heterogeneity interferes with NPC detection in ischaemic brain
This study demonstrated that cellular imaging was in part complicated by the evolution of T 2 and T 2 * heterogeneity in the ischaemic lesion in all animals, obscuring cell detection on MRI at later timepoints. The evolution of T 2 -weighted hypointensity could be related to the accumulation of macrophages and granulocytes to the lesion over time, leading to localized hypercellularity. Accumulation of MHC Class II-expressing cells (such as macrophages, dendritic and B cells) was shown to be spatially coincident with regions of T 2 -weighted hypointensity and reduced ADC (apparent diffusion coefficient) in a model of multiple sclerosis 35 . Hypercellularity was shown to increase from one week, peak at 3 weeks and persist in later timepoints, and regions of T 2 -weighted hypointensity develop by 3 weeks, which is consistent with our results 35 . In another study, We observed accumulated macrophage/microglia in the infarcted cortex and striatum in both FePro-NSC and control groups, and some of them co-labelled with iron ( Figure 6E-F). Iron-positive cells were a mixture of macrophage and non-macrophage cell types. No macrophage/microglia were observed in the contralateral hemispheres ( Figure 6G-H).
Finally to investigate the effect of cell injections on the outcome of cerebral ischaemia, MRI analysis was carried out. Changes in volume of the lesion at the level of the central MCA region and the volume of the whole lesion over time were compared between control ischaemic and FePro-NPC groups. No statistically significant difference in lesion volume over time was observed either for the whole lesion or at the central MCA territory (Figure 7A-B; p=0.275 and p=0.244, respectively). The change in hemisphere volume at the MCA territory was also assessed. We observed that the ipsilateral hemisphere volume was decreased by 28 days post-ischaemia relative to the contralateral hemisphere in both groups (F=17.18; p<0.005; Figure 7C), but we did not observe a difference between FePro-NPC and control groups. This reduction in ipsilateral hemisphere volume over time may represent secondary, delayed neuronal death in the ischaemic hemisphere, lesion compaction, brain reorganisation or a combination of these factors.

Discussion
The main findings of this study were that i) FePro displays a high labelling efficiency and does not affect NPC behaviour in vitro; ii) following brain ischemia, the evolution of endogenous MRI T 2 -weighted contrast and heterogeneity can interfere with NPC detection and iii) soluble cues from injured tissue promote NPC migration in vitro, but no discernible NPC migration is observed in vivo following an extensive stroke.
FePro displays high labelling efficiency and does not affect NPC behaviour We observed that NPC can be labelled with FePro with no significant effect on cell viability, metabolic activity or proliferation relative to  (p=0.577). B, Total lesion volume. There is no significant difference between control and FePro-NPC lesion volumes (p=0.921). C, Hemisphere volume change. Change in ipsilateral:contralateral hemisphere volume ratio between 7 and 28 days (*:ANOVA, p<0.0001), but there is no significant difference between control and FePro-NPC in hemisphere volume. T 2 hypointensity was observed from 4 days post-ischaemia at the lesion border, which interfered with detection of SPIO-labelled injected cells, and accumulation of ferric iron was detected histologically from 6 days post-ischaemia around the lesion 36 . Although we did not observe haemorrhage in any of the animals, the T 2 heterogeneity may also arise as a consequence of the accumulation of iron in the lesion, through the development of microhaemorrhages, the degradation of blood products, or neuronal degeneration.
In the FePro-NSC group, an additional factor is clearly the presence of iron oxide particles in the FePro contrast agent. Palweczyk et al. have shown that iron exchange can occur between labelled cells and macrophages in vitro, although the transfer of iron to macrophages is low 37 . Weber et al. observed the evolution of hypointensity in ischaemic striatum on T 2 *-weighted MRI at 2 and 10 weeks postischaemia in a MCA occlusion model, and found that T 2 * hypointense regions colocalised with iron-containing macrophages 38 . The authors attributed the iron in macrophages to the phagocytosis of red blood cells from damaged blood vessels. Danielisova et al. have observed iron deposition in the striatum and pyramidal cells of cortical layers III and V following 20 minute MCA occlusion 39 . Justicia et al. have reported iron accumulation, T 2 -weighted hypointensity and delayed neuronal death in the thalamus between 3 and 7 weeks following cerebral ischaemia 40 and suggested that iron accumulation was mediated by heme oxygenase-1-positive (HO-1) microglial activity. These studies suggest that iron accumulation at lesion sites may be linked to microglial activity and neurodegeneration. Whereas our previous study has shown in normal brains a good correlation between MRI and FePro-labelled endogenous neural progenitor cells 24 , tracking FePro-labelled cells in the ischaemic brain is clearly more challenging. This study highlights the need to characterise the long-term profile of cerebral ischaemia, and points to confounding image contrast at later timepoints that must be overcome in order for iron oxide-based cellular imaging to be a viable method of cell tracking.

