In Vivo MRI Stem Cell Tracking Requires Balancing of Detection Limit and Cell Viability

Abstract

Cell-based therapy using adult mesenchymal stem cells (MSCs) has already been the subject of clinical trials, but for further development and optimization the distribution and integration of the engrafted cells into host tissues have to be monitored. Today, for this purpose magnetic resonance imaging (MRI) is the most suitable technique, and micron-sized iron oxide particles (MPIOs) used for labeling are favorable due to their low detection limit. However, constitutional data concerning labeling efficiency, cell viability, and function are lacking. We demonstrate that cell viability and migratory potential of bone marrow mesenchymal stromal cells (BMSCs) are negatively correlated with incorporated MPIOs, presumably due to interference with the actin cytoskeleton. Nevertheless, labeling of BMSCs with low amounts of MPIOs results in maintained cellular function and sufficient contrast for in vivo observation of single cells by MRI in a rat glioma model. Conclusively, though careful titration is indicated, MPIOs are a promising tool for in vivo cell tracking and evaluation of cell-based therapies.

Introduction

Mesenchymal stem cells (MSCs) reside in many tissues of the adult body with the bone marrow as an easily accessible source and their multipotent differentiation capacity into mesodermal, endodermal, and, though limited, even ectodermal cell lineages has been shown repeatedly (4, 7, 8, 18, 29, 34). Due to ethical concerns and technical hurdles (i.e., the necessity of immunosuppression and formation of teratomas after transplantation), MSCs rather than embryonic stem cells recently came into focus for the development of regenerative cell-based therapies and even for the therapy of brain tumors (6, 14, 16, 25, 26, 28, 33, 40, 41). Encouraging experimental data have already led to preclinical and clinical trials in the treatment of myocardial infarction, osteogenesis imperfecta, graft versus host disease, and even neurological disorders. However, although proving the safety of MSC transplantation, the benefit varied from promising to none (5, 6, 13, 22, 43). As it has been shown that ex vivo culturing and even the injection procedure influence the regenerative capacity and physiology of MSCs (10, 23), there is general consent that these differing results demand more research on the influence of administration routes, distribution, and differentiation of transplanted cells in vivo during the development of cell-based therapies, pointing towards the necessity for effective in vivo imaging techniques.

In principle, whole-body scans are required after cell engraftment to detect injected cells in various organs, followed by high-resolution scans of the tissues where the cells were found. For this purpose, so far only magnetic resonance imaging (MRI) meets the requirements of penetration depth, spatial resolution, repeated measurements of individuals, and clinical availability (28, 36, 46). In vivo cell tracking using MRI implicates cellular labeling in order to enhance the contrast of injected cells against the background of the host tissue. With regard to the required scanning procedure to detect injected cells eventually dispersed throughout the body, the detection limit of the labeling agent used is a critical factor. Therefore, iron oxide-based contrast agents creating a strong hypointense (dark) signal are favored (11). Among those iron oxide-based particles, micron-sized iron oxide particles (MPIOs) have been proven to meet all the requirements for cell tracking during cell-based therapies, most noteworthy single cell detection even in organs with a high blood supply over a period of several weeks (15, 20, 32, 37–39).

Beyond the detection issue, considering the underlying goal of cell-based therapies, the impact of the labeling agent on cell viability and function is of utmost significance. However, substantial data are sparse. In most studies, iron oxide-labeled cells are detached from the culture dish prior to determination of cell counts, colony forming units, or the rate of apoptosis (1, 12, 21, 31). As we are concerned that the internalized iron exhibits a considerable mechanical burden to the cells during the procedure of detachment and washing, we determined cell counts directly in the culture dish and performed electron microscopy on the labeled cells to evaluate whether MPIOs induce apoptosis or necrosis, because injection of necrotic cells increases the risk of inflammation (7). The proper physiology of MSCs is generally defined by their capacity to differentiate into mesodermal lineages, but for cell-based therapies their migratory potential is also of major interest and respective data in conjunction with iron oxide-labeled cells are lacking. Hence, we evaluated the capacity of MPIO-labeled cells to migrate in a wound-healing assay. Although we used the isolation protocol generating MSC cultures, we here refer to them as bone marrow-derived stromal cells (BMSCs) for ongoing dispute about the precise definition of MSCs (29, 45).

