Characterization of an iron oxide nanoparticle labelling and MRI-based protocol for inducing human mesenchymal stem cells into neural-like cells

The aim of the current study was to develop an iron oxide nanoparticle (ION) labelling and magnetic resonance imaging (MRI)-based protocol to allow visualization of the differentiation process of mesenchymal stem cells (MSCs) into neural-like cells (NCs) in vitro. Ferucarbotran, a clinically available ION, which can be visualized under MRI, is used for tracking cells implanted in vivo. The NCs were verified morphologically and histologically by light microscopy, and their functions were verified by measuring their action potentials. Conformational conversion of axon-like structures was observed under light microscopy. These NCs exhibited frequent, active action potentials compared with cells that did not undergo neural differentiation. The labelling of ION had no influence on the morphological and functional differentiation capacity of the MSCs. We conclude that the MSCs that were differentiated into NCs exhibited in vitro activity potential firing and may be used to replace damaged neurons.

The primary aim of the current study was to develop an MRI-based assay for assessing and comparing the labelling efficiency of ION in hMSCs and hMSC-derived neural-like cells (NCs). The secondary aim was to evaluate and compare the intracellular distribution, cellular toxicity and cell behaviour of the hMSCs and hMSC-derived NCs after ION labelling.

Results
Differentiated human MSCs exhibited neural-like morphology and neuron markers: Directly labelling hMSCs and NCs with ION. To investigate the in vitro differentiation of hMSCs into NCs, hMSCs were incubated in neurogenic induction medium (NIM) for NCs differentiation. Compared with undifferentiated MSCs, NCs exhibited dendrite-like features of long spikes extending into other adjacent cells (Fig. 1A, arrowhead) and lower cell densities ( Fig. 1A and B). There was no morphological difference between ION-labelled and unlabelled cells under light microscopy. Both hMSCs and NCs incubated with ION had blue dots precipitated inside the cytoplasm, whereas unlabelled hMSCs and NCs did not have blue dots (Fig. 1B). TEM images also revealed the presence of internalized ION within the organelles of hMSCs and NCs incubated with ION (Fig. 1C). NCs differentiation was further verified by phosphotungstic acid haematoxylin (PTAH) staining. Additionally, co-staining with Prussian blue revealed iron precipitates inside the cytoplasm. Thin and long dendrite-like structures stained in brown were observed in the NCs. By contrast, cells without neural induction exhibited no axonlike structures, and the cytoplasm was not stained. ION-labelled MSCs and NCs exhibited blue precipitate inside cells (Fig. 1D). TEM imaging of the ION structure revealed an inner layer iron-oxide core (Fe 3 O 4 , dark black colour) and a non-magnetic outside layer coated with carboxydextran (grey colour) (Fig. 2).
The differentiation of NCs from hMSCs were further evidenced by several neural molecular markers at both the mRNA and protein level (Fig. 3). RT-PCR results demonstrated the expression of glial fibrillary acidic protein (GFAP), tyrosine hydroxylase (TH) and NEUROD6 genes at 14 and 21 days after NIM incubation. The mRNA expression of GFAP, TH and NEUROD6 were significantly elevated in the NC differentiation group regardless of ION labelling. However, no differences in GFAP, TH and NEUROD6 mRNA expression were observed in the hMSC groups (Fig. 3A).
Neuron-specific protein markers GFAP, NeuN, TuJ1 and TH were weakly expressed in undifferentiated MSCs. However, the protein levels of these markers were dramatically expressed in NCs (Fig. 3B). With immunofluorescent staining, the expression of GFAP, NeuN and TuJ1 were visualized as strong fluorescent signals inside the cytoplasm after NIM treatment in NCs groups. ION labelling did not alter the expression of the GFAP (stained in green), NeuN (stained in green) and beta-III tubulin (TuJ1, stained in red) (Fig. 3C). Expression levels of GFAP, TuJ1, and TH were also confirmed after neural-like cell differentiation by FACS analysis (Fig. 3D). The NCs had a significantly higher mean expression of NCs / MSCs (GFAP: 958/168, TuJ1: 707/87, and TH: 637/40) compared with hMSCs with or without (w/o) ION treatment.  Fig. 4. The ability acquired by NCs differentiated from MSCs to generate spontaneous firing activity patterns was not altered after labelling with ION (Fig. 4A). The quantitative results and spike frequency of the cells with neural-like morphologies indicated the active membrane properties of the cells, as shown in Fig. 4B.
