SEARCH

SEARCH BY CITATION

Keywords:

  • izon oxide nanoparticles;
  • cancer;
  • cell imaging;
  • high resolution MRI;
  • high-temperature superconducting coil

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. EX VIVO HISTOLOGY
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

We demonstrate the feasibility of detecting individual tumor-infiltrating cells in vivo, by means of cellular magnetic labeling and a 1.5 Tesla clinical MRI device equipped with a high-resolution surface coil. Using a recently developed high-temperature superconducting (HTS) surface coil, single cells were detected in vitro in voxels of (60 μm)3 at magnetic loads as low as 0.2 pg of iron per cell. The same imaging protocol was used in vivo to monitor infiltration of ovalbumin-expressing tumors by transferred OVA antigen-specific cytotoxic lymphocytes with low iron load. Magn Reson Med 60:1292–1297, 2008. © 2008 Wiley-Liss, Inc.

MRI is a method of choice for noninvasive tracking of injected cells in vivo, thanks to its high spatial resolution and the availability of safe intracellular contrast agents. Various methods were developed to label mammalian cells with iron oxide superparamagnetic nanoparticles, making them visible by MRI without affecting cell functionality (1, 2). Particularly, anionic citrate-coated iron oxide nanoparticles (AMNP) have strong affinity for cell membranes, resulting in nonspecific adsorption followed by internalization into magnetic endosomes (3). These nanoparticles efficiently label a wide variety of cells without the use of transfection agents or long incubation times (4–6).

The first MRI-based cell tracking methods could generally only detect large number of magnetically labeled cells, whereas it would be useful to be able to detect individual cells when monitoring stem cell homing (7–9) or T-cell trafficking (10–12). Indeed, only a tiny fraction of injected cells commonly reach their target sites. High-resolution imaging methods capable of tracking one or a few cells in living organisms would have enormous advantages for the development of cell-based therapies.

Most attempts to improve MRI spatial resolution have involved the use of very high magnetic field strengths. This is because nuclear magnetization increases with field strength, and because a better signal/noise ratio yields better spatial resolution. Hoehn et al. obtained a detection limit of 500 cells at 7T after local intracerebral injection of stem cells (8). More recently, Stroh et al. achieved a detection limit of 100 stem cells in rat brains at 17.6T (13) and Kircher et al. a detection of three lymphocytes per voxel in tumor of living mice at 8.5T (12). The possibility of single-cell detection by high-field MRI has been established in vitro (14–16), in embryos (17) and, more recently, in living mouse liver (18) and heart (19). However, high-field MRI is incompatible with clinical use, as clinical magnets have field strengths below 3T (generally 1.5T).

Here we tested an alternative to high-field MRI for the detection of low amount of cell-internalized iron oxide nanoparticles in vivo using a 1.5T clinical MRI machine. The aim was to demonstrate the feasibility of detecting poorly labeled cells which have proliferated in vivo and have been recruited specifically in tumor after systemic administration. We used the well-known immunogenic mouse tumor model, in which ovalbumin-expressing tumors are infiltrated by transferred OVA-specific cytotoxic lymphocytes. High-field MRI allowed monitoring of lymphocytes migration in this model (5, 6, 12) and succeeded in detecting as few as 3 cells/voxel within the tumors in vivo (12). To achieve high resolution with a clinical 1.5T MRI device, we improved here the signal-to-noise ratio by reducing the noise generated by the detection coil. For this purpose, we used a recently developed cryo-cooled high-temperature-superconducting (HTS) surface coil (20, 21).

The animal model and labeling procedure of OVA-specific lymphocytes were the same as in our previous high-field study, in which we demonstrated the proliferation of AMNP-labeled lymphocytes in the spleen and monitored their trafficking from the spleen to the tumor (5). Lymphocytes were loaded with 1.3 pg of iron at time of intravenous injection and iron load per cell decreases after in vivo cell division. Here, 48 h after intravenous injection, we detected heterogeneously dispersed focal signal voids within the mouse tumors. These voids likely corresponded to individual infiltrating lymphocytes as shown by histological studies and comparison with in vitro phantoms experiments.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. EX VIVO HISTOLOGY
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

Animal Model

C57BL/6 mice were obtained from Elevage Janvier (Saint Isle, France). All experiments complied with French legislation and guidelines for animal research.