Materials and Methods

Cells and Culture Conditions

BMSCs were isolated from bone marrow aspirates of femur and tibia of 3–4-month-old Wistar rats by centrifugation at 800 × g for 30 min using a Ficoll gradient. The resulting cell layer in the interface was washed twice with complete culture medium [α-MEM; Biowest SAS, Nuaillé, France); 20% (v/v) FCS Gold (PAA Laboratories, Cölbe, Germany); 200 μM l-glutamine, 100 U/ml penicillin/streptomycin], plated in a 21.5-cm² dish in 5 ml complete culture medium and incubated at 37°C, 5% CO₂. After 24 h nonadherent cells were discarded. Medium was then exchanged every 4 days, and after 7–10 days cells were detached with Accutase® (PAA Laboratories) and plated at 1,000 cells/cm². Cells were passaged when they reached 80% confluence and cells of passage 2 to 6 were used in experiments.

Labeling with MPIOs

The MPIOs (BangsLabs Inc., Fishers, IN, USA) had a mean diameter of 1.63 μm, containing a core of magnetite (42.5% weight; mean iron content 1.1 pg/particle) encapsulated with a neutral styrene/divinyl benzene coating. At the day of labeling culture medium was removed and replaced by fresh medium containing 0, 1, 2, 5, and 25 μl of MPIO solution suspended in 200 μl PBS, respectively. After 24 h of incubation medium was removed and the culture dish washed twice with 5 ml of complete culture medium to remove excessive MPIOs.

Wound-Healing Assay

A scratch was drawn through the confluent cell layer using a pipette tip. Detached cells and debris were removed, fresh medium added, and four sites of the scratch were marked for determination of the scratch width by taking photographs immediately (0), 24, and 48 h later, using an inverted microscope (Zeiss Axiovert, Zeiss, Oberkochem, Germany) with a digital camera (Spot RT Monochrome, Diagnostic Instruments Inc., Sterling Heights, MI, USA). In each photograph the scratch area was divided in five rectangles of identical length and the width of each was given by the widest rectangle fitting into the cell free space. The arithmetic mean of the five determined widths was taken as the scratch width of the specific site.

Microscopy

The actin cytoskeleton was revealed using biotinylated-XX phalloidin (Invitrogen, Karlsruhe, Germany) followed by streptavidin-Cy2 (Amersham Biosciences, Munich, Germany) after fixation with 4% PFA and permeabilization with 0.25% Triton X-100 (Sigma-Aldrich, Steinheim, Germany). Nuclear DNA was stained with Hoechst®-33342 (Sigma-Aldrich) and fluorescence micrographs were taken using a Zeiss Axiophot (Zeiss, Oberkochem, Germany) equipped with a Leica DFC350 FX camera and Leica FW4000 software (Leica, Wetzlar, Germany).

For light and electron microscopy cells were fixed in the culture dish with glutaraldehyde (25%) and picric acid (0.4%) in 0.1 M cacodylate buffer (pH 7.4). After postfixation with 1% osmium tetroxide and dehydration, cells were embedded in Araldit (Serva, Heidelberg, Germany). Culture dish and embedded cells were separated and either semithin sections (0.7 μm, stained with methylene blue) were prepared or ultrathin sections (50 nm) were collected on copper grids and examined by transmission electron microscopy using a Zeiss 902A (Zeiss).

