Mesenchymal stem cells (MSCs) have shown promising results as a cellular therapeutic. Many studies involving MSCs aim to restore damaged tissues, including bone, cartilage, tendon, adipose, and muscle tissue, through tissue regeneration (1, 2). Moreover, several proposed therapies rely on the pleiotropic effects that MSCs impose on their local microenvironment through the release of extracellular vesicles, cytokines, and tropic factors (3–5). MSCs have been shown to exert antimicrobial effects, promote local vascularization and cell growth, and modulate inflammation (1, 2, 6). MSC survival and engraftment in vivo is critical in determining therapeutic outcomes. Unfortunately, many MSCs undergo apoptosis in the days following administration owing to the stresses of administration and subsequent lack of nutrients (7, 8). Apoptotic stem cells release cytokines that attract immune cells (predominately macrophages) to the implant site. A high influx of these cells can ultimately trigger stem-cell rejection (8). The potential of MSC therapy is limited by MSC death and immune rejection; therefore, the development of a technique to quantitatively monitor MSC engraftment and ensuing inflammation over time would be invaluable for evaluating the course of therapy.
Many experimental studies of MSC engraftment have been conducted using histology, which provides detailed molecular and morphological information but is limited to the interrogation of a single time point and portion of tissue. Alternatively, cellular magnetic resonance imaging (MRI) has proven to be an effective technique for noninvasive and longitudinal cell tracking (8–10). To date, most cellular MRI involves tracking cells, which are labeled with superparamagnetic iron oxide (SPIO) nanoparticles. In proton (1H) MRI images, SPIO-labeled cells appear as regions of signal void. In a uniform magnetic field, SPIOs alter the net local magnetization that nearby 1H atoms experience and this leads to increased R2* relaxation rates of these 1H atoms. These voids occupy a volume that is greater than the labeled cells, an effect referred to as the blooming artifact. This effect can lead to enhanced sensitivity of cell detection (11), but it poses challenges for accurate quantification of cell number. Measuring the volume of signal voids is one metric to estimate the number of cells present; however, this is not a direct relationship. One other limitation of SPIO-based cell tracking is lack of specificity in some tissues (9, 12). There is some ambiguity when identifying these cells in vivo, as other regions in anatomic MRI appear dark (ie, the air-filled lungs).
Ultrasmall superparamagnetic iron oxides (USPIOs) are a subset of iron oxides used for MRI cell tracking. These nanoparticles are ∼30 nm in diameter and are coated in dextran, and thus, they are biocompatible and biodegradable. Ferumoxytol is one such USPIO that is FDA-approved for iron replacement therapy for the treatment of anemia in patients and may be used off-label for iron-based MRI cell tracking (13, 14). In this study we will use ferumoxytol for iron oxide cell labeling of MSC, as it is the most clinically applicable agent.
Fluorine-19 (19F) MRI cell tracking is an alternative to iron oxide-based MRI. In this technique, cells are labeled with perfluorocarbon (PFC) nanoemulsions and detected with 19F MRI in a hotspot image. Since there is little endogenous 19F within biological tissues, these cells can be visualized with high specificity. The signal intensity of these images is directly linear to the number of 19F atoms, which allows for the quantification of 19F labeled cells in vivo (8, 9, 15). One limitation of 19F-based cell tracking is that the sensitivity of detection is much less than iron oxide agents, requiring a minimum of 103 - 105-labeled cells per imaging voxel. PFC agents are clinically approved for cell tracking (16).
Magnetic particle imaging (MPI) is an emerging modality that directly detects SPIO nanoparticles. Similar to 19F MRI, MPI produces positive contrast images of the distribution of labeled cells. MPI signal is linearly related to the quantity of iron oxide tracer that allows for accurate quantification of SPIO-labeled cells (13, 17). It is not feasible to achieve this reliable specificity and quantification in SPIO-based MRI cell tracking, although SPIO-based MRI cell tracking may have superior detection sensitivity depending on tissue contrast. MPI has the potential to overcome the limitations of 19F MRI (sensitivity) and iron-based MRI (specificity and quantification).