There was no significant difference between NCs with or without ION labelling. Moreover, the voltage of each action potential signal was higher in the cells in the NCs group (90 mV) than in cells that did not undergo NCs differentiation (50-60 mV) (Fig. 4A,B) Labelling with ION did not alter the voltage amplitude of the spontaneous firing activity. By contrast, MSCs did not express significant spontaneous firing activity patterns. In the 180-s observation period, the cells that differentiated into neural-like cells generated more spikes (80 times/180 s) than the cells that did not undergo differentiation (10-30 times/180 s). Taken together, these data demonstrated that NCs differentiated from MSCs have key features of functional neurons with the long-term ability to generate spontaneous firing activity patterns.
In vitro determination of ION uptake by MRI, inductively coupled plasma mass spectrometry (ICP-MS) and flow cytometry. We next examined whether the ION-labelled cells can be detected under non-invasive MRI. Under T2-weighted images, ION-labelled NCs and MSCs showed dark dots at the bottom of the test tube, whereas no dark signal was detected in cells without ION labelling (Fig. 5A). The iron content of the cells was determined using ICP-MS (Fig. 5B). The intracellular level of iron was significantly greater in MSCs and NCs treated with ION (29.2 ± 1.5 pg/cell and 25.9 ± 2.0 pg/cell, respectively) than in untreated MSCs and NCs (0.38 pg/cell and 0.93 pg/cell, respectively). Cell granularity determined by flow cytometry showed more side scatter counts (SSCs) in the ION-treated cells than in the untreated cells (Fig. 5C).

Discussion
In this study, we demonstrated that ION labelling does not affect the differentiation capability of hMSCs to NCs. Cellular functions, including morphology, viability, oxidative stress, and mitochondrial member energization, of ION-labelled hMSCs and NCs were intact. IONs induced no significant difference in mRNA, protein content, neuron-specific protein marker expression, spontaneous firing activity patterns, or intracellular iron levels in NCs.
The clinically used ION, ferucarbotran, has a high uptake capability by stem cells and can be detected using the 1.5 T MRI at the single-cell level [15][16][17] . Our study revealed a slight increase in cell growth after ION labelling possibly because the cell cycle is involved after intracellular Fe release from lysosomes 18 . The ROS of labelled cells were slightly increased due to the increase in H 2 O 2 in the cytoplasm 19 . These results are consistent with the findings of Chen et al. 14 .
Similar to the previous findings 20, 21 , we found no influence of ION uptake on the MMP. Different sizes of supraparamagnetic ION have been used in labelling and tracking neural-like cells during differentiation using MR imaging methods. Such ION include ultrasmall superparamagnetic iron oxide (USPIO) particles (≤30 nm in diameter), superparamagnetic iron oxide (SPIO) particles (30-200 nm in diameter), and micron-sized superparamagnetic iron oxide (MPIO) particles 22 . After intravascular injection, USPIO enters the capillaries, then the matrix, and ultimately arrives at the lymph nodes, but most SPIO accumulates in the reticular endothelial system 23 . Crabbe et al. compared the labelling efficiency of three different stem cell populations [mouse embryonic stem cells (mESCs), rat multipotent adult progenitor cells (rMAPCs), and mouse mesenchymal stem cells (mMSCs)] with three different (ultra)small superparamagnetic iron oxide [(U)SPIO] particles (Resovist, Endorem, Sinerem). The labelling efficiency with Resovist and Endorem significantly differed among stem cells. They found that the minimum cell density that needed to be detected by the MR imager was 75 cells/μl in Sinerem ® (USPIO, 20 nm), and 5 cells/μl in Endorem ® (SPIO, 80-150 nm) and Resovist ® (SPIO, 60 nm). After labelling with Resovist, rMAPCs can still express the neuroprogenitor genes Sox2 and Pax6. They also observed migration to the injured area 3 weeks after implanting 10,000 Resovist-labelled rMAPCs in the bilateral striatum 24 . Based on this previous research, we sought to label MSCs and NCs with SPIO, i.e., Resovist. Many studies have shown no morphological changes in MSCs, neural stem cells (NSCs), or neural progenitor cells (NPCs) labelled with MPIO, and all cell types expressed TH, TuJ1, and nestin after MPIO labelling. As MPIO particles are larger, they are more easily detected, especially at the single-cell level. Another advantage of MPIO particles for cellular tracking is that they have a longer half-life than (U)SPIO particles in the body. Evaluation of intracellular labeling with micron-sized particles of iron oxide (MPIOs) as a general tool for in vitro and in vivo tracking of human stem and progenitor cells 25 . However, (U)SPIO particles can be metabolized over the long-term, whereas MPIO particles with a size >100 nm cannot.