EG-7 ovalbumin-expressing lymphoma cells were injected subcutaneously into the left flank of C57BL/6 mice. After 15 days, the resulting tumors reached a volume of 200–500 mm3 and 3 × 106 magnetically labeled OT-1 lymphocytes (specific for ovalbumin antigen) diluted in RPMI medium (100 μL) were transferred to the mice by retro-orbital injection.

Magnetic Labeling of Cells

Lymphocytes expressing the T cell receptor specific for ovalbumin were obtained from OT-1 mice and stimulated with concanavalin A (2.5 ng/mL) and interleukin-2 (5 U/mL). The activated cells were labeled with anionic maghemite nanoparticles (AMNP) with a magnetic diameter of 8 nm and a negatively charged citrate coating. The cells were incubated for 15 min at 37°C in RPMI medium supplemented with 5 mM citrate sodium and with nanoparticles at iron concentrations of 0.5 mM, 1.5 mM, and 4 mM for in vitro experiment and of 4 mM for in vivo transfer. Citrate sodium was added to RPMI medium to avoid desorption of citrate ligands on the particles surface and thus ensure stability of the AMNP suspension. After two washing steps, the cells were incubated at 37°C in citrated RPMI for complete AMNP internalization (chase period). The iron load per cell was quantified by means of single-cell magnetophoresis (22). OT-1 lymphocytes were then cultured for 6 to 12 h with interleukin-2 before intravenous injection (3 × 106 cells) to C57BL/6 recipient mice with growing OVA-expressing EG-7 tumors. Trypan blue staining showed >95% viability after labeling and in vitro cell proliferation was identical for labeled and nonlabeled cells (5). As described in (5), the functionality and specificity of labeled OT-1 cells were demonstrated in vivo by partial regression of OVA-expressing EG-7 tumors after cell transfer.

MRI Acquisition

MRI was carried out on a 1.5T clinical whole-body imaging system (Signa, General Electric Medical Systems, Milwaukee, WI) at Centre Inter Établissement de Résonance Magnétique (CIERM, Hôpital Bicêtre, France). A standard whole-body gradient system delivering a maximum amplitude of 22 mT/m with a minimum rise time of 288 μs was used. High-resolution 3D MRI was performed using a 12-mm HTS surface coil similar to that used in (21) cooled to 80K and offering an unloaded quality factor close to 12,800 when positioned at the centre of the 1.5T MR scanner.

For in vitro cell detection, phantoms were made of 0.3% agarose gel (300 μL) containing one thousand dispersed lymphocytes labeled with extracellular iron concentrations of 0, 0.5, 1.5, and 4 mM for 15 min. The Q factor of the HTS coil loaded by the agarose gel was 12,600. A 3D-spoiled GRASS sequence was used, with a field of view (FOV) of 15 mm × 7.5 mm × 7.5 mm, an acquisition matrix of 256 × 128 × 128, an echo time (TE) of 18 ms, a repetition time (TR) of 68 ms, a flip-angle (FA) of 12 degrees and a bandwidth (BW) per pixel of 17.4 Hz. The transmit level, that is, FA, was manually adjusted during scout acquisition with a 2D Spin Echo protocol and a TR of 2 s to achieve the maximum signal at a 1- to 2-mm depth below the surface. The total acquisition time was 18 min, and the isotropic spatial resolution was (60 μm)3.

For in vivo imaging the mice were anesthetized with 2% isofluoran in a mixture of oxygen/nitrous oxide (50/50) administered by means of a face mask. The mice were held in a system equipped with a bed and positioned so as to maintain the tumor in contact with the HTS surface coil. The Q factor of the HTS coil loaded by the mouse tumor was 6330 ± 1190. Eight mice were imaged 48 h after transfer of OT-1 lymphocytes, four with unlabeled cells and four with AMNP-labeled cells labeled with 4 mM iron concentration for 15 min. 3D-spoiled GRASS sequence was used with a FOV 30 mm × 15 mm × 7.5 mm, an acquisition matrix of 512 × 256 × 124, a TE of 14 ms, a TR of 53 ms, a FA of 15 degrees, and a BW of 17.4 Hz. The total acquisition time was 29 min, and the isotropic spatial resolution was (60 μm)3.

We defined the relative signal loss in a voxel as ΔSI/SI = (SImean − SI)/SImean, where SI is the signal intensity in the voxel and SImean is the mean background signal intensity in the agarose gel or tumor.

Histology

After image acquisition, the mice were killed by pentobarbital overdose. The tumors were removed, fixed with paraformaldehyde and embedded in paraffin. Six-micrometer tissue sections were incubated for 30 min in 1% potassium ferrocyanide in 0.12 N HCl (Perls staining). Endogenous peroxidase activity was quenched for 20 min at room temperature on 0.3% H2O2 in methanol. Section were rinsed in PBS and incubated for 6 min in DAB (diaminobenzidine)/ H2O2. Iron-loaded cells were revealed by a brown coloration after DAB-amplified Perls staining.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. EX VIVO HISTOLOGY
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

In Vitro Single-Cell Detection

Mean iron load per cell, as measured by single cell magnetophoresis, was 0.2 ± 0.1, 0.8 ± 0.3, and 1.3 ± 0.6 pg after 15 min incubation with 0.5, 1.5, and 4 mM extracellular iron, respectively. Agarose phantoms containing dispersed labeled cells showed punctuate hypointense signals, whereas the gel with unlabeled cells gave a homogeneous signal with an SNR of 51.4 ± 3.7 (Fig. 1a–d). The number of signal voids was consistent with the number of cells in the phantoms confirming individual cell detection. The modulus signal loss was quantified in 20 different hypointense regions of each gel, both in the central voxel of 60 μm (defined as the darkest voxel of the region) and in the surrounding coronas. The relative signal loss decreased with increasing distance from the central voxel (Fig. 1e) and was no more significant compared with the background signal dispersion beyond the third corona. Signal losses in the central voxel and in the first and second coronas increased with increasing iron load per cell.

thumbnail image

Figure 1. MRI detection of single cells (3D-SPGR, voxel (60μm)3) dispersed in agarose gel. a–d: Overall views and zoom of gel with unlabeled lymphocytes (a,b) and with lymphocytes with 1.3-pg iron load (c,d). The tube diameter is 5 mm. A homogeneous signal is obtained with control cells, while magnetically labeled single cells appear as dark signal voids whatever iron load in the range 0.2–1.3 pg. In plane signal variations in a zoomed region of gel are represented as 3D plots (b,d). e: Signal loss relative to the mean background signal is quantified in the central voxel and in the 1st, 2nd, 3rd, and 4th corona around, for 20 different hypointense signal regions in each agarose gel containing cells with 0.2-, 0.8-, and 1.3-pg iron load, respectively. Bars represent the standard deviation of signal loss between the different analyzed signal voids. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

In Vivo Imaging of Intratumoral Lymphocytes

Three million OT-1 lymphocytes with an iron load of 1.3 ± 0.6 pg per cell were transferred to tumor-bearing mice by retroorbital injection. Imaging was performed 48 h later with a resolution of (60 μm)3. SNR was 34 ± 1.5 in the tumor of living mouse.

Figure 2a shows three successive slices of the tumor in a control mouse transferred with unlabeled cells. A homogeneous signal was obtained over the whole tumor, yielding the precise tumor volume. Figure 2b shows successive slices of a tumor in a mouse transferred with labeled cells. Numerous signal voids were seen throughout the tumor. The focal nature of the signal loss was confirmed by three-dimensional visualization of the tumor volume (Fig. 3a–c). Punctuate signal voids were clearly distinguishable from microvessels irrigating the tumor.

thumbnail image

Figure 2. In vivo MRI detection of labeled lymphocytes in a mouse tumor. a,b: Sequential MR images (3D-SPGR, voxel (60 μm)3) of the tumor in mice 48 h after injection of three million unlabeled lymphocytes (a) or the same number of magnetically labeled cells (b; iron load 1.3 pg/cell at the time of injection). Control tumors (a) give a homogeneous signal, whereas punctuate signal voids (white arrows) distributed throughout the tumor are observed in tumors of mice that received labeled lymphocytes (b). c,d: Zoom of the tumor image containing a signal void (d; labeled lymphocytes) or no signal void (c; control). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

thumbnail image

Figure 3. a–c: The 3D MRI imaging of a tumor in a mouse having received three million labeled cells (1.3 pg of iron per cell) 48 h before imaging, in the transversal (a), sagittal (b), and coronal (c) directions. The exact correspondence of the signal void in the three planes (white cross) shows that the hyposignals cannot be due to a microvessel. d: Signal loss relative to the mean background signal is quantified in the central voxel and in the 1st, 2nd, 3rd, and 4th corona around, for 20 different signal voids in the tumor. Bars represent the standard deviation of signal loss between the different analyzed signal voids.

Download figure to PowerPoint

With the same approach as that used in vitro, the signal loss was quantified for 20 different hypointense points of the tumor (Fig. 3d). On average, the signal loss was 83.7 ± 12.7% in central voxels of the tumor and 26.7 ± 15.4% in the first corona. There was no significant signal loss in the second corona and beyond.

EX VIVO HISTOLOGY

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. EX VIVO HISTOLOGY
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

To confirm that the signal voids were due to labeled cells having migrated to the tumor, tumor sections were stained with Prussian blue amplified with diamino-benzidine, which colors iron inclusions brown. Scarce stained cells were observed, with a heterogeneous distribution compatible with the MRI images (Fig. 4).

thumbnail image

Figure 4. DAB-enhanced Prussian blue staining of tumor sections 48 h after injection of labeled magnetic lymphocytes (×20 and ×40). Red arrows indicate iron-positive brownish cells. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. EX VIVO HISTOLOGY
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

The ability to track a specific cell population in vivo by means of noninvasive methods, and particularly MRI, is drawing increasing interest with the development of cell-based therapies for degenerative, malignant and genetic diseases. In particular, monitoring of tumor-antigen-specific lymphocytes could help to evaluate tumor targeting efficiency (23, 24). High-field MRI and magnetic labeling has been used in three studies as a dynamic method for monitoring T cell homing to OVA-expressing tumors (5, 6, 12). Tumor infiltration was detected in vivo as a heterogeneous decrease in tumor intensity on T2-weighted MR images at 8.5T (12), 9.4T (6), and 7T (5). T cell recruitment was confirmed by immunohistochemical and Prussian blue staining of tumor sections. In the work by (12), the decrease in the MR signal varied with the time after adoptive T cell transfer, with a maximum at 48 h. Using quantitative autoradiography and T2 mapping, a detection threshold of 3 cells/voxel was determined. In the same model as that used here, we previously showed, using 7T MRI, that intravenously transferred AMNP-labeled OT-1 lymphocytes homed within the first 24 h to the spleen, where they divided before migrating to the tumor (5). Ex vivo 9.4T high-resolution imaging of the tumors confirmed the heterogeneous distribution of signal voids corresponding to labeled cells. The recruitment of labeled T cells was highly specific, as only tumors expressing OVA antigen regressed after cell transfer. We show here that single-cell resolution can be achieved in vivo in the tumor at iron load lower than 1 pg using a recently developed HTC coil and a clinical MRI system. Our study overcomes different methodological difficulties: we used a field strength of only 1.5T, cells had low uptake capacity and proliferated in vivo and cell migration was highly specific.

We labeled OT-1 lymphocytes with dextran-free anionic citrate-coated maghemite nanoparticles for a very short time (15 min) without using a transfection agent (5). This simple labeling protocol yielded a maximum iron load of 1.3 pg per lymphocyte. As shown with other cell lines (4, 26), the labeling efficiency is due to the nonspecific affinity of negatively-charged polymer-free nanoparticles for the cell outer membrane. The low uptake by lymphocytes compared with other cell lines can be explained by their small size (and small membrane surface area) and by their low capacity for internalization (5).

