Abstract
Ex vivo labeling of islets with superparamagnetic iron oxide (SPIO) nanoparticles allows posttransplant MRI imaging of the graft. In the present study, we compare two clinical grade SPIOs (ferucarbotran and ferumoxide) in terms of toxicity, islet cellular uptake, and MRI imaging. Human islets (80–90% purity) were incubated for 24 h with various concentrations of SPIOs (14–280 μg/ml of iron). Static incubations were performed, comparing insulin response to basal (2.8 mM) or high glucose stimulation (16.7 mM), with or without cAMP stimulation. Insulin and Perl's (assessment of iron content) staining were performed. Electronic microscopy analysis was performed. Labeled islets were used for in vitro or in vivo imaging in MRI 1.5T. Liver section after organ removal was performed in the same plane as MRI imaging to get a correlation between histology and radiology. Postlabeling islet viability (80 ± 10%) and function (in vitro static incubation and in vivo engraftment of human islets in nude mice) were similar in both groups. Iron uptake assessed by electron microscopy showed iron inclusions within the islets with ferucarbotran, but not with ferumoxide. MRI imaging (1.5T) of phantoms and of human islets transplanted in rats, demonstrated a strong signal with ferucarbotran, but only a weak signal with ferumoxide. Signal persisted for >8 weeks in the absence of rejection. An excellent correlation was observed between radiologic images and histology. The hepatic clearance of intraportally injected ferucarbotran was faster than that of ferumoxide, generating less background. A rapid signal decrease was observed in rejecting xenogeneic islets. According to the present data, ferucarbotran is the most appropriate of available clinical grade SPIOs for human islet imaging.
Introduction
Unlike for other types of organ transplantation, tests allowing the detection of early rejection after islet transplantation are lacking (9,19). Alteration of glycemic control occurs too late in the graft destruction process to enable graft salvage and islet graft biopsy (i.e., liver biopsy) is not sensitive enough, due to sampling issues (3,28). New strategies are currently under development, including immune monitoring, molecular monitoring (2), and islet graft imaging. Magnetic resonance imaging (MRI) after ex vivo islet labeling currently appears to be one of the most promising strategies (11,12,23,29–31).
The aim of the present study is to compare the use of the two commercially available, clinical grade iron oxide nanoparticles, ferucarbotran (Resovist®) and ferumoxide (Feridex®), in terms of iron uptake, toxicity, insulin response, and MRI imaging ability, both in vitro and in vivo.
Materials and Methods
Animal Selection
Male Lewis rats (Janvier, Le Genest, France; 300–500 g) and athymic nu/nu (nude) mice (Janvier, Le Genest, France) were used. All experiments were performed in accordance with a protocol reviewed and approved by the institutional animal care and use committee and by the state veterinary authorities.
Rat Islet Isolation and Culture
Rat islets were isolated by collagenase digestion of the pancreas, followed by Ficoll purification as described previously (25). Rat islets were cultured in Dulbecco's modified Eagle medium (DMEM, Gibco, Basel, Switzerland) (11 mM glucose, 1 mM sodium pyruvate, 10% fetal calf serum, 100 U/ml penicillin, 100 mg/ml streptomycin).
Human Islet Isolation and Culture
Human islet isolation was performed as previously described according to a modified Ricordi method (8,22). Briefly, the pancreatic duct was catheterized and the pancreas was distended by infusion of a cold collagenase solution (Collagenase NB1, Serva, Heidelberg, Germany) (6,8). After digestion (37°C), purification of the cell suspension was performed in a continuous Biocoll gradient (Biochrom, Berlin, Germany) using a refrigerated cell separator (COBE 2991, Cobe, Lakewood, CO, USA). Islet preparations that could not be utilized for clinical islet transplantation were used in these experiments, under a protocol reviewed and approved by the institutional ethical committee of Geneva University Hospitals. Sixteen islets preparations were used in the study.
Human islets were cultured at 37°C, in 5% CO2 atmosphere, in Connaught's Medical Research Labs (CMRL) 1066-based medium, supplemented with 10% FCS, 25 mmol/L HEPES, penicillin (112 kU/L), and streptomycin (112 mg/L).
Islet Labeling with Clinical Grade Nanoparticles
Two superparamagnetic iron oxide nanoparticles (SPIOs) were studied: ferucarbotran (Resovist®, Schering, Baar, Switzerland) and ferumoxide (Endorem® or Ferridex®, Guerbet, Paris, France). Both are widely used in clinical magnetic resonance imaging as liver imaging contrast agents.
