|Home | About | Journals | Submit | Contact Us | Français|
Human mesenchymal stem cells (hMSC) were labeled with Ferucarbotran by simple incubation and cultured for up to 14 days. Iron content was determined by spectrometry and the intracellular localization of the contrast agent uptake was studied by electron and confocal microscopy. At various time points after labeling, reaching from 1 to 14 days, samples with viable or lysed labeled hMSCs, as well as non-labeled controls underwent MR imaging. SE- and GE-sequences with multiple TRs and TEs were used at 1.5T and 3T on a clinical scanner. Spectrometry showed an initial iron oxide uptake of 7.08 pg per cell. Microscopy studies revealed lysosomal compartmentalization. Contrast agent effects of hMSC were persistent for up to 14 days after labeling. A marked difference in the T2-effect of compartmentalized iron oxides compared to free iron oxides was found on T2-weighted sequences, but not on T2*-sequences. The observed differences may be explained by the loss of compartmentalization of iron oxide particles, the uniformity of distribution and the subsequent increase in dephasing of protons on SE images. These results show that viable cells with compartmentalized iron oxides may – in principle – be distinguished from lysed cells or released iron oxides.
In recent years, stem cell research has provided insight into developmental biology, the origin of various diseases, and has initiated the development of numerous stem cell based therapeutic approaches (1–3). Among the various stem cell populations used for cell therapy, adult human mesenchymal stem cells (hMSCs) have been identified as ethical, practical and biologically acceptable cell populations for these purposes. They are easily derived from bone marrow aspirates, are efficiently expanded in vitro and can differentiate into a variety of specialized cell populations. In numerous animal and human studies, hMSCs have displayed excellent tissue regeneration potential (4,5). Treatment with hMSCs resulted in improved coronary perfusion in patients after acute myocardial infarction (5), cartilage repair in osteoarthritis (1) and favorable bone repair in patients with osteogenesis imperfecta (6). While these results paint a promising picture of stem cell based therapies, questions concerning the localization and viability of hMSCs post transplantation remain inadequately answered.
Labeling hMSCs with iron oxide based contrast agents allows for sensitive, non-invasive, real-time cell tracking in vivo via MR imaging (7–10). Various labeling methods and MR imaging techniques have been established for cell tracking in vitro (11–13), both in animal models (7,10,14) and in the clinical setting in patients (15). While most cell labeling studies to date concentrate on the anatomical localization of labeled stem cells (8), recent research attempts have focused on the functional aspects of transplanted cells, such as their viability (16), differentiation capacity (16,17) and long-term residence in the target organ (18,19). These aspects are essential for the assessment of successful stem cell based tissue regeneration.
Migration of transplanted stem cells from one location to another indicates unimpaired viability (19). However, if the cells stay at the site of local injection, for example in the case of hMSC transplantation into cartilage defects, additional criteria to verify the viability of the transplanted cells would be desirable. Recent studies suggest that iron oxides display different signal characteristics when concentrated inside cells versus in solution as free particles (20). These differences in MR signal intensities of intracellular and extracellular iron oxides may be explained by differences in their local concentration, their compartmentalization or their ability to interact with surrounding protons (21,22). Therefore, we hypothesize that this effect can be utilized to differentiate viable cells, which contain compartmentalized intracellular iron oxides, from non-viable cells, which have released iron oxides.
The purpose of this study was to investigate the differences in MR signal characteristics of viable versus dead hMSCs considering two major influencing variables: (1) the time point after the labeling process and (2) the applied field strength. Thus, we investigated the signal intensity of viable versus lysed, iron oxides labeled hMSC at (1) different time points after the labeling process and at (2) two different field strengths, 1.5 and 3 T.
Ferucarbotran (Resovist, Schering AG, Berlin, Germany) is a superparamagnetic iron oxide (SPIO) composed of a 4.2 nm crystalline nonstoichiometric Fe2+ and Fe3+ iron oxide core and a stabilizing carboxydextran coat. This coating ensures aqueous solubility and results in a net negative charge for the particle (23–25). The contrast agent has r1 and r2 relaxivities of 7.2 ± 0.1 mM−1s−1 and 82.0 ± 6.2 mM−1s−1 respectively (in blood at 37°C and 1.5 T) and a hydrodynamic diameter of 62 nm (25). Ferucarbotran is approved for MR imaging of the liver in patients in Europe since 2001 (24).