Ferucarbotran is an ION with a particle size of approximately 60 nm and is composed of an iron oxide core coated with carboxydextran to prevent nanoparticle aggregation. The iron oxide is metabolized and reused as a haemoglobin component. Carboxydextran is degraded by lysosomes 15 . A recent study demonstrated that, after exposure of human neuroblastoma cells to 10 μg/ml of 10 nm ION for 24 h, obvious neurotoxicity, including decreased cellular dopamine, increased ROS, increased neural α-synuclein and activated tyrosine kinase c-Ab1 expression, and inhibition of cell-proliferation, were observed in neural cells 26 .
Another study investigating the size and relationship of ION retention in the central nervous system concluded that 40-nm ION possess more evident detention properties in the CNS than 280-nm ION 27 . Another type of ION, the Molday ION, have also been introduced, which are new ultra-small superparamagnetic ION with a magnetic core and hydrodynamic sizes of approximately 8 and 35 nm, respectively, conjugated with Rhodamine-B (Rh-B) (2 fluorophores per particle). The Molday Rhodamine-B ION significantly reduced the survival, proliferation, and differentiation rate of neural stem cells, and upregulated the immune response in recipient animals in a concentration of 50 μg/mL 28 . Such characteristics are not suitable for use as a labelling agent. We speculate that this might be the result of the very small sizes of the particles, the coating material and its fluorophore ligands. These support the possibility that ferucarbotran may be one of the best ION for labelling cells for studies of the nervous system.
We selected Resovist ION because Resovist is a clinically used MR contrast medium approved by the FDA. Internalization can be performed without any transfection agent. Resovist has several advantages such as its good biocompatibility, low toxicity, and excellent stability 29 . Using Resovist labelling, studies of rabbit MSCs revealed that it induced no changes in morphology, it had no effect on neural differentiation, and the protein expression levels of the mature neuron markers TuJ1 and NSE were maintained 30,31 . These findings are similar to our results reported here. Another report using human amniotic membrane-derived MSCs (hAM-dMSCs) showed no expression of GFAP protein, although there was no resulting morphological change. This finding might be due to the fact that glial formation cannot be detected in a relatively short period of incubation during neural-induction differentiation (7 days) or that a different source of MSCs was used 29 .
To improve biocompatibility and biodistribution and to prevent precipitation and agglomeration under physiological conditions, SPION particles are coated with an amphiphilic layer of polymers. These polymers can be hydrophilic and protect the iron oxide core from degradation 32 . Resovist is coated with carboxydextran, which is a type of natural polysaccharide with a net charge, water solubility, biocompatibility, and biodegradability and is enriched with hydroxyl groups, which have been reported to interact with iron oxide via hydrogen bonding.
In addition, poly(L-lysine) (PLL), a positively charged peptide, can be used to facilitate cellular internalization. Ferumoxides-PLL complex-labelled MSCs does not alter biochemical or haematological measures of organ with PLL-coated Feridex and found no morphological changes after differentiation; NSC and GFAP expression were also preserved. After autologous implantation into the monkey brain, BMSC-D-NSCs were still locally detected by MRI as black dots, and they could express NSE protein according to immunohistochemistry assays 35 . Delcroix and associates used another negatively charged material, 1-hydroxyethylidene-1.1-bisphosphonic acid (HEDP)-coated SPIO particles, to label rat MSCs and found no morphological changes after differentiation and preserved TuJ1 and NeuN expression 36 .
Chen et al. labelled mouse NSCs with hydrogen-terminated ultra-nanocrystalline diamond (H-UNCD) and found that it causes spontaneous neural differentiation of NSCs. They claimed that the characteristics of wear/ corrosive resistance in this crystalline structure might improve cellular adhesion and extension. This was proven in their scanning electronic micrograph study, which showed that the filopodia of NSCs extended more than 100 nm and that the expression of neuron markers such as GFAP and TuJ1 was increased according to immunofluorescence staining 37 . As the underlying mechanism, they proposed that the crystalline structure can increase the linkage between NSCs and cytokine or integrin release, which is not related to the hydrophobicity.
The above studies provide insight into some particular surface structures, such as UNCD, that might have an impact on the differentiation of NSCs, but the charge does not seem to be an important factor.
