Stem cells offer a promising approach to the treatment of myocardial infarction and prevention of heart failure. We have used iron labeling of bone marrow stromal cells (BMSCs) to noninvasively track cell location in the infarcted rat heart over 16 weeks using cine-magnetic resonance imaging (cine-MRI) and to isolate the BMSCs from the grafted hearts using the magnetic properties of the donor cells. BMSCs were isolated from rat bone marrow, characterized by flow cytometry, transduced with lentiviral vectors expressing green fluorescent protein (GFP), and labeled with iron particles. BMSCs were injected into the infarct periphery immediately following coronary artery ligation, and rat hearts were imaged at 1, 4, 10, and 16 weeks postinfarction. Signal voids caused by the iron particles in the BMSCs were detected in all rats at all time points. In mildly infarcted hearts, the volume of the signal void decreased over the 16 weeks, whereas the signal void volume did not decrease significantly in severely infarcted hearts. High-resolution three-dimensional magnetic resonance (MR) microscopy identified hypointense regions at the same position as in vivo. Donor cells containing iron particles and expressing GFP were identified in MR-targeted heart sections after magnetic cell separation from digested hearts. In conclusion, MRI can be used to track cells labeled with iron particles in damaged tissue for at least 16 weeks after injection and to guide tissue sectioning by accurately identifying regions of cell engraftment. The magnetic properties of the iron-labeled donor cells can be used for their isolation from host tissue to enable further characterization.
A novel approach to the treatment of myocardial infarction and prevention of heart failure is cell grafting in the damaged myocardium. Because it is safe and relatively easy to obtain autologous adult stem cells from patients' blood, bone marrow, or skeletal muscle, cell therapy is currently being applied in the clinic (reviewed in ). Clinical trials using mononuclear cells, bone marrow stromal cells (BMSCs) or bone marrow-derived mesenchymal cells, and skeletal myoblasts are completed or under way [2, , , , , , , , , –12]. Initial results look promising, with ejection fractions improving significantly by approximately 5% and cardiac perfusion increasing after cells were infused or injected into the myocardium [2, , , , , , , , , –12].
Experimental models of cell-based therapy for myocardial infarction are being used to identify the cellular mechanisms underlying the functional improvement observed in patients (reviewed in ). To optimize cell-based therapy, the tissue distribution of the administered cells should be known. Donor cells can be identified in tissue sections if they have been prelabeled with fluorescent dyes or particles ; if they are genetically different from the host, such as male cells into females  or human cells into mice ; or if genetically modified to express LacZ  or green fluorescent protein (GFP) . However, because these methods require the sacrifice of the animal, they do not permit serial in vivo monitoring. Methodologies that can track donor cell retention and migration in vivo are needed to optimize stem cell-based therapy. Labeling stem cells with the radioactive compound 111In [19, 20] can show distribution of intravenously delivered cells in live animals, but poor spatial resolution makes specific localization difficult and the radiolabel decays rapidly (In111 1/2 t = 2.8 days), precluding cell tracking studies over several weeks. Cells labeled with particles of iron oxide can be identified in vivo as hypointensities in magnetic resonance (MR) images , as the iron shortens transverse proton relaxation times . The spatial and temporal resolution of magnetic resonance imaging (MRI) has allowed the location of iron-labeled donor cells to be monitored noninvasively over several weeks in heart  and brain . Here, we demonstrate that repeated MR imaging can be used to track iron-labeled BMSCs in the infarcted rat heart over several months and that the magnetic properties of the donor cells can be used to reisolate them from the infarcted heart.
Materials and Methods
Female Wistar rats (200 g) (Harlan UK Ltd., Bicester, U.K., http://www.harlan.com/UK/) were allowed free access to standard rodent chow and water throughout the study. The University of Oxford Animal Ethics Review Committees and the Home Office (London) approved all procedures performed in this study.
Isolation and Culturing of Rat BMSCs
BMSCs were isolated from the tibia and femur of 8-week old male Wistar rats (Harlan UK Ltd.) as described by others . Cells were plated in Dulbecco's modified Eagle's medium (Gibco-BRL, Paisley, U.K., http://www.gibcobrl.com) containing 10% (vol/vol) fetal calf serum (Gibco-BRL), and 7% (vol/vol) horse serum (PAA Laboratories, Somerset, U.K., http://www.paa.at). After 24 hours, nonadherent cells were removed, and the adherent population was cultured at 37°C in 5% CO2 until 80% confluent. Cells used in all experiments were at passage 2.