We first show that cells with 0.2 pg of iron load can be detected individually in agarose phantoms with a contrast to noise ratio of approximately 40 by using a 12-mm HTS coil and a 1.5T clinical scanner. Achieving a resolution of (60 μm)3, signal loss due to individual labeled cell was increasing with cell iron load from 0.2 pg to 1.3 pg. Profiles of signal loss clearly distinguished different iron contents as shown in Figure 1e. Using comparable MRI parameters (gradient echo time and spatial resolution), we then detected isolated focal signal voids in the tumors of mice transferred with labeled lymphocytes 48 h previously. No such spots were detected in control tumors. Only the central voxel usually showed a marked decrease in MR signal intensity (>80%) with a contrast to noise ratio of approximately 30. The volume of the perturbed protons was always within the range of (60 μm)3-(180 μm)3. When compared with in vitro experiments, signal voids observed in vivo had comparable size and intensity as for in vitro single cells containing between 0.2 pg and 0.8 pg of iron. Thus the dark spots in the tumor likely reflect the punctual presence of less than 1 pg of iron, either in single cell or shared by contiguous cells. This is consistent with our previous finding that lymphocytes divide several times in the spleen before migrating to the tumor, thereby reducing their iron content to sub-picogram levels (between 0.2 and 0.3 pg of iron 48 h after injection) (5). Histological examination revealed a very small number of isolated DAB-enhanced Perls stained cells distributed throughout the tumor. Although it was not possible to superimpose histological sections and MRI images, this finding is consistent with signal voids corresponding to single cells or to very small clusters of poorly labeled cells.

In previous studies, single-cell detection in vivo was achieved using cells with high uptake capacity, which have been locally implanted or delivered nonspecifically to the target tissue (18, 19, 27). Single hepatocytes have been detected in the liver after splenic transplantation (18) and single macrophages have been detected in the brain after nonspecific ventricular delivery (27). In both cases, the cells were loaded with more than 50 pg of iron, by using magnetic microparticles and nanoparticles, respectively. In another study, individual macrophages labeled in situ with magnetic microparticles were detected in grafted hearts undergoing rejection (19).

Here high sensitivity to local magnetization (<0.2 pg or iron) was due to the high spatial resolution achieved in vivo. In vitro studies and simulations (28–30) have shown that the cell detection threshold is strongly dependent on the spatial resolution. We achieved a resolution of (60 μm)3 and an SNR of 34 in mouse tumors, with a scan time of only 29 min. SNR was optimized by reducing both the noise of the coil and the noise inductively coupled to the coil by the sample, thanks to the combination of a small coil diameter and of a HTS material at intermediate field strength (20, 21) The coil used in this study was made of YBaCuo deposited on sapphire. Compared with an equivalent copper coil operating at room temperature, the gain in SNR obtained with our 12-mm HTC coil was approximately 18-fold for an ideal, loss-less sample and typically 8-fold for the living mouse tumors in the present study.

Only one previous study has achieved single-cell detection with a 1.5T MRI device (27), involving a cell iron load larger than 60 pg, an optimized FIESTA pulse sequence and a high-performance gradient coil technology. In addition to the conventional high-field approach to improve the SNR (18, 19), there are several advantages to using intermediate MR field strengths for cellular imaging. The first is clearly the possibility of using current clinical MR units. In future, phased arrays of small cryocooled HTS coils could provide high-resolution images of a larger region of interest. Other advantages include low cost, safety and wide bore. More importantly, using a field strength above 1T will no more increase the local field perturbation created by a labeled cell, as nanoparticle magnetization already attained saturation. Conversely, increasing the field strength requires the use of short echo times to maintain susceptibility artifacts below acceptable levels, while the dephasing effect due to the field perturbation increases with the echo time (29). Hence, cell detectability could be optimized by using a moderately low field strength and a relatively long echo time as in the present work.

In conclusion, we achieved an unprecedented spatial resolution by using a 1.5T clinical MRI device and an HTS coil, allowing us to visualize individual cells infiltrating mouse tumors in vivo. This approach opens interesting perspectives for in vivo monitoring of therapeutic cell products, especially rapid proliferating cells or cells with a low capacity for endocytosis.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. EX VIVO HISTOLOGY
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

We thank Dr. C. Ménager for providing us with the nanoparticles, Pr. P. Bruneval for performing histological analysis, Dr. C. Combadicie for fruitful discussion, and Pr. J. Bittoun for giving us access to the MRI CIERM facility.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. EX VIVO HISTOLOGY
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES