Several animal studies have demonstrated improved cardiac function after transplantation of pluripotent stem cells into an injured murine myocardium (1–3). The biological mechanism underlying this functional restoration has not been elucidated. Hypotheses of this therapeutic benefit have ranged from regenerative to paracrine effects (4, 5). Nevertheless, the survival of transplanted cells is critical if those cells are expected to contribute to functional restoration (6). Therefore, a reliable in vivo imaging method is needed to monitor the viability of transplanted cells to evaluate the efficacy of cell therapy.
Magnetic resonance imaging (MRI) may be an ideal noninvasive modality to evaluate the therapeutic effects of cell therapy in the heart (7). It enables arbitrary tomographic views with exquisite tissue contrast at high spatial and temporal resolution. However, MRI suffers from reduced sensitivity in in vivo cellular and molecular applications. Recent implementation of superparamagnetic iron oxide nanoparticles (SPIO) has advanced the sensitivity of cellular and molecular MRI (8, 9). Cells are labeled ex vivo using various transfection agents to facilitate internalization of SPIO into the cytoplasm (10). However, the in vivo MRI signal generated by this labeling method does not provide any biological information of the transplanted cells such as viability, proliferation and teratoma formation (11, 12). Multiple studies have shown transgene and SPIO-conjugated antibody techniques to target specific cell markers in mostly cancer cells (13–16). A novel molecular MRI method has been developed combining the reporter gene (RG) and SPIO-conjugated antibody techniques. Our RG construct has been designed to express antigens on the cell surface of the viable embryonic stem cells (ESCs). In vitro molecular MRI signal has been generated from the viable ESC-RGs by employing SPIO-conjugated monoclonal antibody against these antigens (SPIO-MAb) (13, 14). Furthermore, in vivo MRI allows the assessment of the viability and proliferation of the transplanted ESCs and, subsequent, early teratoma formation.
MATERIALS AND METHODS
MRI Reporter Gene (RG) Construct and Transduction of Embryonic Stem Cells (ESCs) using p2K7 Lentiviral Vector
Firefly luciferase (fluc) was cloned between the N-terminus of HA antigen and the C-terminus of myc antigen of pDispaly (Invitrogen, Carlsbad, CA), generating a RG consisting of the following sequence: Igκ-HA-fluc-myc-PDGFR-TM. The murine Ig κ-chain leader sequence directed the fusion protein to the ribosomal secretory pathway and the platelet derived growth factor receptor transmembrane domain (PDGFR-TM) anchored it to the plasma membrane to express the antigens on the surface of ESCs. Following the cloning of this RG into pENTR™5′-TOPO vector (Invitrogen, Carlsbad, CA), the promoter and the RG were directionally cloned into the p2K7 lentiviral vector using the Gateway LR plus clonase enzyme (Invitrogen, Carlsbad, CA) (17). Two different promoters (EF-1α or human ubiquitin C) were compared for their expression of the RG. The ESCs were incubated with viral supernatant for 24 h followed by blasticidin selection for 3 days, generating RG transduced ESCs (ESC-RGs).
Culture of Undifferentiated ESCs
The hESCs (H9, WiCell and the National Cell Bank, Madison, WI) and mESCs (E14 from Weissman Laboratory, Stanford University) were maintained in undifferentiated and pluripotent state by culturing on irradiated mouse embryonic fibroblast cells in the culture medium (18). Both hESCs and mESCs were cultured at 37.1°C and 5% CO2 in a humidified incubator.
Fluorescence Activated Cell Sorter (FACS) Analysis of RG Expression
To determine the RG expression of myc and HA antigens on the cell surface, ESC-RGs were labeled with either FITC-conjugated anti-myc antibody (FITC-myc-MAb) or PE-conjugated anti-HA antibody (PE-HA-MAb, Miltenyi Biotech, Auburn, CA), followed by analysis for FITC by FL1 channel (excitation at 488 nm by a 15-mW argon laser and detection above 530 nm) and for PE by FL2 channel (excitation at 488 nm by a 15-mW argon laser and detection above 585 nm) with appropriate compensation using FACS Caliber (Becton Dickinson, San Jose, CA). Using FlowJo v7.2.5 (Tree Star, Ashland, OR), histogram was generated and the percentage of myc or HA expressing cells was calculated by subtracting nontransduced cells from the RG transduced cells.