The size and the physicochemical surface properties of Ferucarbotran allow for efficient internalization into macrophages, monocytes and natural killer cells, without additional transfection agents (13,23,26). After uptake, the iron oxide particles are compartmentalized in endosomes and lysosomes within the cytoplasm of the cells (13).
Human mesenchymal stem cells (hMSC) were obtained from a 20-year-old male donor without a bone marrow (BM) disorder, who was admitted to our institution for trauma surgery. The patient gave consent to donate BM via needle aspiration from the posterior iliac crest. The study was approved by the Institutional Review Board. Preparation of hMSCs was done using slightly modified established protocols for primary cultures (27,28). BM mononuclear cells were plated in 100 cm2 pretreated culture dishes (BD Falcon, Franklin Lakes, NJ, USA) and incubated overnight. Non-adherent cells were removed by washing. The remaining cells were grown in standard cell culture condition (37°C in a humidified 5% CO2 atmosphere) using DMEM high glucose medium supplemented with 10% fetal bovine serum (FBS, Hyclone, Logan, UT) and 1% Penicillin-Streptomycin. Culture medium was changed every third day and the cells were split upon reaching 90% confluency. The stem cells were passaged no more than 12 times to preclude senescence.
For cell labeling, hMSCs were plated at 80% confluency and allowed to adhere overnight. The cell labeling solution, consisting of high glucose DMEM and Ferucarbotran at a concentration of 100 µg Fe/ml, was added to the cell culture flasks (100 µl/cm2). After 2 hours of incubation in standard cell culture conditions, 10% FBS was added and cells were incubated for an additional 16 hours. Cells were typsinized, washed 3 times in PBS (25° C, 400 rcf, 5 min) and either replated or used for experiments. For proof of cell viability, a Trypan blue test was performed before and after all labeling procedures.
Before each MR experiment, cells were counted using a hemocytometer. For MR imaging, the cell samples were resuspended in 400 µl of isotonic Ficoll solution (density of 1.07 g/ml), which matches the density of hMSCs. The isodense Ficoll solution prevents cell sedimentation. Ficoll is widely used in the isolation of primary cultures, it preserves viability of the labeled cells during scanning (29,30).
First, reproducibility of the labeling efficiency was studied by labeling three cell cultures, each at three different time points relative to culture passages. The first triplicate was labeled at day 0, the second triplicate was labeled 6 days later and the third triplicate was labeled 12 days later. After measuring viability and cell count 0.5 × 105 cells were taken from each culture for iron content analysis by inductively coupled plasma atomic emission spectrometry (ICP-AES).
In the second experiment, a series of cell solutions were prepared in which 1×106 cells were taken from cultures that remained unlabeled or that were labeled 12, 10, 8, 6, 4, 3 and 1 day earlier, and placed in 0.4ml Ficoll (e.g. cell density of 2.5 × 106 cells/ml) for MRI. Identical series were prepared by adding 0.5×106, 0.25 × 106, 0.125 × 106 and 0.0625 × 106 cells in 0.4 ml Ficoll, yielding an array of 9 by 5 cell solutions with additional control solutions of Ficoll without cells. This experiment was performed to assess changes in iron content and MR relaxation over time during cell division and potential iron metabolism.
In a third experiment, all labeled cells were scanned at a specific time point (i.e. the cells labeled and their progeny). This should show if the iron content and MR signal characteristics of the whole cell population changed over time. For this, cells were labeled 14, 10, 4, 3, and 2 days prior to the experiment. Then, samples of these labeled cells and of unlabeled controls were lysed using a sonic dismembrator. Sonication causes physical disruption of cell membranes by pulsed, high frequency sound waves. Cells are lysed quickly and completely while no metabolism of contrast agent takes place. Sonication was performed at 20kHz for 90 seconds, intermittently sonicating and pausing in 5 second intervals, in order to avoid excessive heating of the sample (Sonic Dismembrator Model 100, Fisher Scientific, Pittsburgh, PA, USA). Finally, for MR imaging duplicates of 1.2 × 106 intact and lysed cells in 0.4 ml Ficoll from all the time points mentioned above were arranged in a tray and imaged.
