Author contributions: E.S.M.L.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing; J.C.: conception and design, financial support, administrative support, assembly of data, provision of study material or patients, data analysis and interpretation, manuscript writing; B.S.: data analysis and interpretation, manuscript writing; L.G.T.: administrative support; M.S.K.C.: data analysis and interpretation; D.L.R.: collection of data; G.S.D.: data analysis and interpretation; J.D.: data analysis and interpretation; S.H.T.: administrative support; O.B.: manuscript writing; A.B.: manuscript writing; K.C.T.: provision of study material; M.C.: conception and design, data analysis and interpretation, financial support, administrative support, provision of study material or patients; S.-C.W.: conception and design, financial support, administrative support, data analysis and interpretation, manuscript writing, final approval of manuscript.
First published online in STEM CELLS EXPRESS April 30, 2009.
Stem cell transplantation for regenerative medicine has made significant progress in various injury models, with the development of modalities to track stem cell fate and migration post-transplantation being currently pursued rigorously. Magnetic resonance imaging (MRI) allows serial high-resolution in vivo detection of transplanted stem cells labeled with iron oxide particles, but has been hampered by low labeling efficiencies. Here, we describe the use of microgel iron oxide (MGIO) particles of diameters spanning 100-750 nm for labeling human fetal mesenchymal stem cells (hfMSCs) for MRI tracking. We found that MGIO particle uptake by hfMSCs was size dependent, with 600-nm MGIO (M600) particles demonstrating three- to sixfold higher iron loading than the clinical particle ferucarbotran (33-263 versus 9.6-42.0 pg iron/hfMSC; p < .001). Cell labeling with either M600 particles or ferucarbotran did not affect either cellular proliferation or trilineage differentiation into osteoblasts, adipocytes, and chondrocytes, despite differences in gene expression on a genome-wide microarray analysis. Cell tracking in a rat photothrombotic stroke model using a clinical 1.5-T MRI scanner demonstrated the migration of labeled hfMSCs from the contralateral cortex to the stroke injury, with M600 particles achieving a five- to sevenfold higher sensitivity for MRI detection than ferucarbotran (p < .05). However, model-related cellular necrosis and acute inflammation limited the survival of hfMSCs beyond 5-12 days. The use of M600 particles allowed high detection sensitivity with low cellular toxicity to be achieved through a simple incubation protocol, and may thus be useful for cellular tracking using standard clinical MRI scanners. STEM CELLS 2009;27:1921–1931
Stem cell transplantation is a rapidly emerging field of regenerative medicine undergoing intensive investigation. Several clinical trials are already in progress for the treatment of various diseases, such as ischemic stroke , skeletal dysplasia , spinal cord injury , and myocardial infarction . Transplanted stem cells have been shown to home and engraft [5, 6] into areas of tissue injury, where they replace the defective cell types and/or produce benefit through paracrine mechanisms .
Development of this field requires identification and tracking of transplanted cells to monitor their survival and localization. This has traditionally been achieved through longitudinal histological analyses, using techniques such as gender mismatches, fluorescent proteins, enzymes (LacZ), and thymidine analogs  (e.g., 5-bromo-2′-deoxyuridine). These modalities, however, require serial sacrifice of animals or multiple biopsies, and are beset by problems of interanimal variability. Moreover, these approaches would not be appropriate in clinical studies, and hence there is an urgent need for the development of noninvasive in vivo imaging modalities .
Magnetic resonance imaging (MRI) is a sensitive three-dimensional imaging method that provides high-resolution images at depth in opaque living animals and patients, without ionizing radiation. Serial tracking of transplanted stem cells is feasible if the cells are labeled with MRI-visible particles. Currently available particles span a wide range of diameters (20 nm to 5.8 μm). Dextran-coated superparamagnetic iron oxide (SPIO) nanoparticles in clinical use, such as ferucarbotran (Resovist®; Bayer Schering Pharma, Berlin, Germany, http://www.bayerhealthcare.com)  and ferumoxide (Feridex I.V., AMAG pharmaceuticals, Lexington, MA, http://www.amagpharma.com) , as well as larger (0.9-5.8 μm) polystyrene-based particles [12, 13], have been investigated as cell labels for MRI tracking. Typically, cells are incubated with the particles for up to a few days, allowing uptake through endocytosis prior to transplantation. Although these particles have been successfully used to image labeled cells clinically and in animal models, the low uptake of particles in nonphagocytotic cell types has resulted in limited MRI sensitivity and hampered imaging. The development of MR contrast particles that can be taken up efficiently by nonphagocytotic cell types is therefore a high priority in this field.