Magnetic Resonance Imaging (MRI)

Phantoms for MRI studies containing labeled cells were produced using culture dishes filled with agar (Sigma-Aldrich, 1.6% in PBS) as previously described (21). Drill holes (4-mm diameter of a defined depth) were filled with 15 μl suspensions of the labeled cells in agar (10–5 × 10³ cells μl¹). MR images were acquired using Bruker Biospec 7.0 (cell phantoms) and 4.7 Tesla (animal experiments) small-animal scanners (Bruker BioSpin, Ettlingen, Germany; horizontal bore, 30 cm) equipped with actively shielded gradients (200 mT m⁻¹ for animal experiments and 400 mT m⁻¹ for agar cell phantoms). A purpose-built radiofrequency surface coil was used for phantom MRI experiments (21). For animal experiments a purpose-build 12-cm Helmholtz coil for excitation and a 3-cm surface coil for detection were used. 2D multislice-multiecho (MSME) experiments were acquired for the calculation of T2 maps [TR = 6,000 ms and 16 TE increments of 8.5 ms, 2562 matrix, FOV: 3.7 × 3.7 cm (agar cell phantoms) or 2.5 × 2.5 cm (animal experiments), 0.8 mm slice thickness]. 3D T2*-weighted MR images were acquired with high resolution using gradient echo sequences (FLASH, TR = 150 ms, TE = 20 ms, flip angle 30°). The field-of-view was 2.0 × 2.0 × 1.2 cm for animal experiments and 3.7 × 3.7 × 1 cm for agar cell phantoms, respectively. The isotropic spatial resolution was usually 78 μm for phantom experiments and 110 μm for animal experiments.

Images were processed using Paravision 4.0 (Bruker BioSpin) and NIH ImageJ. Relative quantification was performed using 3D T2*-weighted MR images. Either the relative mean signal intensity of the respective drill holes was determined relative to unlabeled cells (SI = 100%) or, for quantification of low cell densities, hypointense pixels were counted using ImageJ after deter-mining an intensity threshold (5% of the mean pixel intensity of cell free sections) as previously described (21). In addition, relaxation rates (R2) were determined as mean values of homogeneous sections of the cell loaded areas in the agar phantoms. Values were compared to those of unlabeled cells in the same phantom.

In Vivo Experiments

All animal experiments were performed in agreement with the current German animal protection act. Permission was granted by the district government of Cologne (AZ: 50 203.2-K 35, 6/03).

The rat C6 glioma cell line cultured in complete medium [except for 5% (v/v) FCS] was used for glioma establishment in the animals. Male Wistar rats (300–320 g body weight) were anesthetized with ketamin/Rompun and 20,000 C6 cells suspended in 5 μl sterile PBS were injected stereotaxically into the right striatum using a 26-gauge Hamilton syringe (coordinates: x = 0, y = 3.5, z = 5.4). Seven days after tumor induction 1.5 × 10⁵ MPIO-labeled BMSC suspended in 5 μl sterile PBS was injected into the corpus callosum (coordinates: x = 0, y = 1.2, z = 3.0) contralateral to the tumor. MR images were taken immediately and 7 days after injection of the BMSCs.

Statistical Analysis

Data of MR phantoms are expressed as mean ± SD and significance tests were performed using ANOVA (Origin 7.5) with values of p < 0.05 considered significant. All other data were processed using SPSS® 14.0 for Windows (SPSS Inc., Chicago, IL, USA). Uptake of iron microspheres per cell is presented in box plots and was analyzed using Kruskal-Wallis ANOVA followed by Mann-Whitney U-test for pairwise comparison of samples with previous adjustment of alpha errors by Bonferroni correction. Cell survival and in vitro migration are expressed as mean ± SD and significance tests were performed using one-way ANOVA with Scheffe post hoc test with values of p < 0.05 considered significant.

Results

Concentration-Dependent Uptake of MPIOs and its Effect on BMSC Viability

Concerning correct interpretation of MRI results, precise knowledge of uptake and dilution through cell divisions of the applied labeling is of utmost significance. Therefore, we determined the number of internalized MPIOs per cell after 24 h of labeling and 48 h of further culturing using electron micrographs. Figure 1A clearly points out that the number of iron particles internalized by BMSCs strongly depends on the added number of MPIOs. With the lowest amount of label (1 μl/ml) around one third of the cells were free of iron particles, whereas 2 μl of MPIOs/ml medium already led to a mean uptake of 7 iron particles per cell and approximately 5% without label. Further increase of MPIOs in the medium enhanced uptake, showing a good correlation of labeling amount and particle load per cell. The remarkable spread of MPIOs per cell for each labeling concentration points to different accessibility of microspheres to BMSCs positioned in the center of a cluster and on its border, respectively. Therefore, the upper limit of uptake was around 250 MPIOs per cell because even a fivefold increase of labeling from 5 to 25 μl/ml did not lead to a higher maximum load.