In this study, we use ferumoxytol for labeling and detecting MSCs with MPI, which has recently been shown to permit quantification of MSCs transplanted in a mouse model of osteoarthritis (13). This group also showed that MPI of ferumoxytol-labeled cells was sensitive to changes in cell number in vivo over time, whereas the voids detected in 1H MRI did not detect this change.
One strategy for detecting immune cells in vivo involves the intravenous administration of the labeling agent. This leads to uptake of the agent by phagocytic cells of the reticuloendothelial system (predominately macrophages) in vivo (8, 10, 18). MRI is typically performed 1 day after the administration of the cell labeling agent to permit for the clearance of intravascular agent and the accumulation of the label into cells.
Our aim is to combine the use of iron-based MRI, 19F MRI, and MPI cellular imaging technologies to monitor and quantify the persistence of transplanted MSCs and infiltrating macrophages in vivo. These 3 modalities are complementary and provide additional information (specificity, sensitivity, and quantification of cell number) when integrated together. We explored the ability to label, detect, and quantify MSCs with ferumoxytol for detection in 1H MRI and MPI images, and infiltrating macrophages with PFC for 19F MRI detection.
MSCs derived from the bone marrow of C57BL/6 mice (MUBMX-01101 [BE], Cedarlane, Burlington, Ontario, Canada) were cultured at 37°C and 5% CO2 in low-glucose Dulbecco’s modified Eagle’s medium (Thermo Fisher Scientific, Waltham, Massachusetts). MSCs were passaged every 2 - 3 days for 10 days. When 80%– 90% confluence was met, these cells were labeled with 50.4 μg Fe/mL ferumoxytol (Feraheme, AMAG Pharmaceuticals), along with 60 μg/mL protamine sulfate and 3 U/mL heparin (Sandoz Canada), as per protocol by Thu et al. (19). After overnight incubation, the cells were washed 3 times with phosphate-buffered saline (PBS) before and after trypsin dissociation. Cell counting and viability was determined using the trypan blue exclusion assay (using Countess Automated Cell Counter; Invitrogen). Further, 2 × 105 ferumoxytol and unlabeled cells were collected for Perls’ Prussian Blue (PPB) staining to assess iron labeling (20). Cell pellets (1 × 106 MSCs) were collected to determine iron content using MPI (see below).
15 C57B1/6 mice (Charles River, Canada, Pittsburgh) were obtained and cared for in accordance with the standards of the Canadian Council on Animal Care, under an approved protocol by the Animal Use Subcommittee of Western University’s Council on Animal Care. 1 × 106 UPSIO-labeled MSCs in 25-μL PBS were implanted to the hind limb muscle of 10 immune-competent C57B1/6 mice (day 0). A second cohort of control mice (n = 5) received 1 × 106 unlabeled MSCS in 25-μL PBS. Immediately after MSC implantation, mice were administered 200-μL PFC agent (V-Sense, CelSense Inc.) intravenously via the tail vein to label phagocytic immune cells in situ. Injections were performed under 2% isoflurane in 100% oxygen anesthesia.
1H/19F MRI Acquisition
One and 12 days following MSC implantation, in vivo MRI images of all mice (n = 15) were acquired on a 3 T clinical scanner (Discovery MR750, General Electric) using a 4.31- × 4.31-cm-diameter 1H/19F dual-tuned RF surface coil (Clinical MR Solutions, Wisconsin) as previously described (21). Both 1H and 19F images were acquired with 3-dimensional (3D) balanced steady-state free precession (bSSFP) sequences. Mice were anesthetized with 2% isoflurane in 100% oxygen during these scans.
1H imaging parameters were as follows: field of view (FOV) = 60 × 30 mm; matrix size = 150 × 75, slice thickness = 0.4 mm (0.4 × 0.4 × 0.4 mm3 resolution), repetition time (TR)/echo time (TE) = 12.8 / 6.4 milliseconds, flip angle (FA) = 20°, bandwidth (BW) = ±31.25 kHz, phase cycles (PC) = 12, number of excitations (NEX) = 1, total scan time = 9 minutes. 19F imaging parameters were: FOV = 60 × 30 mm, matrix = 60 × 30, slice thickness = 1 mm (1 × 1 × 1 mm3 resolution), TR/TE = 5.6 / 2.8 milliseconds, FA=72°, BW = ±10 kHz, PC = 1, NEX = 100, total scan time = 18 minutes.