A variety of techniques to increase the cellular uptake of iron oxide particles have been described, including reversible electroporation  and the addition of transfection agents (reviewed elsewhere by Bulte and Kraitchman ) such as poly-L-lysine [16, 17], activated-dendrimer , and protamine sulfate . All these approaches require careful optimization to avoid possible cytotoxicity . Particle conjugation to the HIV tat peptide increases uptake but results in its localization within the nucleus rather than in endosomes, and may thus interfere with nuclear function . The size of the particles appears to influence iron loading of cells, with both ultrasmall SPIO particles (20 nm)  and the larger experimental polystyrene particles (0.9 μm)  resulting in lower cellular loading than obtainable with clinical SPIO particles (62 nm) . However, particles of similar composition in the size range of 100-750 nm in diameter have not been previously investigated for this application.
Among various stem cell types, mesenchymal stem cells (MSCs) have shown the ability to home and migrate toward injury sites , such as cerebral infarcts, where they can induce angiogenesis or functional recovery [23, 24]. Human fetal MSCs (hfMSCs) are primitive MSCs, with greater proliferative and differentiation capacities than their adult counterparts [25–28], and have been under investigation for various cellular therapy applications [25, 28–30]
In this study, we investigated the use of microgel iron oxide (MGIO) nanoparticles with a range in diameter of 100-750 nm as MRI contrast particles for cellular labeling of hfMSCs. We showed that MGIO particles of 600 nm diameter (M600) were taken up more avidly by hfMSCs than ferucarbotran, without affecting the stem cell functions of self-renewal and differentiation. M600 labeling allowed better detection in a rat photothrombotic stroke model on MRI than with the use of ferucarbotran. In the growing field of cellular transplantation applications, the availability of highly efficient, low toxicity MRI contrast particles should greatly facilitate cellular tracking and monitoring of transplanted cells.
METHODS AND MATERIALS
Ethics and Samples
All human tissue collection was approved by the domain specific review board of National University Hospital and was in compliance with international guidelines regarding the use of fetal tissue for research . In all cases, patients undergoing clinically indicated termination of pregnancies gave separate written consent for the use of the collected tissue. Female Wistar rats (200-250 g) were acquired from the Centre for Animal Resources (Singapore) and all procedures were approved by the Institutional Animal Care and Use Committee at National University of Singapore.
Synthesis of MGIO Particles
The development of MGIO particles started with the condensation polymerization of ethyl acrylate, methacrylic acid, and di-allyl phthalate to form a nonmagnetic precursor microgel (PMG), as previously described elsewhere . PMG was magnetized by the coprecipitation of iron salt with ammonia to form primary iron oxide cores within PMG in situ. After purification by density and magnetic field strength, MGIO particles were obtained. The polymerization and coprecipitation conditions were altered to produce MGIO particles with differing diameters of 100-750 nm. MGIO was air-dried on a 200-mesh copper grid and imaged by transmission electron microscopy (TEM) (JEOL JEM-100CX microscope; JEOL, Tokyo, Japan, http://www.jeol.com). The hydrated diameter of MGIO particles in suspension was determined by dynamic light scattering (n = 3–6, Brookhaven BI-200SM system; Brookhaven Instruments Corporation, Holtsville, NY, http://www.bic.com), using a power-adjustable 488-nm argon laser source.
hfMSC Isolation and Differentiation
hfMSCs were isolated from human fetal bone marrow as previously described (8-12 weeks of gestation, n = 2) . Briefly, bone marrow cells were flushed from the femurs using a 22-gauge needle into a culture medium (CM10) consisting of 10% fetal bovine serum (FBS) (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) in Dulbecco's modified Eagle's medium (DMEM) (Sigma-Aldrich, Singapore) supplemented with 2 mM L-glutamine, 50 IU/ml penicillin, and 50 mg/ml streptomycin (Invitrogen, Carlsbad, CA, http://www.invitrogen.com). Single-cell suspensions were plated in 100-mm dishes at 105 nucleated cells per ml and cultured in CM10 at 37°C in 5% CO2. After 3 days, nonadherent cells were removed and the medium was replaced. Adherent cell colonies were detached with 0.25% trypsin EDTA (Stem Cell Technologies, Vancouver, BC, Canada, http://www.stemcell.com), expanded, cultured to subconfluence, trypsinized, and stored in liquid nitrogen.