Figure 1.

Increasing amounts of MPIOs lead to enhanced uptake by BMSCs. (A) Number of iron microspheres internalized by BMSC during 24 h of labeling and 48 h of further culturing with the indicated volumes of MPIOs added to the medium. The box plot shows 10% and 90% (whiskers), 25% and 75% (gray box), and median (black line in box). Single dots are values below 10% and above 90%, respectively. Increase in uptake of iron microspheres was significant for each concentration (Kruskal-Wallis test, p < 0.001; Mann-Whitney U-test after Bonferroni correction, p < 0.01 for all pairs; n = 46 to 94 from three independent experiments). (B) Representative semithin sections (upper row) and transmission electron microscopy (bottom row) of BMSC labeled for 24 h with 2 μl (left) and 25 μl (right) MPIOs/ml medium, respectively, and fixated after 48 h of further culturing. N, nucleus; arrows indicate iron microspheres of different size. Scale bars for both corresponding images are given in the right image, respectively.

After labeling with 25 μl MPIOs/ml medium the BMSC mostly showed a rounded phenotype, densely packed with iron particles, and the cytoplasm was virtually absent (Fig. 1B). Several MPIOs impressed the nuclear membrane, leading to a deformation of the nucleus. On the other hand, the BMSCs labeled with 2 μl MPIOs/ml exhibited the usual spindle shaped form and the cytoplasm as well as the nucleus displayed no signs of major effect. At higher magnification electron microscopy revealed that the massive load with MPIOs under conditions of the highest labeling concentration caused severe damage to the BMSCs (Fig. 2A). The sparse cytoplasm is electron lucent, endoplasmic reticulum undetectable, and electron lucent areas within the nucleus indicate DNA degradation, which altogether are signs of necrosis. In contrast, the cells labeled with 2 μl MPIOs/ml contained normal structured cytoplasm with clearly visible cisterns of endoplasmic reticulum. The integrity of the nucleus was maintained with the typical high amount of euchromatic DNA, indicating that cell viability was not affected. To determine the impact of internalized iron particles on cell proliferation, we determined the cell number of BMSCs 48 h after removing the label (Fig. 2B), demonstrating the relation of label amount and cell viability. The correlation of incorporated particles and cell survival (Spearman rho = −0.67, p < 0.001) indicated a limiting number of internalized microspheres per cell that allowed cell survival and division. Cell viability was only one major concern in the use of BMSCs in cell-based therapeutic approaches; to determine migratory potential, we investigated the possible impact of labeling with MPIOs on the latter using a wound-healing assay.

Figure 2.

Subcritical uptake of MPIOs does not affect viability of BMSCs. (A) Representative transmission electron microscopy of BMSCs labeled for 24 h with 2 μl (left) and 25 μl (right) MPIOs/ml medium, respectively, and fixated after 48 h of further culturing. N, nucleus; arrow indicates an iron microsphere that deeply penetrated into the nuclear membrane. Scale bar for both images is given in the right image. (B) Cell numbers of BMSCs after 24 h of labeling with the indicated numbers of MPIOs and 48 h of further culturing. Data are mean ± SD displayed as percentage of the control dishes in each experiment [one-way ANOVA, p < 0.001 followed by Scheffe post hoc test, *p < 0.05 against control (5 μl) and all other samples (25 μl), respectively; n = 6 from three independent experiments].