Quantification of Signal Loss Due to Iron-Labeled MSC
1H images were assessed for abnormal regions of signal loss (dark regions), in the area where MSCs were implanted. These regions were manually contoured over multiple 1H slices to measure the signal void volume (Osirix, Pixemo SARL, Bernex, Switzerland).
19F Signal Quantification
19F images were overlaid onto 1H images (Osirix, Pixemo SARL) for anatomical reference. 19F signal in the limb ipsilateral to the MSC implant was manually delineated and quantified relative to reference tubes of known 19F content (3.33 × 1016 19F/μL). Owing to the presence of phagocytic immune cells in the bone marrow (BM) and lymph nodes (LN), any 19F signal in the contralateral limb was subtracted from the ipsilateral limb, to isolate the 19F signal detected in the region of MSC implantation.
Evaluation of Ferumoxytol as an MPI Tracer
The particle relaxometer module (RELAXTM) on the MOMENTUMTM system (Magnetic Insight Inc., Alameda, California) was used to characterize ferumoxytol as an MPI tracer. In this mode, the localizer gradient field is switched off and a negative magnetic field is turned on and then switched to a positive field (and vice versa). As a result, iron nanoparticles are driven from a negative magnetic saturation to positive (positive scan) and vice versa (negative scan). This measures the point spread function (PSF) of the nanoparticles and allows for measurements of properties such as full-width half-maximum (FWHM) (spatial resolution) and signal per gram of iron (sensitivity) (22, 23). We used FWHM to define tracer resolution, according to the Houston criterion (24). The shift between the positive and negative PSF is a result of magnetic relaxation as described in some studies (23, 25). We acquired PSFs for 30 μg and 5.5 μg (in 1 μL) ferumoxytol.
For 5 of the 10 mice that had ferumoxytol-labeled MSC implanted and that were imaged with MRI, full-body in vivo MPI images of mice were acquired. Image acquisition occurred on days 1, 5, and 12 following MSC implantation, on a MOMENTUMTM system (Magnetic Insight Inc.) using the 3D isotropic mode. In this mode, tomographic images were acquired using a 5.7 T/m gradient, 35 projections, 1 average, in an FOV of 12 × 6 × 6 cm, for a total scan time of ∼1 h per mouse. We have included day 5 images as an explorative, intermediate time point, to assess MPI detection of MSCs over time. Mice were anesthetized with 2% isoflurane in 100% oxygen during these scans. 3D isotropic images of MSC pellets were acquired using the same parameters on day 0.
MPI Calibration and Signal Quantification
Calibration lines were produced to determine the relationship between iron content in ferumoxytol (30 mg/mL) and MPI signal. To construct this line, 1 μL samples of ferumoxytol were scanned in the same mode as in vivo images (3D isotropic). Samples of iron content over 2 orders of magnitude were tested: 0.75 μg, 1.125 μg, 1.5 μg, 2.25 μg, 3 μg, 7.5 μg, 11.25 μg, 15 μg, 22.5 μg, and 30 μg iron.
All MPI images were displayed in full dynamic range and assessed for MPI signal corresponding to the samples (calibration samples, MSC pellets, or MSCs in vivo) with spatial reference to fiducial markers (Osirix, Pixemo SARL). To quantify the MPI signal in a specific region of interest (ROI), a 3D semi-automatic segmentation tool was used. Before delineating these ROIs, the window/level (W/L) was adjusted to each specific region, such that the full dynamic range of this region was displayed. Total MPI signal for the delineated volume was calculated by multiplying mean signal by volume. With samples of increasing iron content, both MPI signal and volume of the ROI increase. Total MPI signal was plotted against iron content to derive calibration lines, and this relationship was used to quantify iron content in MSC pellets and MSCs in vivo. All MPI images, including the calibration, MSC pellets, and in vivo images, were delineated and analyzed in the same way to ensure consistency.