All samples were scanned on clinical 1.5T and 3T MR scanners (Signa EXCITE HD 1.5T and Signa EXCITE HD 3T, GE Medical Systems, Milwaukee, WI, USA) using standard quadrature knee coils (Clinical MR Solutions, Brookfield, WI, USA). The test tubes were placed in a water-containing plastic container at room temperature (20°C) to minimize susceptibility artefacts from surrounding air. For measurements of T1 and T2 relaxation times, coronal spin echo (SE) sequences were obtained with multiple TR (250, 500, 1000 and 4000 ms) and TE (16, 32, 48 and 64 ms at 1.5T; 15, 30, 45, 60 ms at 3T) values. In addition, coronal gradient echo (GE) images were obtained with a flip angle of 30 degrees, a TR of 500 ms and varying TE values (3.7, 7.2, 14.4 and 28.8 ms). All sequences were acquired with a field of view (FOV) of 160×160 mm, a matrix of 256 × 196 pixels, a slice thickness of 5 mm and one acquisition.
MR images were transferred to DICOM format and processed by a self-written IDL program (Interactive Data Language, Research Systems, Boulder, CO, USA). T1 and T2-maps of the test samples were calculated by fitting the appropriate relaxation equation to the image data assuming monoexponential signal decay on a pixel-by-pixel basis. T1 maps were calculated using SE sequences with a fixed TE (15 ms for 3 T, 16 ms for 1.5 T) and multiple TR values using the equation Mz(t) = M0 (1−exp(−TR/T1). T2 maps were calculated from SE images obtained with TR 4000 ms and multiple TE values using the equation Mxy(t) = M0 exp(−TE/T2). T2* maps calculated from the GE images were inadequate due to low signal-to-noise, therefore T2* values were calculated from regions of interest (ROIs) selected from within the sample vials and the method of (31) was used to calculate T2* using the equation M(TE)2 = M02exp(−2TE/T2*). The relaxation times of intracellular iron oxides in viable cells were derived by ROI measurements of the test samples on calculated maps. SNR were derived by ROI measurements of the test samples on T1, T2 and T2* images.
The iron concentration of all test samples was determined by inductively coupled plasma atomic emission spectrometry (ICP-AES). Samples were dissolved in a microwave by adding 65% HNO3 and 30% H2O2. The obtained solutions were then nebulized into argon plasma. Collaborators from Schering AG Berlin, who were blinded with respect to the content of the samples, performed these analyses.
Labeled cells and unlabeled control cells were plated on glass chamber slides (Nunc, Rochester, NY, USA) and cultured as described above. Cells were fixed at 4°C with Carnoy’s solution. Then, the dextran-coated iron oxides were stained by incubation with an anti-dextran FITC labeled antibody (Stem Cell Technologies, USA) at room temperature for 60 minutes. Finally, a DAPI-counterstain was performed (Vectashield, Vector, Burlingame, CA, USA) and the slides were mounted. Confocal microscopy was performed using a 40x objective (Zeiss LSM 510, Thornwood, NY, USA). Images were acquired using two different filters: for the DAPI staining a UV filter was used (wavelength 351 nm); for the FITC staining an Argon/2 filter was used (wavelength 488 nm).
Adherent cell cultures grown on Thermanox coverslips were fixed with 2% glutaraldehyde in 0.1M sodium cacodylate buffer, postfixed with 1% osmium tetroxide followed by 2% aqueous uranyl acetate. The samples were dehydrated with ethanol and embedded in epoxy resin. Ultrathin sections were stained with 2% uranyl acetate and Reynolds lead-citrate. They were examined and photographed at 80kV in a JEOL 100CX II (Jeol, Peabody, MA, USA).
The experiment to assess the reproducibility of labeling efficiency was performed in triplicate, the comparison of lysed and viable cells was performed in quadruples. Quantitative data from these studies were compared between different groups of cells (e.g. before and after cell labeling, viable and non-viable cells) for significant differences with a t-test and a p-value of less than 0.05. To lower the multiple testing problem, repeated pairwise comparisons in one sample were adjusted with the Bonferroni correction. Statistical computation was done using SPSS 17.0 software (Munich, Germany).