Subcritical Labeling with MPIOs Conserves Migratory Potential of BMSCs, Allowing Single Cell Detection by MRI

In the wound-healing assay a scratch was drawn through a confluent layer of BMSC, which was subsequently closed by migration of the cells. Measuring of the remaining space between the migrating cells at definite time points facilitated quantification of this process, thus allowing comparative studies between cells loaded with different amounts of MPIOs. Figure 3 shows that unlabeled control BMSCs almost completely closed the scratch after 48 h and that labeling with 2 μl MPIOs/ml medium merely affected the migratory potential. The dark dots within the cell layer indicate medium to high labeled cells, according to the range given in Figure 1A. It is clearly visible that these cells were migrating with similar velocity compared to controls. Moreover, a change in the orientation of the cells from a random pattern immediately after drawing the scratch (0 h) towards the migratory direction could be observed after 24 and 48 h in both control dishes and dishes labeled with 2 μl. In contrast, the cells labeled with the highest concentration did not show any migratory movement or orientation toward the scratch, but gave rather the impression of a decreasing cell density. Together with the unusual rounded phenotype of the cells and the occurrence of free MPIOs with time (dark dots in the middle of the scratch), this indicated at least partial necrosis of the cells.

Figure 3.

Migratory potential of BMSCs is maintained using subcritical labeling with MPIOs. Light microscopy of representative wound-healing assays. BMSCs were labeled for 24 h with 0 (control, left), 2 μl (middle), and 25 μl (right) MPIOs/ml medium. Pictures show the identical site of the respective culture dishes immediately (0 h, upper row), 24 h (middle row), and 48 h (bottom row) after drawing the scratch.

Quantitative estimations of the observations made by light microscopy revealed a strong correlation of cell labeling and migratory properties of BMSCs (Fig. 4A). The lowest amount of labeling showed no significant impairment of the BMSCs to close the scratch, whereas the highest labeling abolished any migratory movement. Labeling with 2 μl MPIOs/ml medium led to a slight, not significant, increase of the remaining gap after 48 h, indicating a somewhat slower movement. On the other hand, labeling with 5 μl iron particles/ml medium decreased the migration of BMSCs to an extent that was significantly slower after 24 h of the wound-healing assay. After 48 h it became even more pronounced with the remaining gap being virtually doubled compared to the labeling with 2 μl MPIOs/ml medium. Even though cell viability thus seemed to be similarly affected by labeling with 2 or 5 μl MPIOs/ml medium (Fig. 2B), the impact on the migratory potential was substantially higher using 5 μl.

Figure 4.

Subcritical labeling does not affect velocity of migration and organization of actin. (A) Migrated distance of BMSC labeled with the indicated volumes of MPIOs 24 h (black bars) and 48 h (gray bars) after drawing the scratch, displayed as percentage of the original width [one-way ANOVA, p < 0.001 followed by Scheffe post hoc test, *p < 0.05 against control (5 μl) and all other samples (25 μl), respectively; n = 6 from three independent experiments]. (B) Representative fluorescence micrographs of BMSCs labeled with 2 μl MPIOs/ml medium that migrated into the former free space of the scratch (left) and cells labeled with 25 μl MPIOs/ml (right). Actin was revealed using Phalloidin staining (green) and nuclei are counterstained with Hoechst®-33342. Scale bar for both images is given in the right image.

As the migratory process requires ongoing reorganization of the cells filaments, we investigated organization and orientation of actin in BMSCs labeled with iron particles. Figure 4B shows that the actin filaments of cells labeled with 2 μl MPIOs/ml medium were organized in clearly visible, single fibers that were orientated in the migratory direction (left). In contrast, in cells labeled with the highest amount of iron particles instead of a filamentous organization of actin, a veil of actin-positive structures surrounded the internalized iron particles visible as numerous black holes (right), indicating a disruption of the actin filament network. The recessus resembling iron particles were much rarer in cells labeled with 2 μl MPIOs/ml medium and of low contrast, indicating that only deeper layers of filaments were affected, thus permitting nearly undisturbed reorganization of the actin fibers of the cortical web important for migratory movement.