Labeling of Viable ESCs with SPIO-MAb
To assess in vitro MR viability signal of mouse and human ESC-RGs, the cells were labeled with 20 μL of either SPIO-conjugated anti-myc antibody (SPIO-myc-MAb) or SPIO-conjugated anti-HA antibody (SPIO-HA-MAb, Miltenyi Biotech, Auburn, CA). The SPIO-myc- and SPIO-HA-MAb consist of a superparamagnetic iron oxide core with polysaccharide coating, linked covalently to monoclonal antibodies against myc and HA antigens. The mean diameter of the SPIO is approximately 50 nm. To establish the specificity of the RG-mediated assessment of cell viability, 2 in vitro negative control groups consisting of nontransduced ESCs and apoptotic ESC-RGs incubated under the same conditions were established. After labeling the cells with SPIO-HA- and SPIO-myc-MAbs, all cells were washed twice with PBS (1 mL) and centrifuged at 600 RPM for 5 min. Apoptosis was induced by incubating ESC-RGs with 10 μM of doxorubicin (Sigma, St. Louis, MO) for 2 h prior to labeling by SPIO-MAb (19).
In vitro Optical Bioluminescence Imaging (BLI)
D-luciferin was added to the culture media of ESC-RGs at a concentration of 15 mg/L. Nontransduced ESCs were used for negative control. Cells were imaged using IVIS –Spectrum (Caliper, Mountain view, CA) for 30 min with 1-min acquisition intervals. Bioluminescence was quantified in units of average photons per second per centimeter squared per steradian (P·s−1·cm−2·sr−1) using Living Image 2.5 software (Caliper, Mountain view, CA) (20).
In vitro Molecular MRI
There were three 1 × 106 of SPIO-HA- and SPIO-myc-MAb labeled cell groups: (1) mouse and human ESC-RGs, (2) mouse and human ESCs (nontransduced), and (3) mouse apoptotic ESC-RGs. The cells were suspended in 200 μL of PBS and then placed in a 330 μL PCR microfuge tube. These microfuge tubes containing the cells were stabilized within a phantom made of 0.7% agar and 1% copper sulfate. The phantom was placed in the iso-center of knee coil and scanned using Signa 3.0 T Excite HD scanner (GE Healthcare System, Milwaukee, WI). A GRE sequence using the following parameters optimized T2*-weighted imaging to maximize the signal from SPIO (TR 100 ms, TE 20 to 60 ms, FA 45°, matrix 128 × 128, NEX 1, FOV 12, slice thickness 1 mm). The images were analyzed using ImageJ 1.41 software (NIH, Bethesda, MD). Contrast-to-noise ratio (CNR) was calculated as CNR = (SIcell – SIphantom)/SD of the image noise.
In vivo Molecular MRI
Animal care and interventions were provided in accordance with the Laboratory Animal Welfare Act. The animal protocols were approved by the administrative panel on laboratory animal care at Stanford University. The dosage of SPIO-MAb, the incubation time delay prior to MRI acquisition, and echo time (TE) were optimized using a mouse hind limb model. The hind limb model was generated by transplanting 2 × 106 mouse ESC-RGs into the hind limb muscles of severe combined immunodeficiency (SCID) mice (n = 7). After acquiring preadministration GRE images using the following parameters (TE minimum to 15 ms, TR 500 ms, NEX 1, matrix 128 × 128, slice thickness 1 mm) on Signa 3.0 T Excite HD scanner, three different doses (100 μL, 150 μL, and 200 μL of SPIO-MAb) were administered to different mice through the tail vein. Both hind limbs of these mice were scanned repeatedly using the same sequence and parameters. To achieve optimal T2*-weighted imaging, TE was increased from minimum to 15 ms. The most favorable image quality, based on minimum dephasing artifact extending to the adjacent anatomical structure, was obtained at TE of 10 ms.
For creation of myocardial infarction, SCID mice (8–12 weeks old, n = 18, Charles River Laboratories, Wilmington, MA) were anesthetized by 3% isoﬂurane at 1 L/min oxygen and endotracheally intubated with a 20-gauge angiocatheter (Ethicon Endo-Surgery, Cincinnati, OH). Ventilation was maintained by 1% isoflurane at 1 L/min oxygen with a Harvard rodent ventilator (Harvard Apparatus, Holliston, MA). Myocardial infarction was created in all the mice by permanent ligation of the distal left anterior descending artery, resulting in distal anterior or apical infarct. A total of 0.25 × 106 of mouse ESC-RGs were transplanted into the proximal border of the peri-infarct area at the mid-ventricular level to minimize confounding signal artifacts, which could arise from the infarcted apex. All mice (n = 18) underwent left anterior descending ligation and were imaged before and after intravenous administration of SPIO-MAb at post-transplant days 3, 5, 7, 10, and 14. Two mice were euthanized for histology at each time point. For in vivo MRI, a 1-inch-diameter dedicated small animal receiver coil was built and tuned for 3T imaging.
Based on the data from the hind limb model, 80 μL each of SPIO-myc- and SPIO-HA-MAb was administered into each mouse through the tail vein at each time point. SPIO-MAb was allowed to circulate for 12 to 18 h to enhance specific binding with myc and HA antigens and to remove unbound antibody from the system. To determine the specificity of our in vivo imaging technique, two negative control groups were established. In the first negative control group (n = 4), the mice received 160 μL of SPIO-HIS-MAb (Miltenyi Biotech, Auburn, CA), targeted to unrelated antigen, HIS. In the second negative control group (n = 2), the mice received 40 μL of nontargeted SPIO (Feridex I.V., Advanced Magnetics, Cambridge, MA).
For in vivo MRI, the mice were anesthetized with 1% isoflurane at 1 L/min oxygen and placed in the supine position. The ECG gating was optimized by two subcutaneous precordial leads and the respiratory gating by an abdominal respiratory pad. Both cardiac and respiratory gating were monitored and controlled by PC-SAM (SA Instrument, Stoney brook, NY). The body temperature was monitored to maintain a normal physiologic status. The mice were imaged using cardiac and respiratory gated GRE sequence with the following parameters (TE 10 ms, FA 60°, matrix 256 × 256, NEX 10, FOV 3 and slice thickness 1 mm, space 0 mm). Three to four slices were acquired for a short axis view to cover the majority of the left ventricle and single frame image was acquired per imaging plane. In vivo images were analyzed using ImageJ 1.41. CNR was calculated as CNR = (SISPIO-ESC-RG – SIseptum)/SD of the image noise before and after SPIO-MAb administration. The CNR values in Fig. 6 were obtained by subtracting precontrast CNR in the ROI from postcontrast CNR in the ROI.
In vivo BLI
Mice were anesthetized with 1% isoflurane at 1 L/min oxygen and placed in the supine position within the IVIS imaging chamber. Following intraperitoneal injection of D-luciferin (375 mg/kg body weight), mice were imaged for 30 min with 3-min acquisition intervals using IVIS-spectrum. Images were analyzed in the same manner with the in vitro BLI.
Two mice were euthanized after each time point. Hearts were sectioned along the short axis plane and routinely processed for H&E staining. H&E stained slides were interpreted by a pathologist blinded to the study.
Descriptive statistics include mean and standard deviation of mean. Comparison between two groups was performed using student's t-test. For comparison between multiple groups, ANOVA with post hoc test was utilized. All data analyses were done by SPSS 16.0 (SPSS Inc., Chicago, IL). Significance was assumed when P < 0.05.