Analysis of iron content in labeled hMSCs found reproducible cellular iron uptake, with 7.08 ± 0.48 pg iron per cell after the labeling process. The iron content in unlabeled cells was under the detection limit of the equipment used (Figure 1). This represents a labeling efficiency (amount of iron internalized by cells divided by amount of iron in labeling media) of 0.52% ± 0.074%. No significant reduction in labeling efficiency was evident for cells labeled 6 or 12 days apart from that (Figure 1, p > 0.05). Examination of cells by fluorescence microscopy immediately after labeling and four weeks after labeling showed that the contrast agent was located predominantly inside the cells, with very few iron oxide particles localized at the cell membrane or outside of the cell (Figure 2, a – c). Electron microscopy of labeled cells confirmed intracellular distribution and showed Ferucarbotran particles in lysosomal compartments within the cytoplasm of the cell (Figure 2, d – f). After sonication, the cellular compartmentalization was destroyed and contrast agent was released (Figure 2 g).
Moreover, the cellular structure of labeled cells showed no noticeable morphologic abnormalities when compared to unlabeled controls on fluorescence microscopy or electron microscopy. In addition the labeled cells showed a doubling time of approximately 7 days, which was not significantly different from unlabeled cells. Trypan blue exclusion testing before and after the labeling process showed no significant difference in viability between different samples of labeled cells versus non-labeled controls. All samples exhibited viabilities greater than 97% and the viability of labeled cells did not change significantly at different time points after labeling (Figure 1).
Cell solutions prepared in sample vials containing 1 × 106 cells and serial dilutions down to 6.25 × 104 cells in 0.4mL Ficoll at various time points up to 12 days after labeling were evaluated. ICP-AES showed an initial iron oxide content of 7.07 pg per cell at day 1, followed by a moderate decline in samples for cells labeled multiple days prior to analysis (Figure 3). Corresponding MR studies showed an initial decrease of T2 relaxation times (e.g. increased R2) of labeled cells compared to unlabeled controls followed by a modest increase toward baseline value over the observation period of 12 days consistent with the measured decrease in iron content (Figure 3). In addition, the change in T2 values decreased as expected for serially diluted cells, with a marked change from control evident for the most dilute cell solution (62.5 × 104 cells in 0.4mL). These data were consistent at 1.5 and at 3T. T2* values could be measured only for the most dilute labeled cell solutions due to large signal reductions at even the smallest TE values. T2* of cell solution at day 1 was 8.9 ± 1.6ms, substantially reduced from T2* of 24.4 ± 2.5ms observed for unlabeled cells. Moreover the reduction of T2 effects over time post labeling was not evident in measured T2* values, instead no marked effect of time was observed. T1 relaxation times at the same cell dilution showed an initial decrease from unlabeled controls (T1 = 1481 ms) to labeled cells (T1 = 1219 ms) at day 1. As it was found for T2* relaxation times, these values showed no marked effect over the timecourse of 12 days (T1 = 1208 ms.)
In the second experimental setup, which included all progenies, disruption of cells by sonication to maximally release the internalized SPIO caused great reduction in signal on all SE images. It was not possible to evaluate either T2 or T1 for these solutions because MRI signal of all sonicated cell solutions was at noise level using minimal TE setting. The T2* values of disrupted cell solution were reduced to 3.9 ± 0.3 ms, slightly less than half of intact cell solutions. Therefore, we compared the signal to noise ratios (SNR, Figure 4) for viable and lysed samples based on T1, T2 and T2* sequences. Viable cells showed significantly higher SNRs on T1 and T2-weighted SE sequences compared to released iron oxides. On the other hand, differences in SNR-data on T2*-weighted sequences were only minimal and not significantly different. The differences in SNR data of compartmentalized iron oxides in viable cells and released iron oxides from lysed cells were slightly more pronounced at 3T than at 1.5T (Figure 4). In the second experimental setup, which included all progenies, the iron content of the whole samples did not decrease significantly (p > 0.05).
At both, 1.5T and at 3T, the samples with viable labeled cells displayed markedly greater MR signal when compared to samples with lysed labeled cells, from which iron oxide particles would be liberated (Figure 5, images acquired at 3T). Viable Ferucarbotran-labeled cells showed a T2-effect on our moderately T2-weighted SE sequences while the same quantity of free iron oxides, that had been released from lysed cells, showed a markedly stronger T2-effect. Major signal reduction on SE images was also observed for short echo times as used on T1 sequences (Figure 4 and Figure 5).
Our data showed that (a) labeling hMSCs with Ferucarbotran at the level of roughly 7pg iron/cell did not attenuate viability of the cells over 12 days post labeling; (b) the amount of iron oxide taken up by the cells was reproducible over multiple labeling procedures; (c) the quantity of iron oxide inside single cells declined as the cells divided, but the total intracellular iron of all progeny was not diminished over this time; (d) substantial Ferucarbotran-induced effects on MR relaxation of hMSC solutions were evident up to 14 days after labeling; and (e) sonication-induced cell lysis of viable cells to disperse the internalized Ferucarbotran caused a marked decline in MRI signal intensity and increase in T2-relaxation rate. In our in vitro study, this effect could be used to differentiate intracellular iron oxides in viable cells from iron oxides, which had been released from lysed cells.
The large reduction in MRI signal intensity caused by disruption of cells and liberation of incorporated SPIO particles implies a big change in T2 relaxivity for extracellular versus intracellular particles. It is well known that relaxivity of susceptibility contrast materials is highly dependent on geometry and microscopic distribution of the magnetic centers. The observed differences on SE T2- versus GE T2*-weighted images in our study may be explained by many iron oxide particles being sequestered inside a few cells. Upon lysis of the cells the iron oxide particles disperse more uniformly throughout the sample causing a substantial reduction in distance between iron oxide particles. Using nominal density of 5.1 g/ml for iron oxide, iron fractional mass of approximately 72% and assuming a 4.2 nm sphere diameter for the iron oxide crystal in Ferucarbotran, we have an estimated iron mass of 1.43×10−19 g per crystal. Since the iron assay yielded approximately 7 pg of Fe loaded into each cell, we estimate roughly 4.9×107 crystals per cell. We do not know how many crystals comprise each SPIO particle, but if there are 8–10 crystals per particle it is still possible that distance between superparamagnetic centers, which would be proportional to the cubed root of the number of particles, may be reduced by roughly hundred-fold as a consequence of cellular disruption. Consequently an increased fraction of water molecules would be close to steep field gradients near the particles and would visit microscopic regions of the inhomogeneous magnetic field while diffusing during the TE interval.
On the other hand, the T2* weighted images could exhibit much less effect upon disruption because the static field inhomogeneity would be less altered by cell disruption (22,32). For example, if only a few very large magnetic particles were internalized into each cell, one would not expect to detect any change on T2 weighted SE images upon cell disruption. The observed signal characteristics are in accordance with previous studies: Simon et al. found increased R1 and R2-values for free versus intracellularly compartmentalized USPIO in vitro (20). In that study, monocytes were labeled with Ferumoxides USPIO particles and the cellular iron content was 10-fold less than the labeled hMSCs used in the current study. The difference in iron labeling can be attributed to a combination of factors including that monocytes are physically smaller cells than hMSCs, greater uptake efficiency for larger SPIO versus USPIO, and a larger per-particle iron content for SPIOs versus USPIOs. However the cell solutions used in the study of Simon et al contained roughly 10-fold more cells than the current study, so the total iron content of the cell solutions were similar. Yet in the former study uncompartmentalized USPIOs caused an increase of T1- and T2relaxation rates in comparison to intracellular USPIO, while in the current study disruption of cells containing SPIOs caused an overwhelming T2–effect that severely attenuated signal, such that T2 and T1 could not be evaluated and their relaxation rates could not be measured. In addition, Tanimoto et al. showed in vivo that R2* effects can be evoked by clustering of iron oxides, which would be comparable to the effect evoked by compartmentalization (33). To our knowledge, this is the first study that applies this effect for the differentiation of viable and lysed cells.
We found that the described different T2-signal characteristics of iron oxides in viable cells and iron oxides released from dead cells were consistently observed (A) at different time points over 14 days and (B) at two different field strengths, 1.5 T and 3T.