Flow cytometric analysis of BMSCs was performed using primary (BD Pharmingen, San Diego, http://www.bdbiosciences.com/pharmingen), or isotype matched control (Dakopatts, Denmark) antibodies and detected with a fluorescein isothiocyanate-conjugated secondary antibody (Invitrogen, Paisley, U.K., http://www.invitrogen.com) using a BD LSR II flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com). BMSCs were transduced with a GFP lentiviral vector (LV-GFP) as described previously  (gift from Prof. Adrian Thrasher, Institute of Child Health, London, U.K.) and labeled with iron particles. On the day prior to their injection, BMSCs were incubated with 200 μl of iron particles (Bangs Laboratories Inc.), resulting in approximately 95% of BMSCs internalizing approximately 60 iron particles per cell (iron-labeled BMSCs [Fe-BMSCs]) over 24 hours. Each particle contained 0.125 pg of iron; thus, each cell contained 7.5 pg of iron. The particles consisted of 0.9-μm polystyrene beads containing 62% magnetite in a central core and a fluorescent coating to allow the particle to be located in tissue sections. In our hands, the fluorescence from the particles was quenched during tissue section preparation; hence, BMSCs were transduced with LV-GFP. The labeled BMSCs were removed from culture using trypsin/EDTA (5 minutes at 37°C in 5% CO2), pelleted by centrifugation, and resuspended at 107 cells per ml in phosphate-buffered saline (PBS) at 4°C ready for transplantation.
MRI of Cell Suspensions
Suspensions of 105 BMSCs in 600 μl of 1% (w/v) agarose (Sigma-Aldrich, U.K., http://www.sigmaaldrich.com), containing either 100%, 1%, or 0% Fe-BMSCs were pipetted into Eppendorf tubes. To prevent the susceptibility effects that arise from the air-glass interface distorting the image, Eppendorf tubes were suspended in a glass tube and surrounded by agarose. Phantoms were then imaged at 500 MHz using T1- (spin echo, echo time/repetition time [TE/TR] = 0.02/0.5 second), T2- (fast spin echo TE/TR = 0.16/3 second), and T2*- (gradient echo TE/TR = 4.6/9.2 ms) weighted sequences.
The left anterior descending (LAD) coronary artery was occluded as described previously . Briefly, Wistar rats were anesthetized with 2.5% isoflurane in O2, intubated, and maintained at 80 to 90 breaths per minute with a tidal volume of 2–3 ml. A 3-cm incision was made, and the muscle was blunt-dissected to reveal the ribs. A thoracotomy was performed in the fourth intercostal space. Retractors were inserted, and the pericardium was removed. The LAD coronary artery was ligated approximately 2 mm from the origin using a 5-0 proline suture.
Ten minutes after coronary artery ligation, four injections of 1.25 × 105 iron particle-labeled and LV-GFP-labeled BMSCs in 10 μl of PBS were made into the border zone of the infarcted tissue using a 27-gauge needle (n = 12). In six animals, the same procedure was performed using nonlabeled BMSCs (n = 2) or 10 μl of iron particles (n = 2) into infarcted hearts or Fe-BMSCs into noninfarcted hearts (n = 2).
Cardiac cine-MRI was performed as previously described . Briefly, rats were anesthetized with 2.5% isoflurane in O2 and positioned supine in a purpose-built cradle. ECG electrodes were inserted into the forepaws, and a respiration loop was taped across the chest. The cradle was lowered into a vertical-bore, 500 MHz, 11.7 T MR system with a Bruker console running Paravision 2.1.1 and 60-mm birdcage coil. ECG and respiration trigger levels were adjusted so that acquisitions were triggered at the same point in the cardiac cycle.
A stack of contiguous 1.5-mm true short-axis ECG and respiration-gated cine images (field of view, 51.2 mm2; matrix size, 256 × 256; TE/TR, 1.43/4.6 ms; 17.5° pulse; 25 to 35 frames per cardiac cycle) were acquired to cover the entire left ventricle. Image analysis was performed using Scion Image 4.0.2, and ejection fraction was calculated as the sum of the stroke volume over the end diastolic volume for each short-axis slice. Long-axis two-chamber and four-chamber images were also acquired. The entire imaging protocol was performed in approximately 40 minutes.
High-Resolution Three-Dimensional MR Microscopy
Hearts were excised, fixed in paraformaldehyde (Sigma-Aldrich), and embedded in 1% agarose doped with gadolinium diethylenetriaminepentaacetic acid in a 20-mm NMR-tube. High-resolution MRI was performed in a 20-mm quadrature-driven birdcage coil (Rapid Biomedical, Würzburg, Germany) using a fast gradient echo sequence (TE/TR = 1.8/30 ms; field of view, 20 × 20 × 20 mm; matrix size, 256 × 256 × 512; number of averages, 4). Data reconstruction was performed as described by Schneider et al. .