Reporter Gene (RG) Expression on the Membrane of RG Transduced Embryonic Stem Cells (ESC-RGs)
RG consisting of recombinant fusion protein sequence, 2.2 Kb in length, was constructed as shown in Fig. 1. Flow cytometry demonstrated that 40.5 ± 4.7% of the human ESC-RGs expressed myc and 47.6 ± 5.4% of these cells expressed HA (n = 3), and 52 ± 9% of the mouse ESC-RGs expressed myc and 48 ± 8% of these cells expressed HA (n = 3), compared with the respective nontransduced ESCs (negative control) as shown in Fig. 2A,B. Comparable immunofluorescence signals were achieved by myc and HA antigen expression in both human and mouse ESC-RGs. There was no statistical difference in the RG expression between EF-1α and human ubiquitin C (n = 3). For the subsequent studies, EF-1α was chosen due to the comparatively higher RG expression. The RG expression was validated by positive luciferase (fluc) activity using in vitro BLI as demonstrated in Fig. 3A. Furthermore, quantitative analysis of fluc activity correlated with increasing number of cells (r2 = 0.95, P < 0.05) as shown in Fig. 3B.
In vitro Molecular MRI Detects Mouse and Human ESC-RG Viability
In vitro viability MRI signal was compared among the following three groups (n = 3): (1) viable mouse and human ESC-RGs, (2) nontransduced mouse and human ESCs, and (3) apoptotic mouse ESC-RGs. First, all groups of mouse and human ESC-RGs demonstrated positive molecular MRI signal. Direct comparison of the CNR generated by the MRI signal between the mouse and human ESG-RGs vs. nontransduced mouse and human ESCs showed significant CNR difference (−4.15 ± 0.8 vs. 0.83 ± 0.44 in mouse, P < 0.01 and −3.75 ± 0.46 vs. 0.8 ± 0.35 in human, P < 0.01) as demonstrated in Figs. 4A,B. The in vitro MRI viability signal of both the transduced mouse and human ESC-RGs was validated by positive in vitro BLI signal from the identical groups of cells as demonstrated in Fig. 3A. Second, the MRI viability signal was compared between the viable and apoptotic mouse ESC-RGs. The CNR measured from the viable mouse ESC-RGs was significantly decreased when compared to the apoptotic mouse ESC-RGs and nontransduced mESCs as shown in Fig. 4C. Finally, no significant difference in CNR was measured between the apoptotic mouse ESC-RGs vs. nontransduced mouse ESCs despite some visual difference as shown in Fig. 4C.
In vivo Molecular MRI Detects the Mouse ESC-RG Survival and Teratoma Formation
None of the mice showed any signs of distress after systemic administration of SPIO-MAb. All mice survived throughout the study except one mouse, which died on day 14 due to teratoma formation. On day 3, none of the mice showed in vivo viability signal while BLI showed positive fluc activity in all the mice. On day 5, in vivo MRI viability signal was observed in half of the mice. On day 7, 75% of mice showed in vivo MRI viability signal. On day 10, all mice demonstrated molecular MRI viability signal. The in vivo molecular MRI signal generated from different mice at each time point validated by BLI are demonstrated in Fig. 5A.
To illustrate the specificity of the molecular MRI viability signal, intravenous delivery of unrelated SPIO-HIS-MAb and nonantibody conjugated SPIO was conducted. Longitudinal MRI acquisition of the same mouse demonstrated no signal while concurrent BLI generated positive luciferase signal as shown in Figs. 5B,C.
The CNR measurement of the in vivo MRI viability signal gradually enhanced with proliferation of mESC-RGs as illustrated in Fig. 6A. The precontrast CNR prior to each acquisition was 18.4 ± 14.8% of postcontrast CNR due to residual CNR from the previous imaging session. This increase in in vivo CNR was validated by significant step-up of luciferase activity seen by BLI at each time point as demonstrated in Fig. 6B. Measurement of 3-dimensional volume of the transplanted mouse ESC-RGs based on the MRI viability signal showed that the transplanted mouse ESC-RGs proliferated and expanded as shown in Fig. 6C.
Localization of the mESC-RGs on H&E staining corresponded with the MRI signal as demonstrated in Fig. 7A. The H&E staining confirmed that the mouse ESC-RGs survived, proliferated, and formed teratoma. The teratoma consisted of three different germ layers (endoderm, mesoderm, and ectoderm) on day 14 while only proliferating immature ESCs were observed on day 10 as illustrated in Fig. 7B. Significant increase in CNR, luciferase activity, and cell volume was observed with the early teratoma formation on day 14 as shown in Figs. 6 and 7.
