|Home | About | Journals | Submit | Contact Us | Français|
Monitoring of cell therapeutics in vivo is of major importance to estimate its efficacy. Here, we present a novel intracellular label for 19F magnetic resonance imaging (MRI)-based cell tracking, which allows for noninvasive, longitudinal cell tracking without the use of radioisotopes. A key advantage of 19F MRI is that it allows for absolute quantification of cell numbers directly from the MRI data. The 19F label was tested in primary human monocyte-derived dendritic cells. These cells took up label effectively, resulting in a labeling of 1.7 ± 0.1 × 1013 19F atoms per cell, with a viability of 80 ± 6%, without the need for electroporation or transfection agents. This results in a minimum detection sensitivity of about 2,000 cells/voxel at 7 T, comparable with gadolinium-labeled cells. Comparison of the detection sensitivity of cells labeled with 19F, iron oxide and gadolinium over typical tissue background showed that unambiguous detection of the 19F-labeled cells was simpler than with the contrast agents. The effect of the 19F agent on cell function was minimal in the context of cell-based vaccines. From these data, we calculate that detection of 30,000 cells in vivo at 3 T with a reasonable signal to noise ratio for 19F images would require less than 30 min with a conventional fast spin echo sequence, given a coil similar to the one used in this study. This is well within acceptable limits for clinical studies, and thus, we conclude that 19F MRI for quantitative cell tracking in a clinical setting has great potential.
Cellular therapy is becoming a key treatment modality for conditions ranging from genetic disorders to cardiovascular disease and cancer.1–4 Transplanted cells include various types of stem and immune cells for cell-based vaccines. The success of these therapies hinges on the fate of the transplanted cells and on their accurate delivery to target tissues. Hence, it is necessary to monitor the transplanted cells in a noninvasive manner in vivo. MRI is often the modality of choice for applications requiring long-term monitoring of transplanted cells because of its noninvasive nature and its nonreliance on short-lived radionuclides.
Dendritic cells (DCs) are the professional antigen-presenting cells of the immune system. Their decisive role in inducing immunity forms the rationale for DC immunotherapy: DCs loaded with relevant antigens are injected into patients to stimulate T cells. DCs have been used extensively in clinical trials5 aimed at modulating the immune response in cancer, infection, allergy and transplant rejection.6 DCs may be loaded with relevant antigens before transfer.7 Alternatively, they can also be loaded with mRNA encoding antigen to aid in antigen presentation in vivo.8,9
The functionality of DCs is heavily dependent on their location and migratory ability.10,11 Thus, it is critical to monitor their trafficking in vivo. This was first demonstrated in a clinical trial using scintigraphy and MRI to track DCs in melanoma patients, where it was found that the site of transfer of the DCs influenced their trafficking to lymph nodes.12 This key study was the first example of clinical cell tracking using MRI. However, the difficulty to quantify cell numbers directly from the MRI data using cells labeled with superparamagnetic iron-oxide (SPIO) required the use of scintigraphy. We suggest that the use of 19F MRI for cell tracking would allow both positive localization of the cells and their quantification directly from the image data in a clinical setting.13 Thus, we tested the labeling of human DCs (as used in cancer vaccine trials) with a clinically applicable, commercial 19F agent. This technique has not yet been applied to clinical cell tracking, and this article explores the feasibility of clinical 19F MRI in this setting.
The labels used for MRI-based cell tracking fall into two broad categories, namely contrast agents and tracer agents containing nuclear magnetic resonance active nuclei other than 1H. Contrast agents result in localized changes of intensity in the 1H image. Specifically, iron-based agents such as SPIO often result in a drop in apparent spin–spin relaxation time around the agents and thus a region of hypointensity in the 1H image. DCs have been tracked using SPIOs in melanoma patients.12 Other metal-based agents, such as Gadolinium (Gd), cause a decrease in the spin-lattice relaxation time T1 and thus a localized region of hyperintensity in T1-weighted images.14 For all contrast agents, the large 1H background hampers accurate quantification.15 The use of a second nucleus for imaging is a relatively new paradigm in MR-based cell tracking.16–18 The relevant nucleus is imaged and a 1H image is obtained separately solely for anatomic information. 19F has been the main focus due to its sensitivity, which is comparable to that of the 1H nucleus. Moreover, the negligible endogenous concentration of 19F in vivo leads to a complete absence of background in 19F images, simplifying positive identification of the label and, therefore, also relatively simple quantification of label within a voxel.18 Indeed, 19F imaging has been developed as an in vivo cytometry assay.19
Here, we demonstrate the utility of a novel, potentially clinically applicable 19F agent, CS-1000 (Celsense Inc., USA) for cell tracking and quantification directly from image data, in comparison to conventional metal-based contrast agents, applied to primary human DCs as used in current cancer vaccine trials.
DCs were generated from adherent peripheral blood mono-nuclear cells, from donor blood, by culturing in the presence of interleukin-4 (500 U/ml) and granulocyte-monocyte colony stimulating factor (800 U/ml) (both Cellgenix, Freiburg, Germany). Cells were cultured in X-VIVO 15 medium (Bio-Whittaker, Walkersville, MD) with 2% human serum (Blood-bank; Rivierenland, Nijmegen, The Netherlands)20 at a concentration of 5 × 105 cells per ml of medium. Label consisting of either CS-1000 (Celsense, USA) at the indicated concentrations, 200 μg ferumoxide/ml of Endorem (Gueberet, France) or 1 mM ProHance (Bracco, Italy) was added on Day 3. CS-1000 is an aqueous colloidal nanoemulsion of a perfluorocarbon polymer. The total fluorine content is 100 mg/ml, and the droplet size of the nanoemulsion is 180 nm. It is specifically formulated to facilitate internalization of the reagent into any cell type ex vivo, regardless of inherent phagocytic ability. The perfluorocarbon molecule used is stable at low pH, lipophobic and hydrophobic, and is not degraded by any cell enzymes.21 At Day 6, DCs were matured by the addition of 10 μg/ml PGE2 (Pharmacia & Upjohn, Puurs, Belgium), 10 ng/ml TNF-α, 5 ng/ml IL-1β and 15 ng/ml IL-6 (Cellgenix). On Day 8, the cells were harvested and washed using cold phosphate buffered saline solution to remove excess label. Cell viability was determined by trypan blue exclusion.
Electroporation was done as described elsewhere.8 Briefly, 20 μg of RNA was transferred to a 4 mm cuvette (BioRad, Veenendaal, The Netherlands), and 10 million cells are added over it in 200 μl OptiMEM without phenol red (Invitrogen, Breda, The Netherlands). After a 3 min incubation, the sample was pulsed in a Genepulser Xcell (BioRad) at 300 V and 150F with an exponential decay pulse. After electroporation, cells were transferred to prewarmed X-VIVO 15 without phenol red (Cambrex Bio Science, Verviers, Belgium) with 5% human serum and incubated for 3 hr at 37°C.
1H and 19F MRI for in vitro samples were performed on a 7 T horizontal bore MR system (Surrey Medical Imaging Systems, United Kingdom) with a 10-mm-diameter 1H/19F double-tuned single surface coil. 1H images were acquired by using T1-weighted spin echo and -weighted gradient echo pulse sequences with a repetition time (TR)/echo time (TE) = 1000/4 msec and TR/TE = 2000/15 msec, respectively, resolution 0.125 × 0.125 × 2, matrix size 128 × 128, 4 averages. 19F spin density-weighted images were acquired using a spin echo sequence with TR/TE = 300/4 msec, resolution: 1 × 1 × 2 mm, matrix size: 16 × 16 mm, 1024 averages. 19F T1 and T2 values for the fluorinated label were measured to be 350 ± 20 msec and 44 ± 8 msec, respectively, at 7 T, when internalized into the cells. 1H T1 and relaxation times for different Gd and SPIO-labeled cell concentrations were measured using the inversion recovery pulse sequence and determining the minimum full width at half maximum of the spectrum at the best shimming conditions, respectively. MR Spectroscopy was used to determine 19F content per cell with a calibrated reference of triflouroacetic acid (TFA) by applying an adiabatic 90° pulse to excite the whole sample.
For MRI, a variable number of labeled DCs were embedded in gelatin. Alternatively, 1.5 million labeled cells were injected into bovine muscle tissue and imaged. Relative T1 and were referred to the values of a reference sample consisting of gelatin with nonlabeled cells, and to the TFA reference for 19F images.
In vivo MR images were acquired at 11.7 T using a 89 mm vertical-bore Bruker microimaging system (Bruker Biospin, Billerica, MA) equipped with a 35-mm-diameter volume coil that can tune between 470 and 500 MHz for 19F and 1H, respectively. 19F-labeled human DCs (~3 × 106) were injected subcutaneously into a quadriceps of a female NOD-SCID mouse, 6 weeks of age. The mouse was anesthetized (1.5% isoflurane in 80% O2 and 20% N2O), intubated and placed on a mechanical ventilator with respiratory gating. Body temperature was maintained at 37°C. 19F images were acquired by using a rapid acquisition with relaxation enhancement (RARE) sequence with TR/TE = 500/11.5 msec, RARE factor 8, 64 × 64 image points, 45 × 30 mm in-plane field of view, 2 mm slice thickness, and 256 averages; and 1H images using a spin-echo sequence with TR/TE = 100/15, 4 averages, 256 × 256 image points, and the same field of view and slice thickness as the 19F. Experiments were carried out in accordance with the guidelines provided by the Carnegie Mellon Institutional Animal Care and Use Committee (IACUC) and the National Institute of Health Guide for the Care and Use of Laboratory Animals.
The “relative intensity” for the 1H data (Fig. 3b) was calculated using the relation , where Icells is the total signal intensity over the labeled cells. Icontrol is the total signal intensity over a region of similar size containing non-labeled cells. In all cases, intensity was summed for all voxels in-plane. Note that Icells − Icontrol is the change in signal intensity due to the labeled cells.
To calculate the signal-to-noise ratio (SNR) of 19F MR images, the noise magnitude N was first determined by averaging the intensity of the voxels near the border of the image. Next, voxels with 19F signal intensity higher than 3N were selected, and the signal intensity of each voxel was corrected to account for the Rician-distributed noise present in intensity images with low SNR.22 The threshold for detection was fixed at SNR = 3.
To perform the cell density map calculation (Fig. 4b), the signal intensity values for each voxel and for every slice with SNR > 3 were determined as explained above. The total signal intensity (sum of all signal intensities from all voxels over all slices) was then calculated and associated with the total amount of cells injected into the tissue. Finally, the number of cells in each voxel was calculated by normalizing the signal obtained against the total signal intensity. The density of cells for each voxel was calculated by considering the voxel volume. All errors shown are the standard deviation.
Fluorescence activated cell sorter (FACS) analysis was performed using with a FACSCalibur™ flow cytometer equipped with CellQuest software (BD Biosciences, Mountain View, CA). The primary antibodies used were anti-CD80 (BD Biosciences), anti-CD83 (both Beckman Coulter, Mijdrecht, The Netherlands), anti-CD86 and anti-CCR7 (both BD PharMingen, San Diego, CA). Secondary staining was with isotype appropriate goat-anti-mouse Alexa648 (Invitrogen). Intracellular staining for mRNA expression was carried out on cells permeabilized in PBS with 2% bovine serum albumin (BSA), 0.02% azide and 0.5% saponin (Sigma-Aldrich) (PBA/saponin), and stained with mAb diluted in PBA/saponin with 2% HS, followed by staining with allophycocyanin-labeled goat-anti-mouse (BD PharMingen).
We studied the labeling of human DCs with the 19F CS-1000 agent. Cells were incubated with CS-1000 under various conditions. DCs take up CS-1000 effectively, without a need for transfection aids (Fig. 1a). Label was added to immature DCs, and the cells were incubated for 3 days before the addition of maturation factors. Significant toxicity was observed only at high concentrations of label (Fig. 1a). Uptake appears to saturate at 140 μl of label per million of cells. Based on these data, we used 75 μl of CS-1000 per million cells on Day 3 after starting the monocyte culture, i.e., with immature DCs, for all further experiments. Note that these cells do not undergo cell division. This yielded 1.7 ± 0.1 × 1013 19F atoms per cell, with an acceptable viability of 80 ± 6%. The low standard deviation of 19F content/cell obtained after measuring three independent samples also suggests homogeneous cell labeling.
The DCs were analyzed after labeling, to study any possible effect of CS-1000 on the cells. General visual inspection under the microscope showed no difference with nonlabeled controls, except for the bigger size and the presence of label droplets in the labeled cells. Consistent with this observation, a difference in side scatter between labeled and nonlabeled DCs was clearly seen (Fig. 1b) because of the presence of intracellular label droplets. The expression of the DC maturation markers CD80, CD86, CD83 and CCR7 was unchanged in labeled cells relative to untreated controls (Fig. 1c, left). Furthermore, the label had no effect on protein expression after mRNA electroporation (Fig. 1c, right), which is a typical step in clinical DC vaccination trials.8 A mixed leukocyte reaction using labeled and nonlabeled DCs to stimulate allogeneic lymphocytes showed no difference in T cell activation between the two groups at all DC:lymphocyte ratios tested (p = 0.05; data not shown). From all these data, we conclude that CS-1000 has no negative effects on DC viability, antigen expression and maturation.
Standard imaging sequences were used in our studies, as outlined in Materials and Methods. Cells were resuspended at known densities in gelatin for determination of the minimum number of cells per voxel necessary for detection (Fig. 2a). We found the minimum detectable density of 19F-labeled cells to be comparable with that of Gd-labeled cells when assessed visually, at under 2000 cells/voxel when imaged with a 19F spin density weighted and 1H T1-weighted sequence, respectively. The detection limit using SPIO was lower, ~125 cells/voxel, using -weighted imaging in our system (Fig. 2b). However, these data were acquired with a homogenous 1H background and are not directly comparable to in vivo sensitivity.
The number of labeled DCs in a region of interest was evaluated from the MR image data. Cells were labeled with Gd, SPIO or 19F and suspended at various, known densities in gelatin for imaging, as before. Figure 3a shows the linear relationship between 19F intensity and the number of cells per voxel. A similar linear relationship exists between the SNR and cell density. Thus, the cell number can be calculated directly from image data, if the label content per cell is known.
The plots for Gd and SPIO-labeled cells (Fig. 3b, left and right, respectively) show the more complex relationship between T1, and relative contrast in the image. The change in contrast begins to saturate at about 8,000 cells/voxel with Gd and as early as with 800 cells/voxel using SPIO. These issues complicate cell number quantification using contrast agents but not 19F labels.
DCs (1.5 million) were labeled either with SPIO, Gd or 19F and injected as a bolus in bovine muscle tissue and imaged to study detection sensitivity over tissue background. The corresponding images are shown in Figure 4a. Unlabeled cells were also injected similarly as a negative control. This control shows the typical nonhomogeneous background from biological tissue, unlike with gelatin phantoms (Fig. 2). The contrast due to SPIO is apparent in the bolus injection, but the Gd-labeled cells are not readily detected in this image over the background. The 19F-labeled cells are clearly visible and readily differentiated against the 1H tissue underlay (Fig. 4a). The plots under the corresponding images show the signal intensity through a horizontal line across the center of the sample (plotted in red for the 19F image), where all are set to the same scale. As expected, the SPIO-labeled cells display a sharp drop in signal intensity, and the Gd-labeled cells show a slight peak in its intensity profile. The 19F-labeled cells are readily identified, even though the SNR of the 19F data is lower than that of the 1H images.
The cell density in the slice containing the 19F-labeled cells was calculated from the image data and is shown in Figure 4b. As expected for a bolus injection, the highest cell density was at the center of the injection site. These findings demonstrate the relative ease of quantification using 19F agents. As it can be inferred from Figure 4b, the sensitivity obtained was again in the order of 2,000 cells/voxel, affirming that sensitivity and quality of 19F images are not affected by the presence of tissue background.
To demonstrate the feasibility of in vivo 19F tracking, 3 × 106 mature, labeled human DCs were injected and imaged in mouse. Figure 5 displays the fused 1H and 19F image 12 hr postinjection. Labeled cells are clearly visible in the 19F image, localized in a region anatomically consistent with the proximal draining inguinal lymph node. As the cells were subcutaneously injected in the quadriceps, this image shows cell migration away from the injection site.
This image also demonstrates that 3 × 106 cells can easily be visualized in vivo at 11.7T with a measurement time of 28 min. The absence of any 19F background signal indicates that the presence of isoflurane is under detection limits and it can be neglected in this experiment. However, an alternative non-fluorinated anesthetic should be used when lower cell densities need to be measured or with longer scan times.
We show that a commercially available perfluorocarbon emulsion is a promising label for 19F MRI-based cell tracking in a clinically relevant cell type. We found that the CS-1000 label had minimal effect on DC viability and function, as tested by expression of maturation markers, mRNA expression after electroporation and T cell activation. The minimum number of labeled cells per voxel detectable at 7T with our imaging parameters is ~2000, with 1.7 × 1013 fluorine atoms per cell. This labeling is consistent with previously published data using similar 19F labels in murine cells.18,23 Uptake of label by mature DCs was much lower (data not shown) as maturation downregulates endocytic processes; hence, we labeled the cells in the immature stage. Labeling was observed within 24 hr after addition of label, although we cultured the cells for 6 days after label addition to minimize cell handling steps.
The detection limit using SPIO in vitro is almost 10-fold lower than either the 19F or the Gd labels (Figs. 2 and and3).3). This is comparable with previous reports on the sensitivity of MRI for the detection of human DCs labeled with SPIO at 1,000 cells/mm3 in vitro when loaded with 30 pg Fe/cell.24 SPIO agents can be extremely sensitive, and detection limits as low as single cells have been reported using SPIO agents in vitro.25–27 However, the addition of 1H background from biological tissue complicates identification of labeled cells (Fig. 4a), as endogenous regions of hypointensity are observed in various tissues, blood vessels or blood clots. In addition, the T1 and effects are not independent of each other. Most importantly, the effect of contrast agents on relaxivity saturates at higher concentrations (Fig. 3b). agents are particularly sensitive to this saturation effect (Fig. 3).