Following the last imaging sessions on day 12, mice were euthanized by isoflurane overdose and then perfusion-fixed with 4% paraformaldehyde. The right limb muscle of mice was excised and paraffin-embedded. Embedded tissues were sectioned (5 μm in thickness) every 400 μm to ensure entire sampling of the tissue. These sections were stained with hematoxylin and eosin for general tissue morphology, PPB and nuclear fast red counterstain to identify the presence of ferumoxytol, or F4/80 immunohistochemical staining to identify macrophages. For F4/80 staining, sections underwent antigen retrieval in sodium citrate buffer, permeabilized using 0.4% Triton X-100 (Sigma-Aldrich, Oakville, Ontario, Canada), followed by overnight incubation in rat antimouse F4/80 primary antibody [1:100 dilution] (ab16911, Abcam). The next day, sections were incubated with biotinylated goat antirat IgG antibody [1:300 dilution] (BA-9401, Vector Laboratories) and then processed with ABC solution (PK4000, Vector Laboratories, Burlington, Ontario, Canada). Lastly, the slides were incubated in 3,3′-diaminobenzidine tetrahydrochloride (DAB) substrate solution (SK-4100, Vector Laboratories) and counterstained with hematoxylin. Histological images were acquired on the EVOS Imaging System (M7000, Thermo Fischer Scientific).
Linear correlations were conducted between total MPI signal and iron content to determine Pearson’s correlation coefficient. Student t-tests were used to evaluate temporal changes in signal void volumes and 19F signal (day 1 vs 12). One-way ANOVA was used to determine statistical changes in MPI signal over time (days 1, 5, and 12). These analyses were conducted using Prism software (8.0.2, GraphPad Inc.), where P < .05 was considered statistically significant. Values are presented as mean ± standard deviation.
Evaluation of Ferumoxytol as an MPI Tracer
We used the relaxometer mode on the MomentumTM scanner to measure the FWHM of 30 μg and 5.5 μg (in 1 μL PBS) ferumoxytol. As seen in Figure 1A, we measured a FWHM of 66.335 mT. For a 6.1 T/m gradient, the resolution of this ferumoxytol is 1.088 cm. The amplitude of the 30 μg PSF was ∼4.5 times the height of the 5.5-μg PSF.
Relationship Between Iron Content and MPI Signal
We acquired images of ferumoxytol samples with known iron content (Figure 1B). These samples were separated by 2 cm on the MPI bed (5 samples/scan). There was a strong linear relationship (R2 = 0.992, P < .001) between iron content and MPI signal (arbitrary units, A.U.) (Figure 1C). This relationship holds for small quantities of iron that are relevant for our investigation (Figure 1D). The equation of the line is: MPI Signal = 12.145 × (Iron content) + 2.9034. Using this relationship, iron content may be determined for a given MPI signal. We used these calibration lines to quantify iron content in ferumoxytol-labeled MSC pellets and ferumoxytol-labeled MSCs in vivo.
Assessment of MSC Labeling with Ferumoxytol
Uptake of ferumoxytol by MSCs was assessed using PPB staining (Figure 2, A and B), and iron content in 1 × 106, 0.5 × 106, and 0.25 × 106 cells was quantified using MPI (Figure 2C). These 2 techniques indicated that MSCs were adequately labeled, with 2.430 ± 0.211 pg iron/cell.
We also showed that there is a relationship between MPI signal and the number of ferumoxytol-labeled MSCS in the pellet on day 0. For pellets containing 0.25 × 106, 0.5 × 106, and 1 × 106 cells, iron content was determined to be 0.876 μg, 1.227 μg, and 2.208 μg, respectively (Figure 2D). The viability of these cells did not change with MSC labeling (97% viability before and after labeling).
Detection of Ferumoxytol-Labeled MSCS with 1H MRI and MPI and PFC-Labeled Immune Cells with 19F MRI
Figure 3 shows 1H, 19F, and MPI images of mice which received 1 × 106 ferumoxytol-labeled MSCs (A-C) or unlabeled-MSCs (D,E). In 1H MRI, ferumoxytol-positive MSCs were detected as regions of signal void at the site of implantation in all mice (n = 10), 1 and 12 days following MSC implantation (Figure 3A, labeled Fe). These voids were easily distinguishable on day 1 but were more challenging to locate by day 12 owing to the size reduction of the voids. In some cases, these voids were difficult to distinguish based on other anatomical features that appear dark on 1H MRI.