Whole hearts were fixed in phosphate-buffered saline (pH 7.2) containing 4% (w/v) paraformaldehyde and embedded in paraffin, and serial 5-μm sections were cut. Cells expressing GFP were detected using a monoclonal mouse anti-GFP antibody, clone GEP-20 (Sigma-Aldrich). Immunoreactive cells were visualized using the EnVision+ System-HRP (diaminobenzidine) kit (Dakopatts) according to the manufacturer's instructions. Macrophages were identified by CD68 expression (Serotec Ltd., Oxford, U.K., http://www.serotec.com). Sections were counterstained in hematoxylin solution (Sigma-Aldrich) and photographed using a Zeiss Axiovision camera (Carl Zeiss, Jena, Germany, http://www.zeiss.com) attached to a Nikon Eclipse E600 microscope (Nikon, Kingston upon Thames, U.K., http://www.nikon.co.uk).
Enzymatic Isolation of Iron-Containing Cells
Hearts were rapidly excised and mounted on a cannula attached to Langendorff apparatus and equilibrated with Krebs-Henseleit solution consisting of 119 mM NaCl, 4.7 mM KCl, 0.94 mM MgSO4, 1.2 mM KH2PO4, 25 mM NaHCO3, 11.5 mM glucose, and 1 mM calcium for 5 minutes at 37°C. Low-calcium medium containing 119 mM NaCl, 5.4 mM KCl, 5 mM MgSO4, 5 mM pyruvate, 20 mM glucose, 20 mM taurine, 10 mM HEPES, and 5 mM nitrilotriacetic acid (NTA), at pH 6.96 and containing 12–15 mM calcium was perfused through for 5 minutes, also at 37°C. This low-calcium solution was changed to one of similar composition but without NTA and with 200 mM calcium added. Collagenase (1 mg/ml) (Worthington, Berkshire, U.K., http://www.worthington-biochem.com/) and hyaluronidase 0.6 mg/ml (Sigma-Aldrich) were added, and perfusion continued for an additional 10 minutes. The ventricles were chopped and incubated in the same enzyme solution for 2 × 10 minutes. The medium was shaken gently throughout this incubation and kept under an atmosphere of 100% O2. The dispersed cells were strained through 300-mm gauze to remove tissue. Myocytes were allowed to pellet under gravity for 10 minutes. The supernatant was removed into a 10-ml Falcon tube, and a Neodymium-iron-boron magnet (6-mm diameter, 0.3 T; Maplin, Barnsley, U.K., http://www.maplin.co.uk) was positioned close to the apex of the tube. After 20 minutes, the pellet formed was resuspended in 1–2 ml of embryonic stem medium and transferred to the stage of a Nikon TE200 microscope for fluorescence photography.
Results are shown as means ± standard errors. Differences were considered significant at p < .05.
Isolation, Culture, and Labeling of Rat BMSCs
Rat BMSCs were identified as a fibroblastic population and were characterized by flow cytometry. Cells were negative for CD4, CD11b/c, CD31, CD45R, and CD49d and positive for CD90 (Thy1) (83% ± 3%). To track the transplanted cells in vivo, passage 2 rat BMSCs were double-labeled with iron particles and LV-GFP (Fig. 1A, 1B). Approximately 95% of cells were labeled with iron particles, and 50% expressed GFP. The iron particles did not autofluoresce, as demonstrated by the lack of green fluorescence in Figure 1B from iron-labeled cells that were not transduced with GFP. Cell viability and proliferation were unaffected by the labeling process (data not shown), as previously reported .
MRI of Cell Suspensions
To determine the effect Fe-BMSCs had on MR images, 105 Fe-BMSCs were suspended in 600 μl of agarose, a volume similar to the rat left ventricle, and imaged at 11.7 Tesla. T1-, T2-, and T2*-weighted MR images were acquired, and the size of Fe-BMSC induced signal voids was measured. As expected, the signal void was largest in T2*-weighted images and smallest in T1-weighted images (data not shown). The percentage of Fe-BMSCs in the sample was reduced to identify the limit of detection. It was found that 105 labeled cells caused a large signal void in T2*-weighted images (Fig. 1C). When only 1% of the 105 BMSCs were iron-labeled, small regions of signal void were detected throughout the phantom (Fig. 1D). When none of the BMSCs were labeled, there were no signal voids in the agarose (Fig. 1E). Hence, as few as 1,000 cells in 600 μl of agarose can cause signal voids detectable by T2*-weighted MRI.
MRI Detection of Fe-BMSC Injections
Four injections of 1.25 × 105 Fe-BMSCs in 10 μl of PBS were made into the border zone of the infarcted heart tissue 10 minutes after coronary artery occlusion. MRI at 1 week identified hypointense areas at the predicted sites of injection, confirming that injections were successful. In most animals, two or three of the four injections sites could be located in a stack of images acquired from the base of the heart to the apex (Fig. 2A–2G). Animals were sacrificed after MRI, and hearts were fixed and sectioned. GFP+ cells were identified at locations in the heart where MRI had detected signal voids (Fig. 2H, 2I).
In Vivo MRI at 1, 4, 10, and 16 Weeks Post-Myocardial Infarction and Post-Fe-BMSC Injection
An additional 12 rats were infarcted and received Fe-BMSC injections postinfarction. Signal voids caused by the Fe-BMSCs were detected in the left ventricular wall of all hearts that received Fe-BMSCs at each time point studied (Fig. 3A–3D), with the average injection success rate being approximately 63%. After 16 weeks, animals were sacrificed, and hearts were imaged ex vivo using high-resolution three-dimensional (3D) MR microscopy. Signal voids were identified at the same position in the MR microscopy images as in vivo (Fig. 3E). The 3D nature of this acquisition allowed accurate identification of the location of the iron-labeled cells in the heart, permitting tissue sectioning to be targeted. No voids were identified in hearts that received non-iron-labeled BMSCs (Fig. 3F, 3G). When Fe-BMSCs were injected into noninfarcted hearts, or 10 μl of iron particle solution was injected into the left ventricular wall, signal voids in images were detected at 1 day and 1 week but had disappeared by 4 weeks (data not shown).
Identification of Iron Particle-Labeled and GFP-Expressing Cells Confirmed MRI Tracking of Donor Cells
The presence of iron in grafted hearts surrounding the areas of injection did not guarantee that transplanted BMSCs had survived in the myocardium or that they had integrated into the heart muscle tissue. Iron particles may have been released by BMSCs undergoing apoptosis and taken up by neighboring cells. To confirm that the MRI tracking in vivo correlated with the presence of GFP-expressing transplanted BMSCs, immunohistological analysis was carried out on hearts 1 week (discussed above) and 16 weeks after cell injection. Tissue sectioning of hearts taken 16 weeks after cell delivery was guided using the in vivo and ex vivo MR images. Sections were stained with hematoxylin, and iron particles could be identified in cells at the infarct periphery (Fig. 4A–4D). Adjacent slices were stained with anti-GFP antibody and GFP-expressing cells were present at the same position as the iron containing cells (Fig. 4E–4H). Sections stained for the macrophage marker CD68 revealed large numbers of macrophages in sections taken at 1 week, but very few were detected at 16 weeks. Figure 2J and 2K illustrates that the majority of macrophages present in the 1-week sections did not contain iron and that most of the iron containing cells did not express CD68.
In a different set of rats, Fe-BMSCs were administered using the same methodology, and MRI was performed to determine the presence of iron-labeled cells. At 6 and 20 weeks, hearts were digested with collagenase, and iron-labeled cells were separated from host tissue using a magnet to confirm that, even 20 weeks after cell administration, iron particles were present in the donor cells that expressed GFP (Fig. 4K, 4L). The morphology of the isolated cells was similar to that of the BMSCs prior to injection (Fig. 4I, 4J).
The Signal Void Arising from Iron-Labeled Donor Cells Was Inversely Related to the Ejection Fraction
The size of the hypointense region significantly decreased over time, indicating that donor cells were being lost from the heart (Fig. 5A). The greatest reduction occurred between 1 (51 ± 8 mm3) and 4 (37 ± 6 mm3) weeks (p < .01) with a further reduction by 16 (28 ± 5 mm3) weeks (p < .05). Cardiac function was monitored in every animal, but no significant improvements in left ventricular ejection fraction were found at any of the time points studied. However, there was a strong inverse correlation between the size of the signal void arising from the Fe-BMSCs and the ejection fraction at weeks 10 (Fig. 5B) and 16 (p < .01) but not at weeks 1 or 4. The void volume in the six moderately infarcted animals (ejection fraction >62%) steadily decreased over the 16 weeks, whereas the volume of the signal void in the severely infarcted animals (ejection fraction <62%) remained constant between 4 and 16 weeks (Fig. 5C), indicating that a greater number of donor cells were retained in hearts that had the greatest damage.