Because labeling of BMSC with up to 2 μl MPIOs/ml medium exhibited minimal influence on the desired cell functions, we determined the in vitro MRI detectability threshold of such labeled cells. Figure 5A demonstrates that the lowest labeling concentration facilitated a significant contrast even in highly diluted cell suspensions. Labeling with 2 μl MPIOs allowed visualization of single cells, which was revealed by counting the dark pixel clusters (Table 1). The number of dark pixel clusters was in good agreement with the known cell density as long as it did not exceed 100 cells/μl. At higher cell densities as well as at higher labeling loads the volume effect of the MPIOs, meaning that the area of signal extinction is around 50 times the size of the particle (47), hampered a definite delineation of separated pixel clusters, thus not allowing a correlation any longer (Table 1, Fig. 5A). As relatively low numbers of MPIOs per cell were most suitable regarding cell viability and migration, the leakage of once internalized iron particles from the BMSC was of concern, as this could lead to label dilution beyond MRI detectability and false-positive signal due to free MPIOs, respectively. Therefore, the potential release of MPIOs by the BMSC in the cell culture medium was studied by comparison of the R2 relaxation rates of supernatants from labeled and control cells, respectively, cultured in medium free of iron particles. Repeated determination of the R2 relaxation rates revealed that the R2 values of the supernatant of labeled cells were up to 50% higher for the first 12 h after labeling compared to controls. Thereafter, differences in R2 relaxation rates between supernatants of control and labeled cells were below 10%, indicating that the BMSCs keep the MPIOs inside over time.

Figure 5.

Labeling with MPIOs allows follow-up of migrating BMSC in vivo by MRI. (A) T2*-weighted images (7.0 Tesla MRI scanner, isotropic resolution 78 μm, TR =150 ms, TE = 20 ms, flip angle 30°) of agar phantoms containing 15 μl of cell suspensions with 50 cells/μl (left) or 500 cells/μl (right) in 4-mm drill holes. Prior to measurement cells were labeled for 24 h with 1 μl (top), 2 μl (middle), or 25 μl (bottom) MPIOs/ml medium. (B) In vivo MR images of a rat brain injected with 1.5 × 10⁵ BMSCs labeled with 5 μl MPIOs/ml medium. T2*-weighted image (left) showing the large volume effect of MPIOs that is even visible in the T1-weighted image (right). (C) In vivo MR image (top) 7 days after injection of 1.5 × 10⁵ BMSCs labeled with 2 μl MPIOs/ml medium into a rat brain with a solid tumor established a week prior to BMSC injection. Regions marked by dashed lines and arrow point to areas of hypointense signal, indicating the presence and migration of BMSC containing MPIOs. The numbers refer to the schematic drawing in the inset, showing the injection sites of tumor and BMSCs as well as the details of the brain section analyzed by fluorescence microscopy (bottom) for the presence of PKH67-positive cells, confirming the hypointense signals in the MR image with the presence of injected and migrated BMSCs by histology.

Table 1.

Correlation of Cell Numbers With Hypointense Pixel Clusters in T2*-Weighted MR Images for BMSCs Labeled With 2 μl MPIOs

Cells (μl−1)	Pixel Cluster (μl−1)
1	1 ± 0.1
10	9 ± 0.5
50	47 ± 2
100	97 ± 5
500	600 ± 50
1000	N.S.

Hypointense pixel clusters (1 to 6 pixels, resolution 78 μm). N.S., not separable.

In Vivo Migration of MPIO-Labeled BMSCs Can Be Followed by MRI

Because of their large magnetic susceptibility effect, the MPIOs are also highly efficient for the visualization of labeled cells in vivo. The large volume effect is illustrated in Figure 5B, showing that injection of BMSC labeled with 5 μl MPIOs/ml resulted in a volume effect more than 100 times larger than the engrafted volume. This is not only the case for the highly sensitive T2*-weighted; even the T1-weighted MR images annihilated any cell tracking attempt by masking the migrating cells.