hfMSCs were characterized by immunocytochemistry for CD14, CD34, CD45, CD31, von Willebrand factor (vWF), CD105 (SH2), CD73 (SH3, SH4) (Abcam, Cambridge, MA, http://www.abcam.com), vimentin, laminin, CD29 (Chemicon, Temecula, CA, http://www.chemicon.com), CD44 (BD Biosciences, San Diego, http://www.bdbiosciences.com), CD106, CD90 (Chemicon), human leukocyte antigen (HLA) I, HLA II (Dako, Carpinteria, CA, http://www.dakousa.com), Oct-4, and Nanog (Abcam) and flow cytometry was used to screen for Stro-1 (Chemicon) as previously described elsewhere . Cells at passages 5-6 were used in all experiments. Osteogenic, adipogenic, and chondrogenic differentiation and their respective assays were performed as previously described elsewhere [28, 33].
Cellular Labeling Protocol
Prior to labeling, 5 × 105 hfMSCs were cultured for 24 hours at 2 × 103 cells/cm2 in CM10. The cells were labeled with MGIO or ferucarbotran by incubation at 0.025-0.2 mg iron/ml within a labeling culture medium (CM2: 2% FBS in DMEM supplemented with 2 mM L-glutamine, 50 IU/ml penicillin, and 50 mg/ml streptomycin) at 37°C in 5% CO2. After 24 hours, adherent cells were repeatedly washed with fresh changes of phosphate-buffered saline (PBS) until the PBS appeared free of particles under light microscopy to remove unattached particles. The cells were then trypsinized, resuspended in CM10, and layered on Ficoll-paque PLUS (Amersham Biosciences, Piscataway, NJ, http://www.gelifesciences.com) for density centrifugation at 100g for 30 minutes to remove loosely attached, extracellular particles. Labeled hfMSCs were recovered at the interface between CM10 and Ficoll-paque PLUS and washed with PBS by centrifugation to remove the remaining Ficoll-paque. Mock-labeled cells were used as controls where hfMSCs were subjected to the above procedures but without incubation with any particles.
Cellular TEM and Iron Quantification
To quantify intracellular iron mass, cells were counted, lysed in 0.2 ml of aqua regia, reconstituted to 5 ml with distilled water, and analyzed for iron using inductively coupled plasma optical emission spectroscopy (ICP-OES) (Optima 5300V, PerkinElmer, Waltham, MA, http://www.perkinelmer.com. The iron mass of the mock-labeled control was below the reliable ICP-OES detectability of 0.01 ppm (<0.1 pg/cell). TEM (Leica, Heerbrugg, Switzerland, http://www.leica.com) for labeled cells was performed after the cells were fixed, dehydrated, resin embedded, cut in 100-nm sections, and stained with lead citrate on a copper grid.
Genome-Wide Microarray Expression Analysis
Total RNA was extracted from M600-, ferucarbotran-, and mock-labeled hfMSCs in biological triplicate, using the RNeasy kit (Qiagen, Valencia, CA, http://www1.qiagen.com) in accordance with the manufacturer's protocol. Ten micrograms of total RNA was used to generate labeled cRNA and hybridized to Human Genome U133 Plus 2.0 arrays (Affymetrix, Santa Clara, CA, http://www.affymetrix.com). Differentially expressed genes were identified with GeneSpring GX 7.3.1 (Agilent Technologies, Palo Alto, CA, http://www.agilent.com). The associated gene ontology terms were enriched with the web-based functional annotation software FatiGO . Further details of this process are described in supporting information data.
In Vivo Imaging
Focal Photothrombotic Stroke Induction.
Female Wistar rats were anesthetized with 7.5 mg/100 g body weight (BW) ketamine (Parnell Laboratories, Alexandria, Australia, http://www.parnell.biz) and 1 mg/100 g BW xylazine (Troy Laboratories, Smithfield, Australia, http://www.troylab.com.au) i.p. and mounted in a stereotactic frame. A 7.5-mg/ml solution of Rose Bengal in saline was filtered (0.22 μm) and injected via a tail vein cannula at 1 mg/100 g BW at a rate of 0.2 ml/minute. Simultaneously, the skull, −2 mm anteroposterior (AP) and −3 mm mediolateral (ML) from the bregma, was exposed to 60 W of blue-green passband-filtered (BG39; Schott, Duryea, PA, http://www.us.schott.com) white light from a halogen lamp via a fiber optic waveguide for 10 minutes to generate a photochemical cortical stroke. The spot size was adjusted to a diameter of approximately 3 mm with an optical aperture.
Transplantation of hfMSCs.
Two days after induction of photothrombotic injury to the cerebral cortex (day 0), we xenotransplanted (a) 2 × 104 M600-labeled hfMSCs (M600-hfMSCs, n = 9) or (b) 2 × 104 ferucarbotran-labeled hfMSCs (ferucarbotran-hfMSCs, n = 4) into the contralateral cerebral cortex. A third group (c) had 2 × 106 M600-labeled hfMSCs infused i.v. through the tail vein (n = 2) and a control group (d) of animals without cortical injury was transplanted with 2 × 104 M600-hfMSCs (n = 3). A further two control groups consisted of animals with contralateral stroke injury either without cellular transplantation (n = 3) or transplanted with mock-labeled hfMSCs (n = 1).
A 33-gauge needle was used for the intraparenchymal cortical injection of hfMSCs. Passage of hfMSCs through a 33-gauge needle did not affect cellular viability, as demonstrated through trypan blue exclusion tests, nor did it affect colony-forming unit-erythroid ability (data not shown), suggesting its suitability for transplantation purposes.
For cortical injection contralateral to the stroke site, a burr hole (1 mm) was made on the right side of the skull to expose the dura overlying the cortex. hfMSCs were resuspended by repeated pipetting before loading into Hamilton syringes just before cellular injection to avoid aggregation of the cells. hfMSCs (2 × 104) were injected slowly in 5 μl of PBS over a 10-minute period using a 33-gauge Hamilton syringe into the contralateral hemisphere −2 mm AP, 3 mm ML, and 3.5 mm dorsoventral from the bregma. For i.v. delivery, 2 × 106 cells in 0.5 ml PBS were injected into the lateral tail vein. Immunosuppression with i.p. cyclosporin (20 mg/kg BW on alternate days, Sandimmune Injection; Novartis International, Basel, Switzerland, http://www.novartis.com) was initiated at the time of cellular transplantation and maintained throughout the experimental duration.
In vivo MRI was performed on a 1.5-T whole-body clinical MR scanner with a clinical wrist radiofrequency coil (General Electric, Waukesha, WI, http://www.gehealthcare.com). Anesthesia was induced with 4% isoflurane and maintained with 1.5%–2.5% isoflurane in 100% oxygen delivered through a cone mask. In vivo transverse images were obtained using turbo spin echo (TSE)—field of view (FOV), 5 cm; matrix, 192 × 192 and zero-filled to 512 × 512; voxel dimensions, 260 μm × 260 μm × 1.5 mm; repetition time (TR)/echo time (TE)/echo train length (ETL)/flip angle (FA)/number of excitations (NEX), 2 seconds/81 ms/16/90°/12; acquisition time, approximately 9 minutes—and gradient echo (GRE)—FOV, 5 cm; matrix: 160 × 160 and zero-filled to 512 × 512; voxel dimensions, 313 μm × 313 μm × 1.5 mm; TR/TE/FA/NEX, 280 ms/20 ms/20°/15; acquisition time, approximately 9 minutes—pulse sequences as 10 two-dimensional slices.
Analysis of Hypointense Voxels on MR Images.
A voxel was considered hypointense if its signal intensity was below the signal intensity threshold (SH). Using Rose's criterion, the SH of a day 5 or day 12 image was determined with respect to a reference signal (SR) and the image standard deviation (SDR), as:
The SR was determined from the corresponding day −1 GRE image as the mean signal intensity of an 8 mm2 region of interest positioned at the cortical region contralateral to the stroke. The SDR was calculated from the standard deviation of MR signal from the air, reduced by a factor of 0.655 to account for the non-Gaussian noise of magnitude images. The contrast to noise ratio, k, was assumed to be 5.
Animals (n = 3) were euthanized, and intracardiac perfusion with 250 ml of 2% 2,3,5-triphenyltetrazolium (TTC) (Sigma-Aldrich) 24 hours after the stroke induction procedure was done for confirmation of thrombotic stroke. Following recovery of the brain, 1-mm sections were fixed, laid onto slides, and visualized under light microscopy.
For histological analysis of transplanted cells at various time points, the rats were euthanized and perfused with 4% paraformaldehyde, and the brains were paraffin embedded and 10-μm coronal sections were laid onto polylysine-coated slides for staining.