After labelling hMSCs with ferumoxides, another type of ION, the iron content inside the hMSCs decreased upon cell division 38 . Both decreased iron content and loss of cellular granularity are correlated with a decrease in the long-term detectability of cells by MRI 39 . In a clinical MRI study 40 of traumatic brain injury patients in which autologous neural stem cells were labelled with ferumoxides and transplanted into the brain injury area, strong signal changes were detected from days 1 to days 14. After 2 weeks, the cells began to move around the damaged area, and the signal weakened. No signal was detected until the 7 th week, which may be related to the fact that the calibrated neural stem cells had moved and were scattered in relation to the localization region. Consequently, the detectability of ION-labelled neurons is dependent on the initial iron oxide labelling concentration and the speed of cell division. Higher initial iron concentration might influence cell behaviour that is averse to neuron repair. Our study shows that up to 100 μg Fe/mL for incubation is still within the safe range and has little influence on cell viability and differentiation. The intracellular levels of iron in hMSCs and NCs treated with ION were 29.2 ± 1.5 pg/cell and 25.9 ± 2.0 pg/cell, respectively, in the current study, which are comparable to our previous report (23.4 pg/cell in ION of 100 μg Fe/mL) 15 and slightly higher than the levels reported in another study (15-20 pg Fe/cell in ION of 20 and 30 μg Fe/ml) 24 . This suggests that MR imaging using our ION-labelling protocol can successfully localize the stem cells and can be used to track their persistence and migration over time in animal models. Although a previous report claimed nearly 100% labelling efficiency of ION by Prussian Blue iron stain alone, no quantitative evidence was provided to support the conclusion 31 .
The implantation of MSCs into rats that underwent spinal cord injury or stroke improved their performance in previous studies. However, MSCs implanted into the rat brain do not differentiate into functional neurons [41][42][43][44] . Similarly, the implantation of adult NCs failed to improve cell function and migration, which are critical in neuron repair 45 . However, we showed that the induction of MSCs by growth factors is morphologically effective. Moreover, the migration of cells can also be monitored by non-invasive MRI methods. The combination of growth factor induction and ION labelling will facilitate future research in neuron tissue engineering.
NSCs derived from rabbit BM-MSCs did not affect the morphology of neurons and expressed specific neuroprotective proteins (NSE, MAP) after labelling with ferucarbotran 31 . Whole-cell patch-clamp recordings showed that these NCs exhibited electrophysiological activity. Our labelled hMSCs differentiated into NCs that exhibited significant levels of mature neural biomarkers, including observable dendrites and spontaneous firing activity patterns.
Neural function still cannot be restored in patients suffering from dementia, Parkinson's disease, and stroke even despite aggressive treatment methods. The treatment of these disorders with conventional methods or medication can partially relieve symptoms. Furthermore, the spontaneous generation of neural cells after brain injury is limited. Cell therapy can provide a chance to regain neuronal function by replacing dead or degenerative neurons with newly differentiated cells [46][47][48] .
The induction of pluripotent stem cells that originate from human fibroblast has drawn much attention in cell replacement therapy due to the autologous origin of the fibroblasts. However, the tumorigenicity found in phase I clinical trials has limited the applications of this technique 49 . However, hMSCs have extensive differentiation capacity without inducing tumorigenicity, they are easy to propagate, and they exhibit immunomodulatory properties that are beneficial in damaged brain tissue 50 . However, whether these differentiated NCs retain their immunomodulatory capacity should be further evaluated.
Induced neural-like cells implanted into the spinal cords of rats that underwent spinal cord injury improved neurogenesis in rat models and improved function in models of Parkinsonism 51,52 . Such induced neural cells exhibited higher levels of neural markers 53,54 . It is important to be able to confirm whether hMSCs can express the proteins of nerve cells before or during implantation in vivo and in pre-clinical trials. In our study, the ION-labelled NCs expressed multiple mature neural protein markers and produced many NCs in vitro, and thus, they could possibly be used to replace irreparably damaged nerve cells.

Conclusion
We established an ION-labelling and MRI-based protocol for studying neural stem-like cells differentiation and hMSC-derived NCs. All the derived NCs exhibited significant mature neural biomarkers, including observable dendrites and spontaneous firing activity. In ION labelling, the process of implantation and cell migration can be traced by MRI, which is ideal for the analysis of implanted cells in damaged regions and can shed light on cell therapies applied to the central nervous system.