Due to sufficient MRI contrast and the preservation of cellular function, we injected BMSC labeled with 2 μl MPIOs/ml into rat brains with a previously initiated tumor. Seven days after injection the bulk of implanted BMSC was still located at the site of injection, visible as a contrasted area in the MR image that, due to the volume effect of the iron particles, considerably exceeded the area occupied by the injected cells, as illustrated by fluorescence microscopy (Fig. 5C). A stream of hypointense signal connected the injection site along the corpus callosum with the tumor area. No hypointensity was observed along the corpus callosum in the opposite direction, indicating that a specific migratory stimulus was generated by the tumor. The strong hypointensity above the septum pellucidum and lateral ventricles correlated with a high number of migrating BMSCs. Most noteworthy, even the few cells that migrated from the injection site to the opposite hemisphere produce a significant hypointense signal, indicating the detection of very low numbers of MPIO-labeled BMSCs by MR imaging.

Discussion

Today, BMSCs are on their way to clinical application in cell-based therapies. Nevertheless, the divergent outcome of the various studies performed demand proper analysis of MSC transplantation models with regard to desired and, above all, undesired effects (5, 6, 28, 40, 43). The migratory capacity of MSC is well known and for some applications (i.e., tumor targeting) it is indispensable, whereas in other cases (i.e., cartilage regeneration) it is unwanted (17, 33). For clinical follow-up, therefore, an effective in vivo imaging of as few as possible transplanted cells is desired. Although fluorescence and bio-luminescence-based imaging as well as multiphoton microscopy are rapidly improving techniques, so far only MRI meets all requirements for clinical application, especially concerning spatial resolution, penetration depth, and detection limit when scanning large subjects (28, 36, 46).

Among the available contrast agents for MRI, iron oxide particles are favored for cell tracking because of their high sensitivity. As sensitivity depends on the size of the iron core, MPIOs serve best for detecting even single cells labeled with only a few particles (12, 15, 20, 37–39). Labeling rat BMSCs with 1.63 μm MPIOs for 24 h reveals that up to 250 MPIOs are taken up in a dose-dependent manner (Fig. 1A), thereby showing a remarkable spread in the uptake by individual cells, which we conclude to be due to different accessibility to the particles in cell colonies. For 10 μl of MPIOs/ml, an average uptake of 100 MPIOs was reported for porcine MSCs, but also showing different amounts of particles in the cells (37, 38). Using 0.9 μm MPIOs, MSCs were shown to rapidly internalize the particles and segregate them to both daughter cells during cell division (12). Concerning culture conditions in our laboratory 48 h are equivalent to two cell divisions. Given the strong correlation of labeling amount and MPIOs internalized by BMSCs 2 days after labeling, it can be assumed that MPIOs are indeed equally distributed to daughter cells during cell division and in conjunction with the constant relaxation rates of culture supernatant, that once internalized particles are not released by BMSC, in contrast to a recent report on an embryonic stem cell line (21).

Of major interest in the present study was the impact of the internalized MPIOs on the intracellular integrity of BMSCs, because we assumed the large particles to cause mechanical stress. In accordance with previous studies, we found the particles located in the perinuclear cytoplasm (12). Figures 1B and 2A demonstrate that cells with a high load of MPIOs show severe intracellular damage and multiple signs of necrosis. The strong correlation of internalized MPIOs with cell counts 48 h after labeling (Fig. 2B) indicates that there is a number of internalized particles separating cell survival from cell death. It is likely that in a fully loaded cell, organization and proper function of the spindle apparatus is hampered, leading either to cell death or suppression of mitosis. Interestingly, the necrotic morphology of highly labeled cells indicates that cell death due to particle overload is a process not completed 48 h after labeling, probably because of differing severity of damage in conjunction with internalized MPIOs.

It is important to note that initially after labeling these cells most likely exhibit minor effects of the enzymatic apparatus (see below). Finally, we found that the maximum load not leading to signs of necrosis in this study is between 40 and 50 MPIOs per BMSC. This is contradictory to previous studies, which reported an estimated load of 100 of the same MPIOs in MSC with no signs of cell death (38). As they used light microscopy for particle counting, which we found to be inappropriate to identify single particles especially at higher loads and performed trypan blue exclusion after detachment of the cells, this may be due to an overestimation of particle load and mechanical destruction of overloaded cells in the isolation process, which we both avoided by performing electron microscopy and cell counts in the culture dish without prior detachment. In conclusion, this reveals the importance of electron microscopic investigation for other iron oxide agents as well, because commonly used methods for analysis of cell viability (i.e., trypan blue, WST/MTT assay, propidium iodide, and annexin V) might miss mechanically disrupted cells during isolation and falsely register overloaded cells as unaffected, respectively (1, 21, 27, 31). Instead of inducing apoptosis, an overload of BMSCs with MPIOs exerts gradual necrosis, which is of significance, because the above-mentioned tests might underestimate these cells and transplantation of necrotic cells includes a high risk of inflammation (7).

Irrespective of the contrast agent used for MRI, previous studies did not examine the impact of labeling on the migratory properties of stem cells in vitro, but focused rather on in vivo detection. Whether by injection into vessels or near the corpus callosum of the brain it is not surprising that the cells distribute in downstream organs and tissues (24, 30, 37, 42, 47, 48). We show that migration of BMSCs is gradually impaired with higher labeling loads of MPIOs (Figs. 3 and 4A) and we found that excessive labeling with nanometer-sized particles, which is usually performed to overcome limited contrast of these small particles in MRI, also hampers the migratory potential of these cells (unpublished data). For MPIOs this is most probably due to disturbance of reorganization sequences of the cortical actin filament network, increasing with the number of internalized particles (Fig. 4B), which will hamper and finally abolish migration (44).

Despite the restrictions discussed above, labeling of BMSCs with MPIOs below the critical amount of severe damage serves for significant contrast on a single cell level in vitro (Fig. 5A and Table 1) and in vivo (Fig. 5C), taking into account that even a single particle of the same kind we used here can be detected by MRI in vivo (39). MPIOs have been shown repeatedly to generate more efficient contrast for MRI compared to nanometer-sized small or ultrasmall iron oxide particles (SPIOs/USPIOs). Due to their high iron content, a few MPIOs are able to generate the same contrast as millions of SPIOs/USPIOs (12, 39, 47) and although single cells labeled with SPIOs/USPIOs could be detected in vitro, the in vivo detection limit is still in the range of tens to hundreds of cells (9, 30, 42). Furthermore, it has been shown recently that SPIO-labeled MSCs aggravated the symptoms in a rat model of experimental autoimmune encephalitis, most probably due to released iron by MSCs at the sites of inflammation acting as an adjuvant. As nonlabeled MSCs are shown to have protective effects in this model, SPIOs seem to be inappropriate, at least under inflammatory conditions (35). A major point for criticism in the use of MPIOs is the need of phagocytosis for uptake, but multiple cell types have been shown to incorporate the particles effectively, with hematopoietic and mesenchymal stem cells among them (12, 47, 48).

In conclusion, MSCs display a broad differentiation capacity and immune modulatory and tissue adaptation effects, which makes them a promising tool in regenerative medicine as well as cell-based therapy of autoimmune diseases, neurological disorders, and cancer (3, 5, 6, 22, 40, 43). However, the divergent results of animal and preclinical studies illustrate the lack of knowledge about stem cell differentiation and migration in vivo, with MRI being the only present technique matching the requirements of resolution, penetration depth, and clinical availability to overcome this (36, 46). Therefore, if the inevitable labeling inhibits basic cellular functions like proliferation and migration, the purpose is foiled. Here we have demonstrated that a labeling in terms of “as much as possible” hampers using both MPIOs (Figs. 2B, 4A) or, though to a lesser extent, nanometer-sized particles (unpublished data), but which additionally impair differentiation capacities of MSCs or even worse aggravate disease symptoms (19, 31, 35). Our results point to an inference of MPIOs with the actin cytoskeleton (Fig. 4B), which probably induces cell death through mechanical stress. Nevertheless, we found that labeling BMSCs with an amount of MPIOs conserving proliferation and migration allows in vivo tracking on a single cell level (Fig. 5B). More basic research on the impact of labeling strategies for MRI on cell proliferation, migration, and differentiation is indicated, with the latter alone revealing even less data and inconclusive results (2, 12, 19), because the divergent results of animal and clinical studies are probably at least in part due to different migratory patterns of the injected cells (28, 46).