Soluble cues from injured tissue promote NPC migration in vitro
The NPC demonstrated a migratory response to cerebral ischaemia cues in vitro. These results indicate that fetal NPC are responsive to cerebral ischaemia factors, and the migratory cue is likely to be a soluble extracellular factor(s). In contrast, we did not observe any major cell migration from the injection site to the ischaemic striatum or cortex. It is possible that the soluble cues promoted cell migration in vitro because of the absence of inhibitory or toxic molecules present in the in vivo environment. It is also conceivable that the FePro-labelled NPCs were phagocytosed by macrophages in vivo 41 . The apparent lack of accumulation of NPCs at the infarcted site correlates with the lack of a significant reduction in infarct size. These findings are consistent with a recent report showing that reduction in infarct size is "dose-dependent" and that neural stem cell injection is effective in reducing neural damage only following a moderate infarct 42 . On the other hand, behavioural improvement following injection of NPC has been reported in the absence of a significant reduction in infarct size, and might be attributed to increased plasticity in the grafted brains, possibly via paracrine mechanisms as reported for both neural and non-neural stem cells 16,43-45 .
In our study, NPCs were engrafted at 48 hours post-ischaemia. Data from the in vitro NPC migration assay demonstrated that NPCs responded to migratory cues from this time point, except at the highest concentration. In a recent study, Darsalia et al. demonstrated that this time point was optimal for cell survival and migration in a similar 30 minute occlusion model and intrastriatal engraftment of human fetal NPCs 3 . Similarly, cell migration has been observed in the corpus callosum in other studies engrafting murine embryonic stem cells or human NPC at later timepoints in models of cerebral ischaemia. Injection at the corpus callosum allows for cell injection at a distance from the lesion, reducing exposure to potentially toxic inflammatory environment, while providing a potential pathway for migration towards the lesion site and subsequent migration into the lesion. Clearly, cell type and origin, engraftment timepoint, engraftment site and infarct size are important factors for survival and migration of engrafted cells in ischaemic environments, as well as the parameters used to assess recovery.

Conclusion
We have shown that the MRI cell tracking agent FePro does not affect NPC viability, proliferation or differentiation capacity in vitro. However, in the in vivo studies we have identified possible sources of T 2 -weighted image contrast in the ischaemic lesion that develops over time, creating a challenge for longitudinal cellular imaging studies. In conclusion, this study raises important issues surrounding the use of MRI and MRI contrast agents for longitudinal cell tracking studies in models of injury, where the evolution of endogenous contrast over time within lesioned tissue can be a source of uncertainty in image interpretation.

Grant information
This study was funded by the Child Health Research Appeal Trust and the British Heart Foundation.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

9.
We demonstrated that the contrast agent labelling was stable and that cells were able to retain the contrast agent over a period of at least seven days and upon differentiation. Contrast-agent labelled cells were also capable of proliferation, generating neurospheres over a period of seven days with multiple cells containing the contrast agent. Cell death in contrast agent-labelled cell cultures was not significantly different to unlabelled cells in vitro. Furthermore, our previous in vivo labelling and tracking of endogenous stem cells showed the presence of Doublecortin-FePro-labelled cells 28 days after FePro injection into the lateral ventricle.

What factors present in the ischemic tissue facilitate NPC migration?
We did not investigate the factors present in ischaemic tissue; however we tested factors that had been previously reported to be up-regulated following cerebral ischaemia, and which were associated with enhanced neurogenesis and neuroblast migration in our in vitro migration assays. We found that neurospheres were capable of migration in response to FGF-2 and EGF but did not migrate extensively in the presence of VEGF.

It would be of interest to know if transplanting the cells into the non stroked hemisphere would have allowed for better visualisation of cell migration to the lesion.
We have previously shown that migration of endogenous FePro-labelled cells could be easily visualized, and would expect the same to be the case following cell grafting in a normal brain region. There are a few studies that have transplanted iron-oxide or gadolinium labeled ES or NSC cells into the contralateral, non-ischaemic hemisphere (rather than in the ipsilateral stroked hemisphere), and observed migration of cells towards the ischaemic lesion ( , , , ). In these Modo 2002 Modo 2004 Hoehn 2002 Obenaus 2011 studies, transplanted cells were capable of migration via the corpus callosum into the ishaemic hemisphere, and could be visualised using MRI. However, it should be noted that T2 and T2* heterogeneity in the ischaemic lesion evolves over time, and it clearly becomes an issue particularly at later time points.
No competing interests were disclosed. Competing Interests: