Magnetic resonance imaging (MRI) allows noninvasive and three-dimensional visualization of whole organisms over time, and, therefore, would be ideally suited to monitor cell trafficking in vivo. Until now, systemically injected cells had been difficult to visualize by MRI because of relatively inefficient labeling methods. We developed a novel, biocompatible, and physiologically inert nanoparticle (highly derivatized cross-linked iron oxide nanoparticle; CLIO-HD) for highly efficient intracellular labeling of a variety of cell types that now allows in vivo MRI tracking of systemically injected cells at near single-cell resolution. CD8+ cytotoxic T lymphocytes labeled with CLIO-HD were detectable via MRI with a detection threshold of 2 cells/voxel in vitro and ∼3 cells/voxel in vivo in live mice. Using B16-OVA melanoma and CLIO-HD-labeled OVA-specific CD8+ T cells, we have demonstrated for the first time high resolution imaging of T-cell recruitment to intact tumors in vivo. We have revealed the extensive three-dimensional spatial heterogeneity of T-cell recruitment to target tumors and demonstrated a temporal regulation of T-cell recruitment within the tumor. Significantly, our data indicate that serial administrations of CD8+ T cells appear to home to different intratumoral locations, and may, therefore, provide a more effective treatment regimen than a single bolus administration. Together, these results demonstrate that CLIO-HD is uniquely suited for quantitative repetitive MRI of adoptively transferred cells and that this approach may be particularly useful for evaluating novel cell-based therapies in vivo.
Cell-based therapies (1) have received much attention as novel therapeutics for the treatment of cancer (2, 3), autoimmune (4, 5), cardiovascular (6), inflammatory (7), and degenerative diseases (8, 9, 10). A number of native cells (4, 7, 11), antigen-specific T-lymphocytes (12, 13), or, more recently, stem and progenitor cells have been used for these approaches. These cells alone (14, 15) or armed with additional transgenes (16, 17, 18) have been effective in mediating tumor regression in vivo. Recent advances allowing more efficient ex vivo expansion of such cells could have far reaching implications in a number of therapeutic paradigms; however, there remains a need to better understand the in vivo fate of injected cells, including their distribution, migration, and homing to targeted sites.
Tumor antigen-specific lymphocytes in particular have been used for adoptive transfer and treatment in lymphoma, melanoma, and other malignancies (19, 20, 21, 22, 23, 24, 25, 26, 27). A major obstacle to accurate evaluation of treatment efficacy and antitumor effects has been the inability to track these CTLs4 in vivo at sufficiently high spatial and temporal resolutions (28). The majority of imaging approaches available have involved either mass distribution analysis of radiolabeled cells (29, 30) or bioluminescence imaging of cells stably transfected with luciferase (31) at relatively low spatial resolution, or for greater spatial resolution, invasive intravital microscopy analysis of small, relatively superficial areas of the tumor (32). For an imaging method to ultimately be clinically viable and allow the evaluation of both cell delivery and therapeutic effectiveness in patients, it must be noninvasive, nontoxic, and allow an accurate and quantitative determination of the cell-based therapy. Both MRI and nuclear imaging (positron emission tomography and single-photon emission computed tomography) are in routine clinical use, but nuclear imaging has limited spatial resolution, often requires genetic modification of the administered cells (33), and most radiochemicals have significant cellular toxicity and short half-lives that allow imaging for only 24–48 h. In contrast, MRI is both noninvasive and provides high spatial resolution in vivo. However, until now, adoptively transferred cells have been extremely difficult to visualize via MRI because of a combination of relatively inefficient labeling methods and the dilution of systemically injected cells in vivo. Most published reports have followed the migration of either locally injected cells labeled with different superparamagnetic iron oxide nanoparticles into the adjacent parenchyma (34, 35) or have examined nonspecific accumulation of magnetically labeled cells in whole organs at short time points (36); however, labeling and imaging of immune-specific cells by MRI has not yet been described. Indeed, at present there exists no high-resolution, three-dimensional method to visualize the immune-specific recruitment of systemically administered cells in vivo over time.
Initially, we developed an HIV Tat peptide derivatized magnetic nanoparticle that allows efficient intracellular labeling for MRI (37). Recently, by exploiting principles of multivalency (38) and by improving conjugation strategies, peptide sequences and labeling protocols we developed a nanoparticle (CLIO-HD) that is 200-fold more efficient in lymphocyte labeling. Here, using a model antigen-expressing tumor, we demonstrate quantitative, high-resolution in vivo imaging of CLIO-HD-labeled antigen-specific CTL recruitment to tumors in live animals. Using high resolution MRI and three-dimensional reconstructions, we have revealed the heterogeneous three-dimensional distribution of the recruited cells within the tumor. In addition, we have observed that repeated injections of CTL were apparently recruited to different anatomical regions of the tumor, suggesting that multiple dosing strategies may allow attack of the tumor from multiple fronts simultaneously. Together, these results illustrate that the developed nanoparticle enables quantitative high-resolution repetitive MRI of adoptively transferred cells, and may be particularly useful for tailoring cell-based anticancer therapies and evaluating therapeutic effectiveness in both experimental and clinical settings.
MATERIALS AND METHODS
DMEM, RPMI 1640, HBSS, and fetal bovine serum (for culture of tumor cells) were purchased from Cellgro (Herndon, VA). FCS (CD8+ T cells) was purchased from Sigma (St. Louis, MO). DPBS, DPBS with Ca2+ and Mg2+, and PBS were obtained from BioWhittaker (Walkersville, MD). Recombinant murine IL-2 was purchased from R&D (Minneapolis, MN). The OVA-derived MHC class I immunogenic peptide OVA257–264 (SIINFEKL) was synthesized by Analytical Biotechnology Services (Boston, MA).
Anti-CD3ε (clone 145-2C11), anti-CD28 (clone 37.51), FITC-anti-CD8b (clone 53-5.8), phycoerythrin-anti-CD3 (clone 17A2), rat antimouse CD8a (clone 53-6.7, IgG2a,κ), and purified rat IgG2a were purchased from BD PharMingen (San Diego, California). Biotinylated goat antirat IgG monoclonal antibody was obtained from Jackson Immunoresearch Labs (West Grove, Pennsylvania).
C57Bl/6 (female, 10–14 weeks) were purchased from the National Cancer Institute (Bethesda, MD), OT-I mice were kindly provided by William R. Heath and Francis Carbone (Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia; Ref. 39). All of the animals were maintained in our pathogen-free institutional facilities.
OVA-transfected B16 melanoma cell line (B16-OVA) and B16 melanoma cell line (B16F0) were provided by Drs. Edith Lord and John Frelinger (University of Rochester, Rochester, New York; Ref. 40). Both cell lines were cultured in DMEM containing 10% FCS, 0.075% sodium bicarbonate, and antibiotics, and serially passaged (1:10 split ratio) as required. CD8+ T cells were isolated from the spleens and lymph nodes of mice using the MACS separation system (Miltenyi Biotec, Auburn, CA) and were cultured in DMEM containing 10% FCS, nonessential amino acids, 1 mm sodium pyruvate, 0.075% sodium bicarbonate, 50 μm 2-mercaptoethanol, and antibiotics. For selected in vitro experiments, wild-type CD8+ T cells were proliferated in plates precoated with anti-CD3ε (10 μg/ml), supplemented with 2.4 ng/ml IL-2 and 2 μg/ml anti-CD28 for two rounds of 3 days (41). For functional assays and in vivo experiments, OT-I CD8+ T cells were stimulated with 0.7 μg/ml SIINFEKL-peptide and 10-fold mitomycin C (50 μg/ml)-treated syngeneic antigen presenting cells (APC) fed on day 3 with 0.8 ng/ml IL-2 and harvested on day 5. Murine heart endothelium was isolated and cultured as described previously (42). Endothelium was plated at confluence on 25-mm diameter glass coverslips precoated with 1 μg/cm2 human fibronectin, activated for 4 h with 120 ng/ml murine tumor necrosis factor (mTNF)-α, and then used in experiments.
Synthesis of CLIO-HD, Cell Labeling, and Uptake Studies.