Prussian blue (PB) iron staining was performed by incubating with freshly prepared 5% potassium ferrocyanide in 5% HCl 1:1 for 30 minutes and washing with deionized water. For 3,3′-diaminobenzidine (DAB) enhancement of PB staining , sections were incubated in 3% H2O2 for 3 minutes pre-PB staining, PB stained, and incubated in 0.05% DAB in PBS for 5 minutes followed by another incubation in 0.05% DAB in PBS and 0.03% H2O2 for 3 minutes.
Immunohistochemical staining was done after deparaffinization, rehydration, and an antigen retrieval step performed at 95°C for 30 minutes (H-3300; Vector Laboratories, Peterborough, U.K., http://www.vectorlabs.com). Sections were blocked with 5% goat and fetal calf serum, and the nuclear envelope was permeabilized with 0.2% Triton X-100 for 1 hour before being incubated overnight at 4°C with primary antibodies of mouse anti-rat ED1 1:100 (MCA341R; AbD Serotec, Oxford, U.K., http://www.abdserotec.com) and rabbit anti-human vimentin 1:100 (ab16700; Abcam, Cambridge, U.K., http://www.abcam.com). After washing of the slides with PBS, incubation with secondary antibodies, either goat anti-rabbit Alexa Fluor 488 or goat anti-mouse Alexa Fluor 594 at 1:100, for 30 minutes was performed. Sections were then mounted with 4′,6-diamidino-2-phenylindole or propidium iodide for nuclear visualization (both from Vector Laboratories).
Parametric data are shown as mean ± standard error of the mean. Iron loading at various labeling concentrations was analyzed using two-way analysis of variance with post hoc Bonferroni correction, or with a t-test. A p-value < .05 was considered indicative of a statistically significant result.
Generation and Characterization of MGIO
Nonmagnetic PMG was synthesized and followed by magnetization by alkaline coprecipitation of iron oxide primary particles within PMG. Resulting synthesized MGIO particles spanned a ninefold range of hydrated diameters (87-766 nm) (Fig. 1A) and consisted of multiple iron oxide primary nanoparticles (∼ 5 nm) held within soft, water-swellable polymeric cages as seen on TEM(Fig. 1B).
Isolation and Characterization of hfMSCs
hfMSCs appeared as plastic-adherent spindle-shaped cells in culture. They had an immunophenotype that was negative for the hemopoietic and endothelial markers CD14, CD34, CD45, CD31, vWF, and HLA II and positive for the mesenchymal markers CD105 and CD73, the intracellular markers vimentin and laminin, the cell adhesion molecules CD29, CD44, CD106, and CD90, and HLA I, as previously reported (data not shown) [28, 29]. Under permissive induction media, they underwent osteogenic, adipogenic, and chondrogenic differentiation (data not shown) confirming their bona fide identity as MSCs .
Uptake of MGIO Particles by hfMSCs
In order to test the utility of these novel particles, we labeled primary hfMSCs with MGIO particles of varying sizes and ferucarbotran. By simple incubation of hfMSCs with MGIO particles or ferucarbotran (0.05 mg iron/ml) over 24 hours, we observed the incorporation of nanoparticles into the cytoplasm of hfMSCs (Fig. 2A) and localization to endosomal structures (Fig. 2B). We found 97.3% ± 0.9% M600-labeled cells staining positive for PB (range, 21-38 per low powered field [LPF]; total cells counted, 212) and 98.2% ± 1.1% (range, 20-35 per LPF; total cells counted, 216) ferucarbotran-labeled cells staining positive for PB. Interestingly, we observed a distinct size-dependent uptake of MGIO particles. Using an initial iron concentration of 0.05 mg/ml, we consistently observed threefold greater cellular iron loading with M600 particles (33.3 ± 4.0 pg/cell, n = 9) than with ferucarbotran (9.6 ± 1.3 pg/cell, n = 9; p = .0003), and significantly higher iron loading for M600 particles than for MGIO particles of other sizes (p < .001) (Fig. 2C). This difference in iron uptake between M600 particles and ferucarbotran was even more marked on incubation with increasing iron concentrations. Up to sixfold greater cellular iron loading was achieved at an incubation concentration of 0.2 mg iron/ml (Fig. 2D) (263 ± 27 pg/cell, n = 3, versus 41 ± 6 pg/cell; p < .001).
Upon passaging the labeled cells in culture, we found that the half-life of the cellular iron content was one population doubling, as can be expected when the iron labels have been fully passed onto the two daughter cells. We also investigated the effects of labeling hfMSCs using MGIO particles of 400-750 nm in diameter and ferucarbotran on their stem cell properties. Compared with mock-labeled cells, labeling of hfMSCs with either MGIO particles or ferucarbotran did not affect cell morphology, doubling time, viability (>95% throughout), or trilineage differentiation into osteoblasts, adipocytes, and chondrocytes (Fig. 3).
In order to broaden our understanding of the effects of iron loading on hfMSCs, we performed a global gene expression analysis using a genome-wide microarray (Affymetrix HG U-133 Plus 2.0) approach. We found 114 differentially regulated genes upregulated at least twofold and 102 genes downregulated at least twofold after labeling with M600 particles (Fig. 2E). These genes are largely associated with upregulation of metal and cation binding and downregulation of the cell cycle (supporting information data, FatiGO analysis ). Labeling of hfMSCs with ferucarbotran resulted in a smaller set of differentially regulated genes, with 32 genes upregulated and 29 genes downregulated. The upregulated genes after ferucarbotran labeling are similarly associated with ion binding, and the downregulated genes are principally involved in prostanoid metabolic processes (supporting information data).
MR Tracking of Labeled hfMSCs in a Focal Stroke Model
Next, we carried out a cellular transplantation experiment to test the sensitivity and efficacy of MGIO particles as a cellular label for MRI tracking in a well-established rat photothrombotic stroke injury model . One day after stroke induction, the photothrombotic injury was visualized on MRI through a TSE (Fig. 4, yellow arrowheads) sequence as a wedge-shaped focal hyperintense region, involving the cortex predominantly, and was confirmed by the absence of metabolic activity through TTC staining (supporting information Fig. 1A). In the M600-hfMSC group, intracerebrally transplanted cells (Fig. 4A, green arrowhead) appeared as hypointense regions on GRE images of the transplanted side of the brain. By day 5, a small area of hypointensity could be seen around the peripheral region of the stroke (Fig. 4A, red arrowhead), suggesting the migration of M600-hfMSCs to the stroke site. On day 12, the area of hypointensity around the injury could be clearly seen encompassing the periphery of the stroke (Fig. 4A, red arrowheads). Animals injected with ferucarbotran-hfMSCs showed a smaller area of hypointensity at the stroke site (Fig. 4B). i.v. delivery of M600-hfMSCs was well tolerated and resulted in the appearance of hypointensity by day 5 at the site of the stroke, which increased from day 5 to day 12 on GRE images (Fig. 4C). MRI of control animals transplanted with M600-hfMSCs but without a contralateral stroke demonstrated no development of hypointensity in the contralateral cerebral cortex (Fig. 4D).
Histological Analysis of Transplanted Animals
We harvested animals in the M600-hfMSC group to correlate the MRI findings with histologic and immunostaining evidence of cell and label fate. On day 1, PB+ cells (staining for iron) could be seen only at the injection site (Fig. 5A, blue stain) and not at the stroke site (Fig. 5B, 5C). Double immunostaining of the injection site revealed hfMSCs as human vimentin-positive cells among an infiltration of host ED1+ macrophages (Fig. 5D–5F), with a discernible shift of cells from the injected site toward the stroke area at day 1. There were only ED1+ host macrophages, with no vimentin-positive cells at the stroke site at day 1 (Fig. 5G, 5H).
By day 5, in keeping with the MRI findings, iron-labeled DAB-enhanced PB+ cells appeared at the periphery of the stroke injury (Fig. 6A–6C) along with the appearance vimentin-positive human cells together with a heavy infiltration of host ED1+ cells (Fig. 6D–6F). Histological sections at day 12 showed an abundance of globular heavily PB+ cells (Fig. 7A–7C), correlating well with the increase in hypointensity on MRI. However, immunostaining revealed only ED1+ cells at the stroke area and no human vimentin-positive cells (Fig. 7D, 7E). Inspection of the injection site revealed only few vimentin-positive cells (data not shown) amid a large infiltrate of ED1+ cells.
In animals transplanted with ferucarbotran-hfMSCs, immunohistological staining at day 12 similarly showed PB+ and ED1+ cells at the stroke site (supporting information Fig. 2A–2E) and a few vimentin-positive human cells among a majority of ED1+ macrophages at the injection site (data not shown). In animals that had been transplanted with M600-hfMSCs through tail vein injection, analysis at day 19 demonstrated similar findings of iron-laden macrophages at the stroke site, again with no visible human cells (supporting information Fig. 3A–3E).
Transplantation of mock-labeled hfMSCs into the contralateral cerebral cortex resulted in no MRI-hypointense regions at either the injection or the stroke site, with infiltration of PB− ED1+ cells into both the injection and stroke sites, and only a few vimentin-positive human cells at the injection site by day 12 (supporting information Fig. 4A–4E). Examination of animals with a stroke injury but without cellular transplantation also demonstrated no MRI hypointensity at either site, with infiltration of only ED1+ cells and no vimentin-positive cells at the stroke site by day 12 (data not shown).
Immunostaining for host CD8+ cells revealed an increasing infiltrate at the stroke region between day 5 and day 12, but these were not found at injection sites.
MRI is an attractive tool for the detection of transplanted cells in living organisms, with high spatiotemporal localization in a noninvasive manner. This requires the development of highly efficient iron oxide particles for cellular labeling to improve on detection sensitivities, lower toxicity, and reduce the requirements for more powerful MRI technologies. Several research particles have demonstrated similar efficacies to clinical SPIO particles but have poor efficiencies in labeling primary cell types. The use of M600 particles for labeling of hfMSCs has demonstrated high efficiencies with corresponding low toxicities, suggesting their utility as a cellular label for MRI tracking.
Incubation of hfMSCs with MGIO particles resulted in internalization into endosomes, suggesting an endocytotic mechanism for cellular uptake, as previously described for ferucarbotran . The labeling of hfMSCs with ferucarbotran resulted in similar cellular loading as previously reported by Mailänder et al.  using adult bone marrow-derived MSCs (9 pg/MSC), which is higher than that achieved with ultrasmall SPIO particles (3.8 pg/MSC)  and the larger polystyrene particles (7.5 pg/MSC) . The unexpected finding of a three- to sixfold greater uptake with M600 particles seems to reflect a size-dependent effect that has been previously reported in the uptake of other nanoparticles by HeLa cells  and T cells . The mechanisms for this preferential uptake at 600 nm are presently unknown, but may in part be explained by differences in the pathways involved in endocytotic uptake with different particle sizes. For example, 500-nm polystyrene particles undergo uptake via a caveolae-mediated pathway, whereas those measuring 50-200 nm are taken up by cells through a clathrin-mediated pathway . The lower uptake of the larger M750 particles may be explained in part by a size limitation of the endocytotic machinery of nonphagocytotic cell types .
Labeling of hfMSCs with either MGIO particles or ferucarbotran did not affect either cellular proliferation or trilineage differentiation. This finding is in contrast to an earlier report by Kostura et al.  of the lower chondrogenic differentiation potential of ferumoxide-labeled adult MSCs, although it remains uncertain whether the cause lies with the particle type used or the choice of transfection agent (poly-L-lysine) used to increase particle uptake .
The greater differences in differentially regulated genes in the genome-wide microarray analysis in the M600-labeled hfMSCs may be a result of greater iron loading than with ferucarbotran-labeled cells or differences in the polymeric content between the two particle types. However, we did not observe any phenotypic differences in the labeled cells in terms of proliferation, viability, and differentiation capacity. The longer term effects of such alterations in gene regulation when labeling cells with iron oxide particles, however, are currently unknown. There have been limited studies on the impact of iron oxide labeling on cellular gene expression. Schafer et al.  demonstrated, through flow cytometry, that the transferrin receptor was upregulated after labeling of rat MSCs with SPIO, whereas Berry et al.  demonstrated that cytoskeleton and signaling genes were upregulated in human dermal fibroblasts after iron oxide labeling. To our knowledge, there is no genome-wide array studies performed in human MSC types after labeling with iron oxide particles.
In keeping with the higher iron loading of hfMSCs with the use of M600 particles over ferucarbotran, the hypointensities developing in the stroke region appeared more striking in animals transplanted with M600- than ferucarbotran-labeled hfMSCs. Using the Rose criterion  (as described in supporting information data) to obtain an image-based quantitative assessment for cellular detection, animals transplanted with M600-hfMSCs had greater numbers of hypointense voxels at day 5 (106.2 ± 15.2, n = 5, versus 14.0 ± 5.5, n = 3; p = .002) and at day 12 (235.3 ± 64.7, n = 4, versus 44.0 ± 22.1, n = 3; p = .03) at the stroke region than animals transplanted with ferucarbotran-hfMSCs. Thus, M600 labeling provided a six- to sevenfold higher sensitivity for cellular detection at both time points, which should provide more reliable detection of transplanted cells.