19F MRI was overlaid to 1H images and it revealed PFC uptake in the bone marrow (BM) and lymph nodes (LN), as well as in the muscle surrounding the MSC implant (Figure 3B). This is due to the presence of phagocytic immune cells in these regions. 19F images were first displayed according to their full dynamic range, then W/L to eliminate the background noise in the final images. This signal is unambiguous and directly related to the number of 19F atoms present in the tissue, which allows for assessment of relative cell number.
Ferumoxytol-labeled MSCs were also detected with MPI 1, 5, and 12 days after MSC implantation (Figure 3C; and see online Supplementary Figure 1). These MPI images were windowed to display the full dynamic range from a 1.5-μg ferumoxytol reference (not shown). In day 12 images, MPI signal generated from ferumoxytol-positive MSCs was diminished but more clearly visible with windowing. In MPI images, signal was also detected in the gut regions, presumably owing to the presence of iron in mouse feed. Mouse feed was imaged separately by MPI and had substantial iron content (shown in online Supplementary Figure 2).
In control mice that received unlabeled MSCs (n = 5), no regions of signal voids were detected in 1H images (Figure 3D). In these same mice, 19F signal was detected in the LNs, BM, and in the muscle where the MSC implant occurred, similar to the mice which received ferumoxytol-labeled MSCs (Figure 3E).
Quantification of Temporal Changes in Iron Voids, 19F Signal, and MPI Signal
Over 12 days, the measured iron void volumes in 1H images declined in all 10 mice, by 62% on average (P = .0003) (Figure 4A). On day 1, the average void was 9.216 ± 4.136 mm3 and by day 12, 3.523 ± 2.217 mm3.
19F signal, detected from PFC-labeled immune cells, was detected in both limbs in all 15 mice on day 1 and 12 (Figure 4B). On day 1, 19F signal in the limb ipsilateral to the MSC implant was (1.866 ± 0.5825) × 1019 spins and (2.522 ± 2.101) × 1018 spins in the contralateral limb. Signal in the contralateral limb was only present in the bone marrow and LNs. The difference in 19F signal between these limbs, representing signal solely from immune infiltration as a result of the MSC implantation, was (1.614 ± 0.6604) × 1019 19F spins. On day 12, (1.560 ± 0.6535) × 1019 19F spins were present in the ipsilateral limb and (8.518 ± 7.227) × 1018 spins in the contralateral limb, resulting in a difference of (1.470 ± 0.6565) × 1019 19F spins. There was no significant difference (P = .148) in the 19F signal resulting from MSC implantation (ie, the differences between limbs) when comparing signal on day 1 and day 12. This indicates the persistent infiltration of immune cells. There was no significant difference (P = .841) in 19F signal between mice administered ferumoxytol-labeled MSCs (1.530 ± 0.746 × 1019, n = 10) and unlabeled MSCs (1.577 ± 0.487 × 1019, n = 5).
In vivo iron content in MSCs measured by MPI on day 1 (1.510 ± 0.206 μg, n = 5 mice) was significantly different from day 12 (0.492 ± 0.131 μg, n = 5 mice; P = .002) (Figure 4C). We imaged 3 of these mice on day 5 and the measured iron content (1.072 ± 0.154 μg) was also significantly higher than iron content measured on day 12 (P = .014). Under the assumption that these cells retain iron in vivo (2.430 ± 0.211 pg/cell), we detected 6.21 ± 0.847 × 105 MSCs 1 d after implantation (n = 5 mice), 4.410 ± 0.634 × 105 MSCs on day 5 (n = 3 mice), and 2.027 ± 0.539 × 105 MSCs on Day 12 (n = 5 mice) (Figure 4D).
These quantitative findings are summarized in Table 1.
Microscopy and Immunohistochemistry
MSCs were identified in hematoxylin and eosin sections among connective and muscular tissue (Figure 5A). PPB staining verified the presence of ferumoxytol in these MSCs (Figure 5B). F4/80 staining with DAB identified macrophages infiltrating the MSCs (Figure 5C). These are directly adjacent sections.