The CLIO nanoparticle has been described previously (43). CLIO-SC-Tat preparations were synthesized as described (38) with modifications, with Tat:CLIO ratios (6.4, 7.5, 11.73, 19.5, 22.95, and 23.24 Tat/CLIO), and routinely used with 23 Tat/CLIO (CLIO-HD). CD8+ T cells were incubated with CLIO-HD (2–300 μg Fe/ml/10 × 106 cells) for 4 h at 37°C and washed three times by centrifugation through 40% Histopaque-1077. CLIO-HD-labeled OT-I CD8+ T cells were fixed in 1% paraformaldehyde in PBS for 30 min at 4°C, washed in HBSS, and permeabilized in 0.1% Triton X-100. Cells were then washed and resuspended in 1 mg/ml RNase in PBS for 15 min incubated in PBS-propidium iodide solution (10 μg/ml) for 15 min at room temperature, washed in HBSS, and mounted using Vectorshield (Vector Labs, Burlingame, CA). Dual-channel laser scanning confocal microscopy was performed using a Zeiss LSM 5 PASCAL (Zeiss, Thornwood, NY). Quantification of CLIO-HD uptake was performed as described previously (38).
CD8+ T-Cell Functional Assays.
Toxic effects of CLIO-HD labeling were assessed by functional assays of proliferation. OT-I CD8+ T cells at day 5 of culture were labeled with 75 μg, 150 μg, or 300 μg/ml CLIO-HD or vehicle control for 4 h. Cells were then proliferated as described above in 96-well plates (5 × 104 cells/well). Positive control wells received 10 μl/ml phytohemagglutinin (Life Technologies, Inc.). After 48 h, all of the wells were pulsed for 18 h with 1 μCi of [3H]thymidine (Perkin-Elmer, Boston, MA) and then harvested onto filter mats. [3H]Thymidine incorporation was determined by liquid scintillation counting (44). Target cells (B16F0 and SIINFEKL-coated B16-OVA cells) were labeled with 20 μCi of 51Cr (Perkin-Elmer) for 1 h at 37°C and washed. CLIO-HD-labeled (300 μg/ml for 4 h) or vehicle-labeled OT-I CD8+ T cells were titrated in 2-fold dilutions in 96-well flat-bottomed plates (100 μl volume), mixed with 100 μl of target cells, and incubated for 4 h at 37°C. Total 51Cr release was obtained by adding 0.25% Triton X-100 (final concentration) to the wells. Supernatants were collected and radioactivity quantitated using a gamma counter. Percentage of specific lysis was calculated according to the following formula: percentage of specific lysis = [(experimental cpm − spontaneous cpm)/(total cpm − spontaneous cpm)] × 100 (45). Freshly isolated OT-I CD8+ T cells were labeled as described above with either CLIO-HD or vehicle control and then perfused across 4 h mTNF-α activated murine heart endothelium (106 cells/ml in flow buffer − 1:1 DPBS with Ca2+ and Mg2+:DPBS− containing 0.1% BSA) for 10 min at 0.52 ml/min (estimated wall shear stress of 1.0 dynes/cm2) using a parallel plate flow chamber (42). T cell:endothelial interactions were recorded using video-linked phase contrast microscopy and assessment of attachment, rolling, adhesion, and migration determined offline from four to eight high power fields per coverslip. A minimum of three coverslips per condition were quantified in each experiment.
Purity of OT-I CD8+ T cells was determined by staining with anti-CD8-FITC and anti-CD3-PE using standard protocols (46) and analysis using a FACScalibur (Becton Dickinson, Mountain View, CA). To determine the efficiency of CLIO-HD uptake, labeled cells were analyzed for fluorescence in the FITC channel and compared with an unlabeled control. A minimum of 10,000 cells/sample was analyzed.
C57Bl/6 mice were injected s.c. with 5 × 106 B16-OVA or B16F0 in the right or left flank, respectively, and 10–12 days later, 1–3 × 107 CLIO-HD-labeled OT-I CD8+ T cells were adoptively transferred via i.p. injection. Distribution of CLIO-HD-labeled cells over time was assessed via repetitive MRI. In some experiments, tumors were subsequently excised and used for histological analysis. In additional experiments, tumor-bearing C57Bl/6 mice were adoptively transferred with CLIO-HD/111In-oxine (600 μCi/5 × 108 cells/2 ml) dual-labeled OT-I CD8+ T cells and sacrificed after 12 or 36 h. For biodistribution studies, the organs were excised, weighed, and radioactivity counted to determine the percentage of injected dose/g of organ tissue. For autoradiography studies, tumors were excised from the 36-h animals, fixed in 4% paraformaldehyde, and imaged by MRI (500 μm slice thickness). Subsequently, tumors were cut into slices (same orientation and slice thickness as MR images) and exposed on a phosphorimager (Molecular Dynamics, Sunnyvale, CA) to reveal the distribution of radiolabeled OT-I CD8+ T cells.
MRI and Three-Dimensional Reconstructions.
Under general isoflurane anesthesia (0.5–1.5% at 2 liter/min), mice were imaged at 0, 12, 16, and 36 h after adoptive transfer of CLIO-HD-labeled CD8+ cells (Bruker DRX 360, 8.5 T magnet, 2-cm diameter birdcage-coil, T2-weighted spin-echo sequences; TR = 3000 ms; TE = 15–60 ms; matrix size 256 × 256; in-plane resolution 75 μm; slice thickness 500 μm; NEX = 2; imaging time 25 min). Phantoms were imaged with identical parameters, except TE = 10–240 ms in 10 ms intervals Carr-Purcell-Meiboom-Gill sequence (CPMG). Distribution of 111In-oxine/CLIO-HD dual-labeled OT-I CD8+ T cells was determined by MRI, and correlated with autoradiography of identical tumor sections. Image segmentation, T2 analysis and three-dimensional volume rendering were performed using CMIR-Image (developed in Interactive Data Language; Research Systems Inc., Boulder, CO). ROIs were defined manually, and serial images of a subject were registered using a rigid body algorithm to generate optimal coincidence of the tumor ROIs for each dataset. T2 maps were constructed by performing fits of a standard exponential relaxation model to the data on a pixel-by-pixel basis within the ROIs. Only pixels with intensity greater than a threshold level (mean background +5× SD) were considered during the fitting process. Registered T2 maps of each mouse were color mapped and rendered using α blending, and then superimposed on transparent outlines of the mouse surface and the segmented tumors. For data in Fig. 6, tumor volumes with T2 changes of >20 ms (corresponding to 2 cells/voxel) were extracted from the registered T2 maps of the mouse after each injection of CLIO-HD-labeled CD8+ T cells. These volumes were rendered as red, green, and blue, respectively, and superimposed on a transparent outline of the tumor surface. Renderings were performed at multiple angles to highlight the three-dimensional nature of the calculated T2 maps.
After MRI, animals were sacrificed, and B16-OVA and B16F0 tumors excised preserving the capsule, fixed for 2 h in 2% paraformaldehyde, and equilibrated in 18% sucrose, embedded in OCT (Sakura, Torrance, CA), snap-frozen in liquid N2, and sectioned in 10 μm slices. Tissue sections were incubated with antimouse CD8a or isotype-matched control monoclonal antibody. Intrinsic tissue avidin or biotin was blocked (avidin/biotin blocking kit; Vector Labs), and biotinylated goat antirat-IgG was used to detect the 1° antibody, and visualized using Vector ABC Elite and Vector NovaRed (Vector Labs). Sections were counterstained with Gills No.2 Hematoxylin (Sigma), dehydrated, mounted with Vectashield, and examined via bright field (Nikon) with ×20 magnification. Serial sections were viewed unstained in the FITC channel using an inverted epifluorescence microscope (Zeiss Axiovert). For Prussian Blue staining, the same sections were then incubated for 30 min with 2% potassium ferricyanide (Perls’ reagent) in 2% HCl, washed, and counterstained with Eosin.
Labeling of OT-I CD8+ T Cells with CLIO-HD Does Not Influence Cell Behavior in Vitro.
To optimize the labeling protocol for CD8+ T cells, we examined the kinetics of CLIO-HD uptake in these cells. The uptake of CLIO-HD by CD8+ T cells was proportional to both the incubation time and CLIO-Tat concentration (Fig. 1,B). Using the FITC label incorporated into the Tat peptide, we were able to additionally validate the correlation between incubation time and labeling efficiency via FACS analysis (Fig. 1,C), confirming that the cells were uniformly labeled by CLIO-HD. When we examined the intracellular distribution of CLIO-HD via laser-scanning confocal microscopy, we determined that the particle localized to both the nucleus and cytoplasm of CD8+ T cells (Fig. 1,D). Intracellular labeling of CD8+ T cells with CLIO-HD was not cytotoxic (Trypan Blue exclusion) in concentration ranges of 50–300 μg Fe/ml/106 cells (data not shown), with cells retaining >95% viability. In additional experiments, we examined the effects of CLIO-HD labeling on OT-I CD8+ T cell functions of in vitro tumor cell killing (Fig. 1,E) and proliferation (Fig. 1,F). Treatment of OT-I CD8+ T cells with up to 300 μg Fe/ml/106 cells did not affect their ability to kill target cells or to proliferate in vitro. In addition, we examined the interactions of OT-I CD8+ T cells with 4 h mTNF-α activated murine heart endothelium under in vitro flow conditions, and did not observe any significant difference in their initial rate of attachment, rolling fraction, stable arrest, or transmigration (Table 1). CLIO-HD-labeled CD8+ T cells exhibited a total accumulation of 217 ± 26 cells/mm2 after 15 min of perfusion, compared with an accumulation of 217 ± 46 cells/mm2 for unlabeled cells, and 6.0 ± 0.6% of CLIO-HD labeled cells underwent transmigration compared with 5.2 ± 0.9% unlabeled cells.