Materials and Methods
All experimental procedures for hMSC culture were approved by the Committee on Biological Research of National Taiwan Normal University and implemented under the guidelines of the Committee.
ION labelling. For incubation with or without ION, ferucarbotran ION (Resovist ® , Bayer Pharma AG, Berlin, Germany) were added to the culture medium at 100 μg Fe/mL. After 24 hours incubation at 37 °C in 5% CO 2 , the cells were further evaluated using Prussian Blue staining (Sigma-Aldrich Co.) and MRI. Examination of neural protein expression through flow cytometry, immunofluorescence, Western blotting and RT-PCR analysis were also conducted.
Morphological analysis. Transmission electron microscopy. We used permanent magnets to observe the concentration of iron oxide ions by placing 10 µg/ml and 100 µg/ml ION in a centrifuge tube. The intracellular 100 µg/ml ION uptake by cells was confirmed by transmission electron microscopy (TEM). 1 × 10 4 labelled or unlabelled cells were cultured in a plastic chamber slide (Lab-Tek, Nunc, Naperville, Il, USA) overnight. After washing with phosphate-buffered saline (PBS; Sigma-Aldrich Co.), the cells were fixed with Karnovsky's fixation solution containing 2% paraformaldehyde (Sigma-Aldrich Co.) with 2.5% glutaraldehyde (Sigma-Aldrich Co.) in 0.2 M cacodylate (pH 7.4) (Sigma-Aldrich Co.) for 2 hours at 4 °C, followed by incubation with 1% osmium tetroxide (OsO 4 ) buffer for 1.5 hours in the dark for post-fixation, rinsing, dehydration, and embedding. Ultra-thin slices were cut from the dried sections with a diamond knife and placed on the grids. Photographic images were taken using a TEM with a CCD camera (Hitachi H-7100; Hitachi, Ibaraki, Japan).

Co-staining with Prussian Blue and phosphotungstic acid haematoxylin (PTAH).
To localize the intracellular ION, 1 × 10 5 hMSCs were exposed to 100 μg Fe/mL ION (Resovist, 45-60 nm) (Schering AG, Berlin, Germany) for 24 hours and then transferred to NIM and incubated for 2~3 weeks. The hMSCs were treated with a 1:1 mixture of 2% potassium ferrocyanide (Prussian Blue) and 1 M hydrochloric acid for 5 minutes. Furthermore, the hMSCs and NCs were stained for astrocytes, fibroglia and myoglia using PTAH (Sigma-Aldrich Co.) for 20 minutes at room temperature. The cells were then washed twice and imaged using a Nikon TE2000-S inverted microscope.

Reverse transcription polymerase chain reaction (RT-PCR).
Total RNA from hMSCs and NCs was extracted with TriZol Reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. Briefly, the cells were harvested after 21 days of incubation and lysed in supplied lysis buffer. Total RNA concentration was determined by measuring the optical density at 260 nm (OD260) using a spectrophotometer (DU800; Beckman Coulter, Fullerton, CA, USA). One microgram total RNA was used to generate cDNA using a SuperScript III First-Strand cDNA Synthesis System (Invitrogen, Carlsbad, CA, USA).
Polymerase chain reaction (PCR) conditions and cycle numbers for a linear amplification range were determined and optimized. The primers used are shown in Supplement Table 1(S1). The PCR amplification was carried out using PCR Tag Master Mix (Applied Biosystems, Foster City, CA, USA). The thermal profile for PCR was 96 °C for 5 minutes, followed by 40 cycles of 96 °C for 50 s, 59~63 °C for 90 s, and 72 °C for 90 s. The products were examined by electrophoresis in a 2% agarose gel, stained with ethidium bromide and visualized under UV light. Actin was used as a housekeeping gene. The gene expression level was quantitated using the National Institutes of Health (NIH) ImageJ program (National Institutes of Health, USA). The expression level of housekeeping gene was defined as 1.0. The expression ratios of hMSC-derived neural specific genes to the housekeeping gene were determined.