SPIOs contain superparamagnetic iron oxide crystals [maghemite (Fe2O3) and magnetite (Fe3O4)] coated with a dextran shell that links several crystalline molecules together. Ferumoxide has an incomplete nonionic dextran-10 coating and a particle size of 120–180 nm, while the coating agent of ferucarbotran is ionic carboxydextran, with a smaller particle size of 60 nm. A comparison of physicochemical properties of both agents is provided in Table 1 (10).
Iron labeling was performed by incubating islets in CMRL 1066-based medium, as described above, with SPIOs at iron concentrations increasing from 0 to 280 μg iron/ml, as specifically stated. Incubation was carried out at 37°C for 24 h, in 5% CO2 atmosphere. Duration of incubation was selected in accordance with previous publications, showing a saturation of intracellular iron concentration after this incubation time (11). After incubation, islets were washed three times in culture medium and resuspended in 1.5 ml aliquots for transplantation.
Islets selected for labeling had a purity of 80% or more, in order to limit labeling of nonendocrine tissue.
Islet Viability and Function Assessments
Viability was assessed by incubating the islets at room temperature (RT) in 0.2% Dead Red solution (Molecular Probes, Basel, Switzerland) for 30 min. With this assay, dead cells are stained green on UV fluorescence.
In vitro islet function was assessed by static incubation assays run in triplicate. Briefly, 100 islet equivalents (IEQ) were washed with Krebs-Ringer-HEPES buffer. After 1 h of preincubation, islets were incubated for 1 h at low glucose (2.8 mmol/L; basal condition) followed by 1 h at high glucose (16.7 mmol/L; stimulated condition), with or without the addition of 0.1 mmol/L IBMX, 5 μmol/L forskolin, and 100 nM PMA (Sigma, Buchs, Switzerland). Cellular insulin was extracted in acidic ethanol after overnight incubation at 4°C. Samples were analyzed by the Ultrasensitive Insulin ELISA kit (Mercodia, Uppsala, Sweden). Stimulation index (SI) was calculated as the ratio of stimulated-to-basal insulin. All values were normalized according to total insulin content. This test was performed with islets from at least three different isolations.
In vivo islet function was assessed by transplanting 2,000 IEQ human islets under the left kidney capsule of nude mice. Time to reverse streptozotocin-induced diabetes (blood glucose <16 mmol/L) was recorded and compared between control (n = 8), ferumoxide (n = 5), and ferucarbotran (n = 9) labeled islets.
Human and Rat Islet Transplantation in Rats
Labeled islets were used without delay at the end of the incubation period, in order to be in the same condition as in clinical islet transplantation.
Islet transplantation was performed into the portal vein under general anesthesia (Isoflurane®, Abott AG, Baar, Switzerland) after laparotomy. Animals (n = 18) were transplanted with 5,000–10,000 IEQ/kg body weight or with a predetermined number of islets (500, 1,000, 2,000, or 3m000 EIQ), as indicated. Hemostasis was performed using a ready-to-use surgical fibrinogen and thrombin sponge (TachoSil®, Nycomed, Switzerland). Control animals (n = 8) were transplanted with nonlabeled islets (n = 5) or infused with transplantation medium (n = 3).
Ferucarbotran and Ferumoxide Liver Clearance Study
The basal clearance from the liver in vivo of ferucarbotran or ferumoxide was investigated (n = 6/condition), in order to determine how long extracellular SPIO signal was retained after intraportal injection. Animals received SPIOs intraportally at the same iron concentration (280 μg/ml). Sequential MRI studies were performed up to 20 weeks after injection in order to determine the clearance time of each contrast agent in vivo.
Ex Vivo and In Vivo of SPIO-Labeled Islets Imaging
Islet phantoms were made by embedding 100–200 iron-labeled islets in 2.5% gelatin (Gibco, Basel, Switzerland) in 24-well culture plates. MRI of islet phantoms was performed in order to optimize MRI acquisition sequence protocols.
Images were also acquired after transplantation of human and rat islets in rats. Three groups were studied: control (no islets, n = 8), ferucarbotran-labeled islets (n = 10), and ferumoxide-labeled islets (n = 8). During MR acquisition, animals were anesthetized with Isoflurane®, and a breathing cycle around 40 respirations/minute was maintained. After the last MRI follow-up, animals were sacrificed and the liver processed for histology.