In order to quantify the superior cell detection with M600 labeling over ferucarbotran labeling, we measured the signal-to-noise ratio (SNR) in our GRE images to be 15.2 ± 0.2 . By applying the detection threshold reported by Heyn et al. , our detection limit was calculated to be 777 ± 126 pg iron per voxel. In terms of cell numbers, the lower limit at which M600-hfMSCs (33.3 pg/cell) can, therefore, be detected is ∼ 23 cells, compared with ∼ 81 cells for ferucarbotran-hfMSCs (9 pg/cell). Under the microimaging conditions possible with research scanners or custom-built hardware (100-μm isotropic voxel dimensions and an SNR of 60) , the detection limit would be lowered from 777 to 1.34 ± 0.22 pg iron per voxel, which should allow a single M600-hfMSC to be detected even after four cellular divisions, assuming that the intracellular iron halves with each cellular division.
Histological analysis of the harvested brain suggested that M600-hfMSCs demonstrated cellular migration toward the injury from day 1, arriving at the stroke area by day 5, as confirmed by both MRI and immunohistological findings. However, by day 12, none of the hfMSCs at the stroke site had survived, and their label had been ingested by host ED1+ macrophages, which appeared as heavily label-laden PB+ cells. This in turn was matched by a similar finding of the disappearance of human cells at the injection site, and corresponding macrophage infiltration. The increase in hypointense voxels in the stroke region between day 5 and day 12 is, therefore, most likely a result of migration of host ED1+ macrophages, following ingestion of iron label from necrosed hfMSCs at the injection site.
This unexpected finding may have been caused by acute graft cellular apoptosis associated with cellular transplantation paradigms [48, 49], resulting in the infiltration of host macrophages attracted to areas of cell death as early as day 1 after cellular transplantation. Coyne et al.  and Amsalem et al.  had previously documented inflammatory rejection of allogeneic rat MSCs post-transplantation into neural and cardiac injury models, and thus this phenomenon is not limited to xenotransplantation models.
We did not find any heavy mononuclear cellular infiltrate into the stroke or injection areas to suggest an adaptive immune response, which would normally take 10-14 days to develop. Moreover, the observation that cytotoxic CD8+ T cells were found only at the stroke sites on day 5 and day 12, and not at the injection sites, suggests that the T-cell infiltrate occurred in response to the stroke and not the cellular transplantation. Furthermore, we used an immunosuppressive dose that has previously been shown to enable human-rat xenotolerance for up to 6 weeks [52–54].
Previously, we demonstrated hfMSC survival in a fetal-to-fetal human-mouse xenotransplantation paradigm with chimerisms of up to 5% for up to 19 weeks in duration [25, 29]. Intracerebral fetal injections in a similar model led to oligodendrocytic differentiation and survival for at least 35 days . Indeed, postnatal transplantation of hfMSCs in adult severe combined immunodeficient mouse muscle was associated with their survival for up to 28 days post-transplantation . The observed cell death cannot be attributed to label toxicity, because a similar finding occurred in both ferucarbotran-labeled and M600-labeled hfMSCs and in mock-labeled hfMSCs (supporting information Fig. 4), and may therefore be a result of a difference in the host environment specific to this model.
We have described a new class of MGIO nanoparticles for cellular labeling and MRI tracking, synthesized in a range of diameters, 100-750 nm, and demonstrated significantly higher iron loading of M600 particles in primary hfMSCs without affecting the stem cell properties of self-renewal, differentiation, and migration capacity. Gene expression analyses showed greater numbers of differentially regulated genes in hfMSCs labeled with MGIO than in hfMSCs labeled with ferucarbotran, which may have been a result of differences in either iron loading or particle composition. Despite the evidence of acute cellular death of hfMSCs in our rat stroke model, we found histological evidence of donor cell migration to a thrombotic stroke after M600 labeling. Greater MRI sensitivity for visualization of cells labeled with M600 than with ferucarbotran was demonstrated at the stroke site. The greater detection sensitivity available with MGIO labeling of hfMSCs should allow better MRI tracking of small numbers of migrating stem cells. This technology could enable in vivo monitoring of cell therapy on standard, widely available clinical 1.5-T MR scanners.
This work was funded by a grant from the Singapore Bio-Imaging Consortium (SBIC) of the Agency for Science, Technology and Research (A*STAR), 023/2005. J.C. received salary support from an Exxon-Mobil-NUS Fellowship.
Shih-Chang Wang is now affiliated with the Department of Radiology, University of Sydney, Australia.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.