In Vitro Detection Threshold of CLIO-HD-Labeled OT-I CD8+ T Cells and Stability of Intracellular Labeling.
To determine the threshold of detection for CLIO-HD-labeled OT-I CD8+ T cells and to create a T2 map standard curve, we next performed MRI phantom in vitro experiments. CD8+ T cells were labeled with CLIO-HD (300 μg Fe/ml/106 cells for 4 h) and mixed with unlabeled CD8+ T cells in the ratios described (Fig. 2,A). Cell pellets were implanted in agar plugs and imaged by MRI using the identical sequences used for in vivo experiments. MRI signal intensity reduction induced by CLIO-HD was observed with as few as 3 × 104 labeled cells/50 μl (Fig. 2,A), representing an in vitro detection threshold of <2 cells/voxel. These data were used to prepare a standard curve of 1/T2 against Fe concentration (Fig. 2,B) that would be used later to quantitate cell numbers within MRI slices. In additional in vitro experiments we examined the stability of CLIO-HD-mediated signal reduction over time in CD8+ T cells. Isolated CD8+ T cells were labeled with CLIO-HD (300 μg Fe/ml/106 cells for 4 h), and 8 × 105 cells placed into each well of a 24-well plate and cultured as described in “Materials and Methods.” At the intervals indicated (Fig. 2,C), total cells from each well were washed, counted, and implanted in agar plugs with unlabeled cells (8 × 105) used as a control. Cells were then imaged by MRI using identical sequences as those described above. The signal reduction caused by the presence of CLIO-HD is similar in each of the wells (Fig. 2,C), indicating that intracellular labeling of OT-I CD8+ T cells with CLIO-HD was maintained for at least 120 h. Furthermore, when resting OT-I CD8+ T cells and activated OT-I CD8+ T cells were labeled with CLIO-HD, there was no detectable difference in the degree of MR signal reduction induced by CLIO-HD (Fig. 2 D), confirming that this technique is equally applicable to the study of either resting or activated T-cell trafficking.
Recruitment of OT-I CD8+ T Cells to B16-OVA in Vivo Can Be Visualized by MRI.
We next performed in vivo experiments in which C57Bl/6 mice were implanted with both control B16F0 (left side) and the antigen-presenting B16-OVA (right side) melanoma in the thigh so that each animal served as its own control (n = 8). After 10–12 days, when the tumors had reached 5–10 mm diameter, serial MR image slices were taken of the animals before adoptive transfer (Fig. 3,A). OT-I CD8+ T cells were then labeled with CLIO-HD as described, and 3 × 107 cells were injected i.p. into the recipients. Animals were imaged again at 12, 16, and 36 h after adoptive transfer (Fig. 3, B–D), and the T2 data used to create three-dimensional reconstructions of the images (Fig. 3, E–L). In one representative example shown in Fig. 3, very little change in signal intensity was observed in the control B16F0 tumor, indicating that few OT-I CD8+ T cells had been recruited. In contrast, a significant signal reduction (dark areas) was observed in some regions of the B16-OVA tumor at 12 h after adoptive transfer. At later times, a significantly greater signal reduction was observed (Fig. 3, C, D, G, and H), reflecting the increase of CD8+ T-cell recruitment across the tumor. The heterogeneous nature of the T-cell recruitment is especially noticeable in the three-dimensional reconstructions (Fig. 3, E–I) of the MRI slices, and in the additional axial, sagittal, and coronal slices presented (Fig. 3, J–L). The mean MRI signal intensity also decreased with time after adoptive transfer, indicating the continued recruitment of CLIO-HD-labeled OT-I CD8+ T cells to the tumor. For a 360° animation of the data see web-movie 1 in Supplemental Data.
To control for signal intensity reduction caused by tumor necrosis, we performed parallel experiments with unlabeled OT-I CD8+ T cells and saw no significant change in signal intensity within the time frame of the experiment (data not shown). Recruitment of unlabeled cells to B16-OVA tumors was confirmed by histology (data not shown). When we performed biodistribution studies of CLIO-HD/111In-oxine dual-labeled OT-I CD8+ T cells at 12 and 36 h after injection, we determined that ∼25% of T cells accumulated in the spleen, 7% in the liver, and 2% in the lung, with 6% in the B16-OVA tumors, compared with 2% in the B16F0 tumors (expressed as % injected dose/g of tissue). Therefore, OT-I CD8+ T cells demonstrated a 3:1 difference in recruitment to the target tumor as compared with the null tumor.
Signal Reduction by MRI Correlates with Recruitment of CLIO-HD-Labeled OT-I CD8+ Cells.
To confirm that the signal reduction within the B16-OVA tumor was because of the specific recruitment of CLIO-HD-labeled OT-I CD8+ T cells, we performed MR imaging of recipient mice carrying both B16F0 and B16-OVA tumors, 12 h after adoptive transfer of labeled T cells. Subsequently, both tumors were excised, sectioned, and used for histological analysis. A representative axial MRI slice (Fig. 4,A) was correlated with serial sections of the same region of each tumor, staining for the presence of CD8+ T cells (CD8a, Fig. 4, B–D), CLIO-HD (FITC visualization, Fig. 4, F–H), and Fe content (Fig. 4,I). In the control B16F0 tumor, very few CD8+ T cells could be visualized (Fig. 4,B); however, there were numerous CD8+ T cells present within the B16-OVA tumor (Fig. 4, C and D), the density of which correlated with the degree of signal reduction observed via MRI. Because CD8a antibody is not specific for OT-I CD8+ T cells, but will also stain host CD8+ T cells, we confirmed the presence of adoptively transferred OT-I CD8+ T cells by examining the presence of CLIO-HD via fluorescence microscopy (Fig. 4, F–H) and Prussian Blue staining (Fig. 4,I). There was no visible FITC staining present within the B16F0 tumor (Fig. 4,F), but the presence of CLIO-HD within the B16-OVA tumor (Fig. 4, G and H) correlated with both CD8a staining and the MRI signal reduction observed. In a parallel experiment, CLIO-HD/111In-oxine dual-labeled OT-I CD8+ T cells were adoptively transferred into a C57Bl/6 mouse carrying B16F0 and B16-OVA tumors. After 36 h, the tumors were excised, sectioned, and imaged via MRI. The identical sections were then exposed for autoradiography. As can be seen in Fig. 4, J and K, the regions of MRI signal reduction present within the B16-OVA tumor correlated with the darkened areas observed via autoradiography, indicating that the recruited OT-I CD8+ T cells were responsible for the MRI signal reduction observed. Using quantitative autoradiography we also validated MRI measurements of accumulated cells and determined the in vivo detection threshold to be <3 cells/voxel.
To determine the in vivo longitudinal limit of detection for CLIO-HD-labeled OT-I CD8+ T cells within the target tumor, we performed an additional series of experiments. We adoptively transferred CLIO-HD-labeled OT-I CD8+ T cells into recipient tumor-bearing mice and continued to image these animals up to 60 h after administration (Fig. 5). We observed that the changes in signal intensity observed via MRI persisted for 48 h (Fig. 5,A) and then receded by 60 h (Fig. 5,B), suggesting that CLIO-HD-labeled OT-I CD8+ T cells could be detected in the tumor for a maximum of 48 h. The loss of MR detectability by 60 h in vivo could be because of several factors, including: (a) dilution because of rapid cell division; (b) emigration of OT-I cells from the tumor; (c) OT-I cell apoptosis and removal; or (d) intracellular biodegradation of the superparamagnetic iron oxide core. From our in vitro experiments (Fig. 2 C) we had observed that CLIO-HD-induced MRI signal reduction persisted for at least 120 h in actively proliferating cells; however, OT-I CD8+ T cells divided only once every 44.5 h in vitro, compared with division of up to once every 4–6 h in vivo. This difference in proliferation rate could explain the discrepancy between the in vitro and in vivo temporal detection limits.
Serial Administrations of Effector Cells Are Recruited Heterogeneously into the Target Tumor.
We hypothesized that serial doses of adoptively transferred cells might provide a better therapeutic approach than a single bolus, but because of the dynamic environment within the tumor we aimed at investigating the intratumoral recruitment of these cells with each administration. In the next experiments we used the same tumor model with C57Bl/6 mice (n = 3) carrying both B16F0 and B16-OVA tumors and performed serial injections of 107 CLIO-HD-labeled OT-I CD8+ T cells into the same animal at different times (0, 48, and 96 h). Mice were imaged via MRI before adoptive transfer, and at 12, 60, and 108 h (12 h after each cell administration). The selection of these time points allowed the signal changes induced by the previous injection to return to baseline before the next injection. Representative axial MR images are shown in Fig. 6, A–D, indicating a signal reduction in the B16-OVA tumor because of recruitment of CLIO-HD-labeled T cells, but not in the B16F0 tumor. The image slices for each time point were then reconstructed into three-dimensional volumes, and the signal intensity reduction for each administration color coded (Fig. 6, E–H) in red (0 h), green (48 h), and blue (96 h). When the three reconstructions were merged (Fig. 6 E) there was very little overlap between the color-coded maps, indicating that the cells in each administration had been recruited to different regions of the tumor mass. A 360° animation of this data is shown in the Supplemental Data web-movie 2.
Previous reports have demonstrated the feasibility of intracellular labeling with iron oxide particles (37) and imaging of locally injected cells (34). However, systemically administered cells undergo huge dilutions, and existing intracellular magnetic labels and protocols have not allowed us to track cell recruitment in live animals. In our own experience, previous studies have achieved limited success and have required a combination of high field strengths (9–14 T), long imaging times (>8 h), and ex vivo imaging of resected tissues (37, 47). Currently there exists no high-resolution three-dimensional method to quantitate the recruitment of systemically administered cells over time. In the present study, we have used an improved superparamagnetic particle (CLIO-HD) and optimized the labeling protocol to efficiently label lymphocytes, at levels that can be detected in vivo via MRI, that are not cytotoxic and that do not influence cell behavior or effector function. CLIO-HD-labeled OT-I CD8+ T cells remained >95% viable and exhibited the same profile of proliferation in vitro as unlabeled cells. In addition, their interactions with activated endothelial monolayers in an in vitro flow model were identical to those observed for unlabeled cells, indicating that CLIO-HD does not influence expression of, or the activation of, cell surface adhesion molecules or cytokine receptors required for CD8+ T-cell recruitment. CLIO-HD-labeled CD8+ T cells retained the ability to kill target cells in vitro and in vivo, indicating no impairment of their cytotoxic capacity by CLIO-HD labeling.
The present study has indicated significant intratumoral heterogeneity in the recruitment of CTL from the systemic circulation. In our antigen-specific model at least, OT-I CD8+ T cells were recruited primarily to focal regions within the tumor and areas of the tumor capsule, with some areas of the tumor apparently spared of OT-I CD8+ T-cell accumulation. Currently very little is known regarding the intratumoral distribution of infiltrating cytolytic cells. In experimental studies in rats carrying mammary tumors, CD8+ T cells were observed largely in the tumor periphery (48), with limited infiltration into the tumor mass. A similar pattern was observed in human tissue sections from B-cell non-Hodgkin’s lymphoma patients (49). However, these studies use histological analyses of limited serial sections, and do not describe the true three-dimensional distribution of CD8+ T cells within the whole tumor. One of the most powerful aspects of our approach is the ability to describe the distribution of infiltrating CD8+ T cells in three dimensions, across the entire tumor simultaneously, and in a quantitative and repetitive manner in the same animal. Thus, we were able to determine the true three-dimensional heterogeneity of CTL recruitment not fully appreciated in previous studies. Furthermore, the ability to examine both cellular recruitment and therapeutic response (tumor volume changes) using the same imaging modality is highly desirable. Consistently we observed continued rapid growth of the control tumor and tumor reduction or tumor stasis of the targeted antigen-expressing tumor, using data from the same MR images analyzed for CTL recruitment (Fig. 7).
When we followed the recruitment of CTL over longer time periods, we observed that the intratumoral pattern of CD8+ T-cell recruitment also exhibited a temporal heterogeneity. This shift in the CD8+ T-cell recruitment pattern appeared to correlate with the time course of recirculation through the draining lymph nodes (data not shown), and most likely reflects additional CD8+ T cell activation within the lymph node environment. However, the mechanisms that mediate the heterogeneity of this recruitment have yet to be elucidated, and this technology provides an avenue to investigate this additionally.
Clinically, cell-based therapies for cancer have attracted a great deal of attention in recent years. In particular, dendritic cells (2, 50) and tumor antigen-specific T cells (3, 13, 24) have been considered the best candidates for this type of approach, used either alone, in combination with cytokines, or armed with transgenes. However, this type of treatment protocol is far from optimized, and insight into the molecular mechanisms that mediate the recruitment of these cells to their target, their proliferation, and cytolytic activities within the tumor would provide invaluable information to improve these therapies. Recent data have indicated that particular subtypes of CD8+ T cells may exhibit differences in their temporal regulation of recruitment to and persistence at a tumor site (51), and that the effectiveness of these specific subtypes in tumor regression may be mediated via different mechanisms. The ability to effectively evaluate the roles of different T-cell subsets in tumor regression would provide unique insight for the development of novel cell-based therapies. This approach is uniquely suited to address these questions and would allow an investigator to readily evaluate the responses to specific cell-based therapies and develop specific dosing schedules for these therapeutic strategies.
In short, we have described a novel, quantitative, noninvasive high-resolution imaging approach to follow the recruitment of antigen-specific CD8+ T cells to target tumors. For the first time, we have been able to determine immune-specific cellular recruitment in vivo in live animals via MRI. We have determined that these cells are recruited in a heterogeneous manner, both spatially and temporally, and that multiple administrations could provide a more effective therapy for solid tumors than a single dose. Because the magnetic iron oxide core of CLIO-HD is widely used in MRI contrast agents to induce T2 shortening, at greater doses than we have used in the present study (52), these data indicate the potential to image similar cell behaviors in patients in a noninvasive manner and would allow evaluation of existing treatment protocols or modified protocols that might ultimately provide a more favorable outcome for the patient.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org).
Supported by NIH Grants AI86782-02 (to R. W., M. F. K., and J. R. A.), CA96978-01 (to J. R. A.), HL36028 (to A. H. L.), GM64931 (to V. L.), and a grant from the Charles A. Dana Foundation (to J. R. A.). M. F. K. was supported by a fellowship from the Deutsche Forschungsgemeinschaft (DFG).
The abbreviations used are: CTLs, cytotoxic T lymphocytes; MRI, magnetic resonance imaging; CLIO-HD, highly derivatized cross-linked iron oxide nanoparticles; OT-I, ovalbumin-specific, MHC class I-restricted T-cell receptor transgenic; DPBS, Dulbecco’s PBS; IL, interleukin; OVA, ovalbumin; ROI, region of interest; MR, magnetic resonance; FACS, fluorescence-activated cell sorter.
|.||Accumulation cells/mm2 .||Rolling fraction (%) .||Transmigrated cells (%) .||Initial rate of attachment cells/min/mm2 .|
|CLIO-HD labeled||217 ± 26||1.6 ± 1.5||6.0 ± 0.6||47 ± 9|
|Unlabeled||217 ± 46||1.7 ± 1.2||5.2 ± 0.9||60 ± 18|
|.||Accumulation cells/mm2 .||Rolling fraction (%) .||Transmigrated cells (%) .||Initial rate of attachment cells/min/mm2 .|
|CLIO-HD labeled||217 ± 26||1.6 ± 1.5||6.0 ± 0.6||47 ± 9|
|Unlabeled||217 ± 46||1.7 ± 1.2||5.2 ± 0.9||60 ± 18|
We thank Terry O’Loughlin for mathematical modeling of the CLIO particle, Anna Moore for technical advice, Nikolay Sergeyev for CLIO-HD preparation, and William R. Heath and Francis Carbone (Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia) for OT-I mice. We also acknowledge Michael Delfs and Nir Grabie for helpful discussions, Deborah Burstein, Jeeva Munasinghe, and Alik Petrovsky for assistance with MR imaging, and G. Stavrakis for advice with immunohistochemistry.