We have detected and quantified the presence of MSCs and infiltrating macrophages over time using a unique combination of cellular imaging technologies. MSCs were labeled with ferumoxytol and detected in 1H images as regions of negative signal (signal void) and in MPI images as regions of positive signal. Macrophages that accumulate at the site of MSC implantation were labeled in vivo with a PFC agent, then detected with 19F MRI. The direct quantification of MPI and 19F MRI signal was used to estimate the relative number of MSCs and macrophages over time. This multimodality imaging approach allowed for the confirmation of MSC delivery, the measurement of MSC number over time (post implantation), and quantification of inflammation.
Ferumoxytol as a Dual 1H MRI and MPI Cell Tracking Agent
In this study, we measured a decrease in the volume of signal loss generated by ferumoxytol-labeled MSCs in 1H images and a decrease in MPI signal detected over 12 days following MSC implantation. This occurred in all mice and is consistent with several previous MRI cell tracking studies from our laboratory (8, 15, 26). Microscopy obtained on day 12 confirmed that PPB-positive cells were present in muscle tissue at the site of implantation. The decrease in the region of signal loss in MR images and the MPI signal is likely because of MSC apoptosis and clearance of these cells by the immune system. The use of MRI to measure signal void volume gave us an indication that there were fewer cells at day 12 than at day 1. However, this is not a direct measure of the number of MSCs present owing to the blooming artifact and the nonlinear relationship between signal loss and cell number. With MPI we detected a decrease in positive signal produced by ferumoxytol-labeled MSCs over 12 days. This finding was in agreement with our MRI measurements; however, the MPI signal is directly related to iron concentration, which can be related back to MSC number. We can estimate the number of MSCs in vivo from the MPI data by comparing the measurements of MPI signal with the mean iron uptake per cell for MSC labeled in vitro. MPI data for cell samples showed 2.4 pg iron per MSC on average. This was used to estimate the number of MSCs in vivo over time (Figure 4, C and D).
The use of ferumoxytol as a tracer for both iron-based 1H MRI and MPI cell tracking is appealing, as it is a clinically translatable iron nanoparticle. However, other iron nanoparticles (ie, ferucarbotran, an SPIO) have superior MPI SNR and spatial resolution compared with ferumoxytol (13). In this study we reported an FWHM of 1.088 cm for 30 μg of ferumoxytol (30 μg iron/μL). We also reported a PSF for 5.5 μg ferumoxytol (in 1 μL PBS) to easily compare FWHM and SNR of ferucarbotran (Vivotrax, 5.5 μg iron/μL) in the future. The ideal MPI nanoparticle is still under investigation, considering the effects of nanoparticle size and biological properties, that is, surface composition and cell labeling process. The Langevin model predicts a cubic improvement in spatial resolution with increasing nanoparticle size (23, 25). USPIO (ferumoxytol) nanoparticles have a diameter of <50 nm, which is smaller and contain less iron than other available nanoparticles such as SPIOs (50–100 nm) and micron-sized superparamagnetic iron oxides (∼1 μm) (27).
In our study, we used protamine sulfate and heparin to increase uptake of ferumoxytol by MSCs. There is evidence (28) that the use of transfection agents (protamine) can reduce MPI detection; however, this effect is seen mainly at lower drive frequencies (0.4 kHz). The MomentumTM MPI scanner uses a 45 kHz alternating magnetic field to excite iron nanoparticles.
We showed that the formation of a calibration line was a robust technique to quantify iron content from measured MPI signal. In this process, the samples of iron (0.75–30 μg) were imaged using the same settings as the other in vivo scans (3D isotropic mode). This linear relationship persists at low iron concentrations (0.75–3 μg), which is useful in the quantification of ferumoxytol-labeled MSCs (which contained ∼2.4 μg in 1 × 106 cells on day 0).