Within the 14-day observation period, we found consistently different signal characteristics of intracellular iron oxides and free iron oxides. It can be deduced that neither iron oxide metabolization nor cell proliferation altered this effect. The cellular iron content and T2-effect stayed nearly constant for the whole proliferating cell population and slowly decreased for individual cells. This can be explained by a distribution and dilution of the internalized iron oxides in proliferating progenies and minimal or absent metabolization of the contrast agent for up to 14 days after labeling. The positive staining for the dextran coat of the iron oxides for up to 4 weeks after labeling further supports the conclusion that no major metabolism of the contrast agent particles occurred during our experiment (Figure 2). Other authors described that they found a persistent T2-effect of iron oxide labeled mesenchymal stem cells in vivo 7 days after localized injection in the myocardium (34) or in the kidney after injection into the renal artery (9). Even after injection of highly proliferative neural stem cells detection was possible for up to 32 days (35). However, Walczak et al reported that the proliferation-induced dilution of the contrast agent in cell lines with a high rate of proliferation - neural stem cells in that case - limits the long-term detectability by MR imaging (36). In slowly proliferating cell types like the mesenchymal stem cells used in this study, this effect should not affect the detectability.
Studies with longer follow up intervals showed that iron oxides are slowly metabolized within the lysosomes and incorporated into the body’s iron metabolism (37). Arbab and coworkers reported, that the intracellular iron oxide metabolization is dependent on the contrast agent type, the applied labeling technique and the pH of the cells cytoplasm (38). All these discussed mechanisms, cell proliferation, dilution in progenies, iron metabolization, and potential cellular iron elimination, are expected to result in a slow decline in cellular iron content, and consequently, a decline in the cells T2-signal effect over time. We hypothesize, that any increase in T2-effect at the transplantation site after local transplantation of labeled cells would be an indicator of iron oxides release and cell death.
Another important factor for potential in vivo applications will be the proton density and proton diffusion capacity within the target organ. Since the observed T2-effects are apparently dependent on the degree of interaction between protons and iron oxides, our results are more likely reproducible in vivo in a proton-rich pathology (e.g. edema) than in a proton-deprived environment.
In vivo applications of our technique are also dependent on additional biologic factors. Our technique to differentiate viable and dead cells is based on differences in T2-effects and dependent on a lysis and iron release of dying cells. There are two major mechanisms how cell death can occur – necrosis and apoptosis (39). Necrosis results in cell lysis and subsequent inflammatory response (40). It is to be expected that processes resulting in necrosis and cell lysis will produce a similar signal behavior as we have shown. On the other hand, apoptosis is a regulated cell death where controlled degradation occurs leading to fragmentation of the cell. This type of cell death will result in the same iron content being compartmentalized among many cell fragments. This would be similar to the behavior seen in our scans with increasing cell counts – the R2 relaxation rate would be the parameter most sensitive to change in this case. In many instances, both mechanisms may be involved after cell transplantation.
In addition, necrotic or apoptotic cells as well as released iron oxide particles may undergo a secondary phagocytosis by macrophages. The relation between the number of lysed cells and the number of macrophages would determine the resulting rate of compartmentalization and thereby a possible contrast agent effect. Therefore, the consistency of the surrounding tissue (fluid necrosis vs solid scar tissue) in which the release and the re-compartmentalization of iron oxide particles takes place and the number of macrophages will determine the possibility to detect a graft failure. Thus in vivo applications will be more complex and will have to evaluate various different pathologies in order to clarify, if the results from our in vitro model are applicable for certain pathologies and target organs in vivo.
In conclusion, our in vitro data show a persistent contrast agent effect and an increase in T2-relaxation rate induced by cell lysis. In our in vitro study, this effect could be used to differentiate intracellular iron oxides in viable cells from iron oxides, which had been released from lysed cells. If these results could be applied in vivo, it would allow for a non-invasive diagnosis of a treatment failure after hMSC transplantation. An increased T2-effect at the transplantation site would then be indicative of graft failure.
This work was supported by a seed grant from the Department of Radiology, University of California of San Francisco. Tobias D Henning was supported by a research stipend from the German Research Association (DFG, HE 4578/1).