Cell number quantification using agents is prone to error due to the uncertainty in agent relaxivity in situ and at higher cell densities due to the saturation of the effect, as demonstrated in Figure 3b. Cheung et al. carried out quantification of cells in phantoms using a ultrasmall SPIO agent and found that the cell numbers were underestimated at cell densities greater than approximately one million cells per voxel.28 This may be a limitation of the technique given that the typical cell numbers injected in a clinical DC vaccination trial is about 15 million12 and cell numbers can be even higher with other cell therapies. Other techniques to decipher the complex change in observed contrast with SPIO labels involving quantitative measures of relaxation rates are also susceptible to the same saturation effect.15 Recent advances with contrast agents include the use of “white markers” for cell localization using agents.15,28–31 In this technique, the background is made dark by dephasing gradients, and the intensity of the signal near the contrast agent dipoles is dependent on relaxivity, echo-time, slice thickness and gradient strength.30 Endogenous iron stores and blood clots at the site of injection also complicate relaxometric measurements at the site of cell transfer.32
It is also worth to mention that as we used monocytes-derived DCs, these cells do not divide.32 This fact makes the method used for cell quantification particularly suitable for tracking DCs since no signal reduction nor data misinterpretation are caused by the uncontrolled process of cell division. However, the technique of quantitative cell tracking using 19F MRI has also been applied to actively dividing T cells, which were successfully tracked for up to 3 weeks.32 The underestimation of cell numbers that occurs due to cell division is often within tolerable limits, two–fourfold depending on the division rate and length of time. This error can be reduced if the cell division rate is known. Exo-cytosis of the label after cell death could also be a source of error in the quantification method, but only relevant when the number of dead cells is in the same order of magnitude as living cells. This issue is discussed in more detail elsewhere.13
Together, our findings demonstrate the feasibility of 19F labeling of human DCs and cell quantification using MRI. The label shows comparable detection sensitivity to cells labeled in vitro using Gd-complexes when imaged using conventional contrast mechanisms. The main limitation of 19F imaging is the low concentration that necessitates the use of signal averaging and potentially longer scan times. Our current detection limit of 2,000 cells/voxel at 7T with 1.7 ± 0.1 × 1013 19F atoms/cell translates to a limit of about 9,000 cells/voxel at 3T (considering a strong SNR∞Bo7/4 dependence with the magnetic field intensity Bo for small coils), with other factors remaining constant. To put this into perspective, a typical DC vaccination study utilizes an intranodal or intradermal injection of about 10 million cells, with between 30,000 and 200,000 cells migrating to secondary lymph nodes, as detected using scintigraphy on 111In-labeled DCs.33 We note that the size of a voxel in MRI is determined by the practitioner, and high resolution imaging is not necessary for 19F-based cell localization as these images are overlaid onto high resolution 1H images for anatomy. Importantly, the SNR needed for 19F images can generally be much lower (e.g., SNR~5 or less) than normally acceptable for 1H images, as the 19F is not used to provide tissue contrast or organ definition, but only cell population localization. Moreover, further improvements in 19F MRI methodologies might yield significant imaging acceleration, for example using restricted k-space acquisitions schemes.34
Currently, clinical 19F MRS requires a concentration of 19F in the millimolar range.35,36 Our current detection limits of 2,000 or 9,000 cells/voxel at 7 T and 3 T, respectively, translate to concentrations of 19F at 1 and 5 mM. The clinical MR spectroscopy work suggests that these concentrations can be achieved and detected using human scanners.
Thus, to image a lymph node with 30,000–200,000 cells based on our system (acquired using the same coil) would take 30–0.7 min at 3T, while maintaining a minimum SNR of 3 for 19F (sufficient for quantification18,19) when using a single spin echo sequence. The use of a fast multiecho imaging sequence,18,19 typical in clinical applications, with a turbo factor (RARE factor) of 4 would result in imaging times of 8–0.2 min for 30,000–200,000 cells, which are within reasonable limits for clinical use. Overall, realistic advances in hardware and imaging protocols may allow in vivo 19F-based cell tracking to become clinically relevant in the near future.
Grant sponsors: Netherlands Organization for Scientific Research (NWO, VISTA grant) and by Netherlands Organisation for Health Research and Development (ZonMw);
Grant numbers: 911-06-021, 917-76-363;
Grant sponsor: National Institutes of Health;
Grant numbers: R01-CA134633, R01-EB003453, P01-HD047675, P41-EB001977;
Grant sponsor: European Commision grant ENCITE;
Grant number: Health-F5-2008-201842
The authors would like to specially thank Andor Veltien for the 19F/1H coil construction and very helpful discussions. They also acknowledge to Mandy van de Rakt, Nicole Meeusen-Scharenborg, Annemiek de Boer, Dr. Gerty Schreibelt, Dr. Annechien Lambeck and Mark van Uden.