Our RG-mediated molecular MRI technique generated key features not readily achieved in other in vivo molecular imaging methods such as PET, SPECT, or BLI. First, it monitored the biological activity of the transplanted ESCs such as viability, proliferation and the subsequent teratoma formation while also providing precise evaluation of the anatomy at the site of cell transplantation. Second, our RG enabled dual modality imaging of cellular viability by double fusion protein, coupling MRI, and BLI. This capability not only combined high resolution of MRI and sensitivity of BLI but also served as a reliable validation tool of MRI signal. Third, the MRI signal generated from the RG did not dilute out with in vivo cell division or differentiation, allowing longitudinal tracking of the transplanted ESCs in contrast to the traditional prelabeling technique (11). As the transplanted ESCs formed early teratoma, significant increase in CNR, volume, and luciferase activity was noted. Finally, our RG will be able to select specific cell population magnetically by employing SPIO-MAb in conjunction to molecular imaging (21).
Teratoma formation is a significant consideration in clinical translation of ESC therapy. PET is the currently preferred method to detect teratoma. However, sensitivity for this imaging modality has been limited to lesions larger than 1 cm and the uptake of 2-18fluoro-2-deoxy-D-glucose has been confounded by inflammatory changes and granulomatous tissues (22). Our molecular MRI technique addresses some of these shortcomings by sensitive detection of cell proliferation and, subsequent, early teratoma formation with high spatial resolution. The MRI signal generated from early teratoma formation indicated a 2-fold increase in CNR and cell volume. While specific MRI diagnostic criteria for teratoma formation have not been established, the rate of increase in CNR and cell volume measurements may provide future guideline for molecular MRI to distinguish cell proliferation from teratoma formation. In our study using murine myocardial injury model, teratoma was localized within the myocardium and characterized initially by endodermal development followed by mesodermal and ectodermal differentiation. None of the mice showed distant metastasis of the transplanted ESCs.
Antibody-mediated molecular imaging, which targets cell surface receptors, may potentially modify biological function of the ESCs (21, 23). Therefore, nonfunctional cell surface receptors were selected as optimal targets for molecular MRI. Although no known surface marker is expressed throughout the differentiation process, HA is not normally expressed by the mammalian cells and c-myc transcription factor is a nuclear phosphoprotein (24). This property is especially critical in ESC biology since pluripotency and subsequent differentiation capability must be preserved.
Several groups have reported various types of MRI RGs based on the modification of the enzymatic activities of protease, metalloproteinase, and ß-galactosidase, and cell surface receptor expression (25–28). While most of these novel approaches demonstrated in vivo sensitivity in high magnetic field strength, our study generated in vivo molecular MRI signal of viability, proliferation, and teratoma formation of the transplanted mESCs in a clinical grade scanner. Furthermore, an important consideration in any molecular imaging method is the ease of production and dissemination of the technique (29). All components including the lentiviral vector and SPIO-MAb for our imaging method are commercially available, which should lead to consistent duplication and faster dissemination.
One limitation of our in vivo molecular MRI of stem cell viability signal is the known blooming artifacts generated from the SPIOs. This issue is addressed by our previously reported off-resonance positive contrast technique (11, 30, 31). Although endogenous iron products from bleeding and susceptibility artifact at air-tissue interface have confounded the MRI signal from the SPIO, our recently developed sequence based on SSFP and slice select capability reduced this problem (32, 33). Second, residual SPIO from previous scans can be a limitation in longitudinal studies, since it necessitates subtracting precontrast CNR from postcontrast CNR. Third, to maximize the cell transduction efficiency, lentiviral vector was employed to generate ESC-RGs. However, nonviral methods are also investigated in our current effort. Furthermore, ectopic expression of myc and HA may provoke an immune response targeting and clearing the transplanted ESCs in immunocompetent subjects. Thus, RG is effective for immunosuppressed preclinical models to assess ESC survival and teratoma formation. Finally, this study is limited by the fact that neither cardiac function nor myocardial infarct size was measured by MRI, although current efforts are underway to evaluate the effects of teratoma formation on cardiac function and on myocardial injury.