Immunofluorescence staining. HMSCs were plated for 24 hours in 48-well culture plates at 2.5 × 10 3 cell/ wells before the experiment. After incubation with or without ION for 24 hours, the hMSCs were cultured in CM, and NCs were cultured in NIM for 21 days. The cells were then washed three times with PBS before fixing in 4% paraformaldehyde solution (Sigma-Aldrich) in PBS at room temperature for 10 minutes. The cells were then washed twice with PBS and permeabilized with 0.1% Triton X-100 (Sigma-Aldrich) for 5 minutes. Nonspecific binding sites were blocked using a 2% BSA solution for 30 minutes at room temperature. hMSCs were incubated with neural primary antibodies including GFAP (1:50) for astrocytes, NeuN (1:100), TH (1:500) and Tuj1 (1:200) overnight at 4 °C. The cells were washed and then incubated with fluorescent (FITC/Rhodamine) secondary anti-mouse or anti-rabbit IgG antibodies (Millipore) at room temperature for 45 minutes. The platelets were washed again with PBS, follow by staining with DNA binding dye, 4′,6-diamidino-2-phenylindole (DAPI; 5 μg/ mL; Molecular Probes) in PBS for 5 minutes at room temperature. The cells were then washed twice and imaged using an inverted microscope (Eclipse TS100; Nikon, Tokyo, Japan).
Flow cytometry analysis of neural markers. hMSCs were plated 24 hours before the experiment in 6-well culture plates at 1 × 10 5 cell/well. After incubation with or without ION for 24 hours, the hMSCs were trypsinized and washed three times with PBS. The cells were fixed in 4% paraformaldehyde solution in PBS at room temperature for 10 minutes, washed twice with PBS and permeabilized with methanol at 4 °C for 15 minutes. The non-specific binding sites were blocked with 2% BSA solution at room temperature for 30 minutes. After centrifugation at 1,500 rounds per minute (rpm) for 5 minutes at 4 °C, the cells were re-suspended in PBS. Electrophysiological recording. A multi-electrode recording MED64 system (Alpha Med Scientific, Japan) was used to observe the changes in hMSC-derived nerve action potentials. Each MED probe contained 64 electrodes in an 8 × 8 grid set point (Alpha Med Science, MED-P515A) and was coated with 5 μg/mL poly-lysine/ laminin (Sigma-Aldrich, St. Louis, MO, USA) at 37 °C in a 5% CO 2 , 95% air atmosphere for 2 hours. After washing with ddH 2 O, 1 × 10 5 of cells were seeded onto the MED probes and incubated at 37 °C in a 5% CO 2 , 95% air atmosphere overnight. CM and NIM were changed every 2 days for 3 weeks. The multi-electrode recordings used a sampling rate of 50 kHz. Total spikes were counted, and the frequency was analysed with a spike sorting analysis system.
Flow cytometry detection of ION particle uptake. hMSCs were plated 24 hours before the experiment in 6-well culture plates at 1 × 10 5 cell/well. After incubation with or without ION for 24 hours, CM and NIM were changed every 2 days for 3 weeks. hMSCs were trypsinized and washed three times with PBS. ION particle uptake by the cells was determined by the number of SSC measurements. For each group of stem cells, 1,500 cell counts were measured via FACS Calibur flow cytometry (FACS Calibur; BD Biosciences, Franklin Lakes, NJ, USA) and CellQuest Pro software (Becton Dickenson, Mississauga, CA).
Magnetic resonance imaging (MRI). MRI was performed using a 7 Tesla animal MR system (Bruker Biospec, 70/30, USR). After ION labelling, undifferentiated and differentiation cells in 6-well plates (1 × 10 5 cells per well) were collected by trypsinization and then washed, centrifuged, and placed in 300-μl Eppendorf tubes (1 × 10 5 cells per tubes) in a water tank. T2-rapid acquisition with relaxation enhancement (RARE) pulse sequences were used (TR/TE = 3000/12.276 ms, flip angle = 180°, matrix size = 256 × 256). The slice thickness was 1.0 mm with a 1.0-mm gap. The field of view (FOV) was 80 × 80 mm for coronal scanning of the test tubes and 10 minutes and 40 s for sagittal scanning at the NEX of 5. All images were then analysed using the Import Bruker NMR Files and ImageJ software (http://rsb.info.nih.gov/ij/plugins/bruker.html).
Intracellular iron content determination. The total uptake of Fe by cells was analysed using ICP-MS (Agilent 7500ce, Agilent Technologies, Palo Alto, CA, USA). 1 × 10 5 cells treated with or without 100 μg Fe/mL ION were collected after 24 hours. The cells were trypsinized, washed and centrifuged. Cell pellets were lysed with 1 ml of 3% HNO 3 (65% HNO 3 ) acid solution. Samples (10 µl) were diluted in 10 ml acid solution and injected to ICP-MS. The Fe concentration was obtained by interpolating to a standard curve obtained from serial dilutions of 0, 10, 20, 50 and 100 ppb Fe.