The ideal method for noninvasively tracking stem cells in live animals would permit accurate monitoring of donor cell distribution at multiple time points after administration, so that the optimal cell delivery time after infarction and most efficient administration technique could be established. Tracking of radioactively labeled cells has shown the tissue distribution of donor cells [19, 20], but the method suffers from poor resolution. With greater spatial resolution, MRI has been used to detect iron-labeled stem/progenitor cells in live pig [23, 31] and rodent [32, 33] hearts. Himes et al.  located 3 × 105 superparamagnetic-iron oxide-labeled embryonic stem cells in mouse heart up to 5 weeks after administration, and Cahill et al.  tracked iron-labeled myoblasts up to 4 weeks after delivery.
Here, we used MRI to track Fe-BMSCs for 16 weeks after administration to the infarcted rat heart. We detected labeled BMSCs using a cine-MR sequence that acquired between 25 and 35 frames per cardiac cycle. This frame number allowed the tracking of signal voids as they moved with the heart during contraction, which slightly altered the location of the MRI slice. The signal voids could be measured in systole and diastole, and iron-labeled cells could be tracked within the time required to monitor heart function, meaning that no additional scanning time was needed for the cell tracking studies.
Injecting cells into the beating rat heart is difficult in that cells are lost because the contracting myocardium forces them back up the needle track and out of the heart, or they can be injected through the thin left ventricular wall and into the ventricular cavity. It is important that the success of injections can be determined so that improved cardiac function can be related to the number of cells delivered. A group with extensive experience in cell administration has reported that only 54% of their injections of radiolabeled cells into the mouse heart were successful . Here, MRI was able to confirm injection success. In every animal, at least one of the four injections was observed, with the average injection success being 63%. We were also able to accurately locate several individual injection sites in the hearts.
Two methodologies were used to confirm that the iron particles remained in the grafted donor cells over the 16 weeks. Firstly, tissue sectioning was targeted to regions of donor cell engraftment using the in vivo and ex vivo MR images. Cells labeled with iron particles and GFP were found in the regions of signal void in the MR images, and iron particles and GFP staining were identified in the same regions of adjacent 5-μm slices. Further confirmation came from whole-heart digests from which magnetic, iron-labeled, and GFP-expressing donor cells were separated. In general, the morphology of the donor cells magnetically isolated from hearts grafted with Fe-BMSCs remained similar to that of the pre-implanted cells, whether grafted 6 or 20 weeks earlier. Labeling cultured cells with magnetic antibodies against cell surface markers is a common method for separating and sorting cells, but to our knowledge this is the first time that internally labeled donor cells have been isolated from digested whole hearts using their magnetic properties. This should be a useful method for determining donor cell numbers in hearts that received cell injections, and it could be used to isolate and characterize the specific and rare cell populations that remain engrafted for long periods after cell administration.
This MRI tracking technique, although not truly quantitative, can be used to determine cell retention in live animals. Here, a significant inverse correlation between the size of the signal void and the ejection fraction of the hearts was observed after 10 weeks, with the greater number of Fe-BMSCs retained in hearts with more impaired function. This relationship was not observed at 1 or 4 weeks, indicating that cell retention at these early time points had a nonspecific component.
Subpopulations of BMSCs have the ability to differentiate into cardiomyocytes  and improve cardiac function after administration to infarcted rats [36, 37]. However, we did not observe any improvement in ejection fraction over the 16-week experiment, perhaps because the degree of infarction in these animals was relatively low and variable.
Here, we show that MRI can track rat BMSCs for up to 16 weeks after injection into the infarcted rat heart, yield information on stem cell retention that can be related to cardiac function, and identify individual injection sites and regions where administered cells have engrafted so tissue sectioning can be targeted. This methodology could help optimize cell therapy techniques, being a useful tool for identifying the best time and route of cell administration and the cell type with the greatest regenerative capacity. The recent development of safe iron-based contrast agents approved for clinical use  makes it a possibility that MRI can be used to track stem and progenitor cells in patients and give valuable information concerning the location of injected cells in human stem cell trials. Finally, we have used a novel method for the isolation of live engrafted iron-labeled cells from the heart several months after cell administration using their magnetic properties, which has allowed the characterization and culture of the donor cells after engraftment. This technique would be useful for stem cell and cell therapy research in general.
The authors indicate no potential conflicts of interest.
We thank Peter O'Gara for assistance with the enzymatic heart digests and cell isolation experiments and Dr. Craig Lygate for assistance in establishing the rat infarction model. This work was supported by the British Heart Foundation and had funding from the National Health Service Research & Development Directorate.