Magnetic Resonance Imaging
All MR sessions were performed on a clinical 1.5 Tesla MRI scanner (Achieva, Release 1.5, Philips Medical System, Best, The Netherlands). The same acquisition parameters were used for in vitro (phantoms) and in vivo imaging.
Anesthetized animals were positioned prone and head first on the 4.7-cm-diameter circular MR coil (Philips Medical System, Best, The Netherlands). MR acquisition was performed during a 1.5-h protocol. The liver was localized on images acquired with a spin echo sequence. The acquisition parameters were as follows: slice thickness = 5 mm, field of view (FOV) = 200 × 200 mm2, echo time (TE)/repetition time (TR)/flip angle (FA) = 22 ms/187 ms/90°, acquisition time (AT) = 7.2 s per slice, and acquired in-plane resolution = 0.6 × 0.9 mm2, reconstructed in-plane resolution = 0.4 × 0.4 mm2 in the coronal and transverse planes.
To visualize iron-labeled islets in the liver and detect iron susceptibility signal voids, a 15 transversal slice T1 weighted fast field echo (T1w FFE) acquisition was performed on the liver volume; MR parameters were: slice thickness = 2 mm, FOV = 150 × 150 mm2, TE/TR/FA = 5.1 ms/500 ms/50°, AT = 5 min 34 s, acquired in-plane resolution = 0.3 × 0.45 mm2, reconstructed in-plane resolution = 0.15 × 0.15 mm2, number of scan averages (NSA) = 2. Two saturation bands with thickness of 60 mm were placed above and below the imaging plane to reduce blood and motion artifacts. Identical geometrical parameters and positioning were conserved for 15 transversal slices T2 weighted fast field echo (T2w FFE) acquisition; for this contrast weighting, altered MR parameters were TE/TR/FA = 15 ms/500 ms/50°. The change in TE modified contrast and sensibility to iron susceptibility effect: larger TE increased dark signal void appearance.
To calculate T2 maps, 16-echoes spin-echo images were obtained with the following parameters: TE/ΔTE/TR/FA = 5.9–94.4 ms/5.9 ms/333 ms/90°, slice thickness = 3 mm, FOV = 75 × 75 mm2, acquired in-plane resolution = 0.5 × 0.65 mm2, reconstructed in-plane resolution = 0.3 × 0.3 mm2, NSA = 4. The 10-slice volume was acquired in 10 min. A single exponential fit of the signal intensity at multiple echo times, in all pixels of a region of interest defined in the liver allowed determination of T2 respectively (4).
To study iron nanoparticle clearance from the liver, the same T1w and T2w MR images were performed 1 day before injection, 6 hours later, and every 2 days during 10 days, then at various time points until the signal reached baseline. T2 MR parameters were also evaluated.
Histological Studies
Islet pellets incubated with various concentrations of iron nanoparticles were fixed in 10% formalin, and embedded in paraffin. Similarly, histological studies were also performed using livers of transplanted rats after MRI assessment. Briefly, the portal vein was perfused with NaCl 0.9%. The liver was further fixed with an intraportal perfusion of 10% formalin, dehydrated in successive ethanol baths of increasing concentration, followed by a 100% xylol bath prior to paraffin embedding. Liver sections were cut according to MRI liver orientation in order to be able to compare MRI images and histology. Sections were stained with H&E (Sigma, Buchs, Switzerland), insulin (guinea pig antibody anti-insulin A564, Dako, Zug, Switzerland), and Perls staining (Sigma, Buchs, Switzerland) for iron uptake evaluation.
For electron microscopy, islets were fixed with a 2.5% glutaraldehyde solution in 0.1 M phosphate buffer (pH 7.4), postfixed in 1% osmium tetroxide in the same buffer, dehydrated, and embedded in Epon (13). Thin sections were examined and photographed with a Philips CM10 electron microscope. The different types of islet cells were identified by the typical appearance of their secretory granules (5).
Statistical Analysis
Results were expressed as mean ± SEM. Continuous variables were compared with the Student t-test, and two-way ANOVA test with Bonferroni's post hoc, and categorical variables with the chi-square test. Statistically significant difference was defined as p < 0.05 (95% confidence limits).
Results
Islet Function and Viability Assessment
Islet viability was similar in all conditions (80 ± 10%, p = ns). Of note, ferucarbotran was found only in viable cells of labeled islets, whereas ferumoxide was preferentially observed in dead cells.