Imaging Inflammation With 19F MRI
This is the first study to use PFC to indicate inflammation associated with iron-labeled stem cells and to track this over time. Following the intravenous administration of PFC, we detected prominent regions of 19F signal in the limb muscle surrounding the MSC implant using 19F MRI. This in vivo labeling technique is known to label resident phagocytic immune cells of the reticuloendothelial system (8, 10). We detected a large number of 19F atoms (on the order of 1019) in the ipsilateral limb on both day 1 and 12. Microscopy obtained on day 12 confirmed that F4/80-positive cells were present in muscle tissue at the site of implantation. This suggests that the number of PFC-positive macrophages remained constant over this time. We can get a rough estimate of macrophage cell number by comparing the value for total 19F atoms with the mean 19F uptake per cell for macrophages labeled in vitro. Our previous work has measured 2.12 × 1011 19F spins per macrophage using NMR (21). Using this value, we would estimate that ∼7.44 × 107 cells are present at the site of the MSC implantation on day 1 and 6.93 × 107 cells on day 12.
This is the first study to demonstrate the ability to image macrophage infiltration in vivo using 19F on a clinical (3 T) MRI system. Compared with iron-based cell tracking, 19F MRI has lower sensitivity and consequently, preclinical 19F cell tracking has only been performed at relatively high magnetic field strengths (>3 T). The bSSFP imaging sequence and surface RF coil play a major role in enhancing sensitivity to enable detection and cell tracking of 19F-positive cells at 3 T.
The MRI and MPI cell labeling agents used in this study (iron and PFC) may be diluted over time as cells proliferate, and thus, there is some ambiguity in the number of cells detected. Since MSCs are implanted in vivo to a suppressive environment that lacks nutrients, we do not expect that MSCs are proliferating substantially. We presume that the reduction ferumoxytol in MSCs, by detection of voids in 1H MRI and by MPI, is predominately resulting from MSC death. The presence and detection of ferumoxytol do not reflect cell viability because this agent is retained within apoptotic MSCs. Phagocytic immune cells uptake these apoptotic MSCs and the USPIO nanoparticles for clearance in the liver. Thus, immune cells may be labeled with PFC and ferumoxytol, in a process called bystander labeling. Hitchens et al. (29) previously showed that if iron and 19F agents are in the same cell, iron-mediated quenching of 19F signal can occur. This may contribute to ambiguity when detecting and quantifying PFC-labeled macrophages that are involved with clearance of MSCs. However, we did not detect a difference in 19F signal between mice, which received ferumoxytol-labeled MSCs or unlabeled MSCs. This indicates that while quenching of 19F signal may be occurring in the presence of ferumoxytol, this effect does not significantly alter the quantification of PFC-labeled macrophages in this application.
We detected unwanted signal in the mouse gut in all 5 mice imaged with MPI (Figure 3) owing to the presence of iron in mouse feed. This gut signal is also present in mice that do not have iron-labeled cells implanted (data not shown). This signal can complicate analysis of MPI using the automatic thresholding tool if the gut signal is much brighter than the region of interest. Because of this, the signal from ferumoxytol-labeled MSCs was manually delineated for 3 of the mice on day 12. This has negligible impact on the signal quantification, rather it is more time-consuming for the user. This gut signal can also create problems if it is in close proximity to other target sources of iron. This did not impact the quantification of iron in this study, as the gut signal was distant enough from the MSC implant; however, this should be considered when designing future experiments.
The in vivo scan times for MPI (1 h) are considerably longer than MRI (9 minutes for 1H scans and 18 minutes for 19F scans). This much time under anesthesia is undesirable for cell tracking when images are collected at multiple time points. Although 2-dimensional MPI scans of mice can be acquired within 3 minutes, these images (which appear as maximum intensity projections) do not have volumetric data for accurate quantification of iron present within a 3D geometry.
In this study, we have shown that iron-based 1H MRI, 19F MRI, and MPI can be used together to noninvasively monitor the fate of 2 cell populations in vivo (MSCs and macrophages). This is the first time that these 3 modalities are combined to monitor cell populations in vivo. We propose that these cellular imaging techniques could be used to monitor MSC engraftment over time and detect the infiltration of macrophages at transplant sites. This could enhance therapeutic monitoring to confirm appropriate MSC delivery, measure the number of MSCs present over time, and quantify immune infiltrate to identify MSC rejection.