Glucose-induced insulin release was similar after labeling with both ferucarbotran and ferumoxide, at all tested concentrations (p = 0.1, two-way ANOVA) (Fig. 1A). Insulin exocytosis was assessed using a mix of three different secretagogues (IBMX, PMA, and forskolin) (Fig. 1B), demonstrating similar patterns in all groups (p = 0.4, two-way ANOVA), but with a trend towards lower levels of insulin release from ferumoxide-labeled human islets.
Static incubation, measurement of insulin secretion (basal and stimulated), or stimulation index. (A) High glucose stimulation (16.7 mM) or (B) high glucose stimulation and addition of forskolin, PMA, and IBMX (N ≥ 3 ± SEM). There is no significant difference between conditions (two-way ANOVA).
Human islet function was tested in vivo after transplantation in nude mice. Controls and ferumoxide or ferucarbotran iron-labeled islets had similar patterns to reverse diabetes (mean time to diabetes reversal was: 4.3 ± 0.47 days; log-rank, p = 0.59).
Iron Content Assessment
We evaluated islet iron uptake according to two different methods, first by light microscopy after Perls and insulin staining, and then by electron microscopy (Fig. 2). On conventional histology, iron particles were seen both inside islet cells and in the islet stroma. Immunohistochemistry showed colocalization of iron and insulin staining, indicating uptake by β-cells, but iron was also seen in other islet cell types. Iron uptake was very heterogeneous between islets of the same preparation. Overall, iron labeling was more intense with ferucarbotran than with ferumoxide at the same iron concentration.
Iron oxide nanoparticles label the extracellular spaces of human islets and different types of islet cells. (A) After labeling with ferucarbotran (280 μg/ml iron), electron microscopy revealed the presence of the electron dense nanoparticles used for MRI contrast within the extracellular spaces of isolated human islets (asterisk), within large invaginations of the cell membrane (open arrowheads), and within lysosomes (solid arrow heads) of β-cells, which were identified by the presence of typical insulin-containing granules (arrows). (B, C) Similar observations were made in the glucagon-containing α-cells (B), and in islet macrophages (C). (D) No labeling was seen in islet exposed to ferumoxide (280 μg/ml iron). Scale bar: 430 nm (A, D), 180 nm (B, C).
On electron microscopy, iron labeling did not alter islet morphology (Fig. 2). After 280 μg/ml ferucarbotran labeling, electron-dense iron oxide nanoparticles were found in sizeable amounts within the extracellular islet space, at the periphery of the islet. Cells bordering these spaces showed large invaginations of the cell membrane filled with contrast agent that was also observed within intracellular, single-membrane-bound organelles (Fig. 2A). From the ultrastructural appearance, these organelles were identified as lysosomes of β- and α-cells, as well as multivesicular bodies of islet macrophages (Fig. 2A–C). No labeling of the other organelles, including nucleus, mitochondria, and secretory granules, was observed in these different cell types. Also, no extra- or intracellular nanoparticles were observed in islets exposed to concentrations lower than 140 μg iron/ml of either ferucarbotran or ferumoxide (Fig. 2D). Finally, cell iron uptake was low, but stronger with ferucarbotran than ferumoxide (proportion of islet cells labeled: 9.3 ± 3.42% for ferucarbotran and 2.0 ± 0.36% for ferumoxide, p = 0.07).
Liver Clearance of Iron Nanoparticles
To rule out that the measured signal in the liver was in fact due to free iron particles retained inside the liver, we measured the time to complete clearance after intraportal injection of both contrast agents at the highest iron concentration (280 μg/ml).
As shown in Figure 3, the ratio of liver-to-muscle signal of intraportally injected ferucarbotran went completely back to baseline levels within 10 days of injection. In contrast to iron-labeled islets, which display easily identifiable individual spots, the pattern of T1 and T2* effect of free iron is diffuse throughout the liver. Surprisingly, ferumoxide labeling did not show the same clearance pattern as ferucarbotran. A hypointense diffuse signal remained within the liver for more than 90 days in T1 and T2*, clearing started only after 42 days.
Clearance of both nanoparticles is shown. SPIOs were directly injected into the portal vein into the liver at the same iron concentration (280 μg/ml of iron). On the T1 MR images, signal decreased, with a full clearance obtained in the ferucarbotran condition within 10 days. Ferumoxide liver signal started to return towards to normal levels after 42 days. (B) Liver/muscle ratio showing a rapid clearance of the ferucarbotran condition compared to the ferumoxide condition, which is still far from normal after 22 days.
In Vitro and In Vivo Islet Imaging
We first performed in vitro imaging of iron-labeled islets in phantoms. At the same iron concentration, iron-induced signal voids were markedly more conspicuous with ferucarbotran than with ferumoxide labeling (data not shown).
MRI signals of iron-labeled islets after intraportal transplantation exhibited a different pattern with ferucarbotran compared to ferumoxide. After ferucarbotran labeling, hypointense spots were scattered throughout the liver, whereas ferumoxide-labeled islets only generated spots centrally, presumably along the main branches of the portal system (Fig. 4). This was consistently observed in eight experiments. Additionally, in contrast to ferucarbotran, T2* localized hypointense signal completely disappeared within 10 days after ferumoxide labeling.
Typical patterns in the rat liver, 6 h after transplantation with transplantation medium (A), ferucarbotran (B), or ferumoxide (C). Gradient echo images were acquired with two different echo times, TE = 7 ms (first line) and TE = 15 ms (second line). No difference has been found on sham-transplanted livers compared to pretransplant acquisitions. Dark voids linked to iron presence were disseminated into the liver lobes in the ferucarbotran conditions, although they were preferentially concentrated around the central portal vein in the ferumoxide conditions.
The number of hypointense spots increased with the number of transplanted islets. Figure 5 shows T2* images of livers transplanted with 500–3,000 ferucarbotran-labeled human islets. The number of spots increased proportionally to the number of transplanted islets.
Typical images of the rat liver 6 h after injection of islets labeled with ferucarbotran. Number of injected islets was proportional to the signal seen on MRI here from 500 to 3,000 labeled islets (ferucarbotran concentration of 280 μg/ml of iron).
Because of the impossibility to obtain a long-term signal with ferumoxide at 1.5T and of the markedly better labeling ability of ferucarbotran, this agent was selected for long-term studies. Long-term monitoring was performed with ferucarbotran-labeled islets in a syngeneic model. We observed the presence of pancreatic islets in the liver of the transplanted animals for at least 8 weeks. Over this period, few dark spots linked to the presence of iron disappeared whereas most remained at the same position and retained the same intensity. On the other hand, as expected, the signal generated by xenogeneic (human) islets rapidly decreased (Fig. 6).
Eight-week follow-up MR images after injection of ferucarbotran-labeled rat islets (A, syngeneic model) and human islets (B, xenogeneic model). Before injection (D\min\1), local dark voids correspond to hepatic vessels; in both models, 6 h (D0) after injection, dark voids linked to iron presence are visible throughout the liver (blacks arrows). One (W1), 2 (W2), 4 (W4), and 8 (W8) weeks after injection, dark voids persisted in the syngeneic model and progressively disappeared in the xenogeneic model.
Radiohistological Correlation
In order to demonstrate that hypointense spots were indeed generated by iron-labeled human islets, correlation of the radiological images with histology was performed. After removal of animal livers transplanted with ferucarbotran-labeled human islets at different time points (pretransplant, posttransplant, W1, W2, W4, W8), histological exploration was performed over the whole liver. Only a portion of intrahepatic islets presented iron inside the cells or in the islet stroma. Iron was only found in and around viable transplanted islets. No iron was seen anywhere else in the rest of the liver (Fig. 7).
Cross-section of the liver in the same plane as the MRI acquisitions were performed, allowing retrieval of labeled islets in liver section with a good correlation with hypointense spots on MR images. Retrieval of human islets in the liver was possible over 6 weeks after transplantation. Some islets are not labeled, while others show iron uptake in the islet stroma (red arrow) and in the β-cells (green arrow). No iron was found in the liver parenchyma except around islets.
Persistence of iron-labeled human islets within the liver was observed on histology up to 8 weeks after transplantation. A comparison of histological images according to the corresponding orientation in MR showed a good correlation between histology and radiology (Fig. 7). Histology was also performed on livers transplanted with ferumoxide-labeled islets. Islets were seen, but none of them exhibited any iron labeling even immediately after transplantation.
Discussion
The present study reports a method of islet cell labeling with superparamagnetic iron oxide (SPIO) nanoparticles for direct, noninvasive imaging of islet grafts, and demonstrates the superiority of ferucarbotran compared to ferumoxide.
Pretransplant iron labeling of islet grafts faces several issues that must be addressed in order to validate the method. One should first demonstrate the lack of toxicity toward the islets; second, the appropriate iron uptake by the islet cells; third, the rapid clearance of the agent from the environment after cell destruction; and last the correlation between the radiologic images and the reality, as assessed by histology.
In this study, in an effort to adopt a translational approach, we elected to use approved contrast agents that are widely used for clinical MRI, as a source of SPIO nanoparticles to label human islets, and to perform our studies on 1.5T clinical MRI scanners. This is in contrast to previously published studies in which rodent islets were used (26), or images were obtained on MRI scanners with a 4.7T (18) or 9.4T (12) magnet, which are unavailable, and inapplicable, in the clinical setting. Our first step was to compare the two most widely used agents, ferucarbotran and ferumoxide, in their labeling ability and toxicity, and select the most appropriate for in vivo experiments.
We have shown a lack of human islet toxicity of both types of nanoparticles, with viability and in vitro glucose-stimulated insulin secretion of labeled islets essentially identical to those of native islets. In vivo islet transplantation into diabetic nude mice correlates with the in vitro test in terms of graft survival and function (24).
It became obvious from our cell uptake histology studies that ferumoxide lacked the labeling ability observed with ferucarbotran with “passive” iron labeling by incubation at iron concentration below 300 μg/ml. This was corroborated by the much lower signals obtained with ferumoxide both for in vitro phantoms, and in vivo. Of even more concern was the fact that ferumoxide could not be seen integrated within live islet cells, in sharp contrast to ferucarbotran. This may seem to contradict the absence of negative effect of ferumoxide uptake on islet cell viability. A possible explanation is that islet cell uptake of ferumoxide is too low, in terms of proportion of cells labeled, to have an impact on cell viability. On the other hand, ferumoxide was used for the same purposes in a recent study, reporting a good labeling ability of this agent (20). This apparent discrepancy can be attributed to the four to five times higher iron concentrations used in this publication. The iron concentrations we applied were determined according to manufacturers' instructions, in order not to exceed the maximal amount of 2.24 mg iron/kg body weight in our patients (160 mg iron for a 70-kg subject). One hypothesis explaining the difference in iron uptake is the different coating of the particles (Table 1). The presence of dextranase in human cells could lead to a rapid destruction of the ferumoxide coating, while no carboxydextranase was found so far in human cells, which could explain a preservation of ferucarbotran coating. Another explanation could be the difference in the range of the size of the nanoparticles, with a much wider range for ferumoxide particles compared to ferucarbotran (120–180 nm vs. 60 nm hydrodynamic diameter), which could explain the lack of efficiency in islet labeling and a heterogeneous iron uptake. Finally, differences in ionic properties of the nanoparticles coating have been shown to differentially affect cellular uptake (10).
Electron microscopy analysis provided insights on the localization of the nanoparticles, and on the mechanisms of uptake. The images we have obtained clearly show that ferucarbotran particles are taken up nonspecifically by all types of islet cells, including β- and α-cells, and resident macrophages. In accordance with light microscopy studies, no uptake of ferumoxide was observed. Interestingly, iron nanoparticles were seen within large invaginations of the cell membrane and within vesicles, which suggests that they are integrated within islet cells by endocytosis mechanisms. Iron uptake had no deleterious effect on the islet ultrastructure. On the other hand, a large part of the material remained in the extracellular compartment, in the peri-islet stroma. These observations concur with and complement those already reported by the Prague group (1).
In this study, cellular uptake of SPIOs ranged between 2% and 9%, as measured by light and electron microscopy. This suggests that improved techniques of SPIO uptake will have to be developed in order to optimize the technique for further long-term clinical studies. Transfection techniques (using l-lipofectamin and electroporation or poly-l-lysine) were successfully used, but this has not yet been applied to human islets (16,26).
The question then arises whether islets destroyed by rejection would be able to maintain an MRI signal in spite of the absence of remaining live cells, due to a local release of their iron load by damaged islet cells, or, more critically, by local persistence of the iron located around the islets. To address this point, we performed clearance studies by direct intraportal injection of SPIO nanoparticles. After an initial diffuse hypointensity due to iron overload, signal went back to baseline within 10 days for the ferucarbotran conditions, indicating that SPIOs are not retained by the reticulo-endothelial system in the liver. This is in keeping with previously published clearance studies of iron-based MRI contrast agents in the rat model (15). In the ferumoxide condition, clearance time was more than 90 days. Such a large difference was never reported even though it has been demonstrated that the type of coating material significantly influences the rate of iron oxide clearance in rat liver (7).
Histology provided a second line of evidence, showing that, by 6 weeks posttransplant, some iron was still located within and around live engrafted human islets, and in no other place. It should finally be mentioned that iron nanoparticles (17) released by dead islets are too small in size to individually disturb the MRI signal. These observations led us to abandon ferumoxide as a labeling agent for our purpose and to concentrate our efforts on ferucarbotran. A recent study, from the Verona group, on rat islets at 4.7T corroborates our finding and found that the use of ferucarbotran rather than ferumoxide should be recommended (21).
With ferucarbotran-labeled islets, we were able to generate hypointense spots on T2*-weighted images after transplantation of syngeneic and human islets into rats, using a clinical 1.5T MRI scanner. Spots were distributed throughout the liver lobes, in a rather even manner. As expected, resolution did not reach that obtained with higher magnetic fields, but spots were unmistakably detectable, in numbers increasing in proportion with numbers of islets transplanted. Therefore, the method is clinically applicable using routinely available MRI, as was further demonstrated in a recent pilot study done on human subjects at our institution (29). Signal progressively decreased to eventually disappear after xenogeneic islets transplantation, whereas signal persisted long-term after transplantation of “syngeneic” islets.
Correlation demonstration of the MR images and the histology was a challenge. To do this, we elected to perform MR imaging ex vivo on explanted livers. This allowed histological staining on liver sections cut along the same planes as those of the MR acquisition sequences. Spatial correspondence of MRI spots and iron-labeled viable islets on histology unquestionably validates the method.
With this in-depth study of human islet labeling with SPIO nanoparticles with the aim of MR imaging of intrahepatic islet grafts, we have demonstrated (i) that iron-based MR contrast agents are taken up by islet cells, (ii) that this uptake probably occurs by endocytosis mechanisms and is not cell specific, (iii) that incorporated iron nanoparticles are not toxic to the β-cell, and (iv) that labeled islets are able to generate an MR signal and thus imaging of the islet graft, using clinical MRI equipment, is possible. We have also been able to show that the labeling and imaging abilities of commercially available SPIO nanoparticles is far from being equivalent, and have selected the most efficient available agent for this purpose. These findings demonstrate the proof of concept of the approach of ex vivo islet cell labeling for long-term islet graft imaging and monitoring in the clinical setting.
At this point, it is quite clear that islet labeling for the purpose of islet graft imaging is still a work in progress. Rodent studies have shown good imaging ability and correlation between graft rejection and alteration of the MRI signal, but these studies have been performed on non-clinical grade MR scanners, with high magnetic fields (1,11,12,18). The only pilot study performed on human subjects with a 1.5 T MR scanner showed that islet graft could indeed be imaged, but that signal was rather weak and did not last beyond 6 months posttransplantation (29). Some issues must still be tackled in order to progress in this field: iron uptake by islet cells is not optimal, and therefore techniques aimed at improving tracer uptake or novel labeling/MR imaging agents must be investigated. The quality and definition of the signal also has to be improved. This can be done by enhancing islet labeling, but also by improving signal detection. Finally, signal quantification methods must be developed, which is an extremely challenging goal to achieve on a technical standpoint. We are currently developing novel 3D sequences on a 3T MRI scanner to provide better signal resolution and allow signal quantification, while keeping a translational approach using clinical equipment.
Footnotes
Acknowledgments
The authors wish to thank S. Masson and the staff of the Cell Isolation and Transplantation Center for excellent technical assistance. This study was supported by grant #5R01AI74225 from the National Institutes of Health (to T.B.), grants #3200B0-113899 and PPOO33-116901 from the Swiss National Science Foundation (to T.B., D.B., P.M., C.T.), by an unrestricted educational grant from Astellas (to T.B.), and by a grant from the “Insuleman” Foundation (to T.B.). C.T. is supported by a SCORE grant from the Swiss National Science Foundation. Human islets were obtained thanks to grant #31-2008-416 from the Juvenile Diabetes Research Foundation. The Meda team is supported by grants from the Swiss National Science Foundation (310000-122430), the Juvenile Diabetes Foundation International (1-2007-158), and the European Union (BetaImage, FP7, 222980; IMIDIA, IMI, C2008-T7).