Clinical Magnetic Resonance Imaging of Pancreatic Islet Grafts After Iron Nanoparticle Labeling


* Corresponding author: Christian Toso,


There is a crucial need for noninvasive assessment tools after cell transplantation. This study investigates whether a magnetic resonance imaging (MRI) strategy could be clinically applied to islet transplantation. The purest fractions of seven human islet preparations were labeled with superparamagnetic iron oxide particles (SPIO, 280 μg/mL) and transplanted into four patients with type 1 diabetes. MRI studies (T2*) were performed prior to and at various time points after transplantation. Viability and in vitro and in vivo functions of labeled islets were similar to those of control islets. All patients could stop insulin after transplantation. The first patient had diffuse hypointense images on her baseline liver MRI, typical for spontaneous high iron content, and transplant-related modifications could not be observed. The other three patients had normal intensity on pretransplant images, and iron-loaded islets could be identified after transplantation as hypointense spots within the liver. In one of them, i.v. iron therapy prevented subsequent visualization of the spots because of diffuse hypointense liver background. Altogether, this study demonstrates the feasibility and safety of MRI-based islet graft monitoring in clinical practice. Iron overload (spontaneous or induced) represents the major obstacle to the technique.


Islets of Langerhans transplantation has come to the forefront as a promising approach in the quest for a cure for type 1 diabetes. Unfortunately, outcomes of 80% insulin independence at 1 year have not been sustained (1–3), and the latest updates report insulin independence rates of approximately 10–15% at 5 years, although graft function (C-peptide positivity) was retained in a vast majority of patients (4).

Islet loss occurs in a two-step fashion. Early loss is linked to the low rate of engraftment. It is thought to be the result of damage during the isolation procedure or in the graft microenvironment within the liver, secondary to ischemia-reperfusion-like injury and to nonspecific inflammatory phenomena. Subsequent losses are usually more progressive, and are thought to involve allogeneic rejection, recurrence of autoimmunity, islet toxicity of the immunosuppressive drugs or ‘exhaustion’ of the islet graft. Many of these events are likely to impact similarly on other types of cell transplant, including bone marrow, hepatocytes, neural cells and various types of stem cells, with wide areas of application.

There is undoubtedly a lot of room for improvement in the short- and long-term survival rates of all types of cell transplants, but this will only be achieved when mechanisms of destruction have been fully understood and characterized. In this regard, there is a blatant lack of monitoring tools that are able to detect graft damage or loss in a timely manner.

An ideal radiological monitoring modality should be noninvasive, should not be harmful to the graft, should enable a repeated and accurate assessment and should detect graft damage early enough to allow treatment. In the situation of islet transplantation, in order to be useful, a new tool should detect islet damage prior to a drop in serum C-peptide levels, an increase in fasting blood glucose levels or the need to resume exogenous insulin intake.

Two strategies can be theoretically applied, namely, in situ and ex vivo labeling. In situ labeling by posttransplant injection of a labeled tracer highly specific for islet cells would be the preferred method, since it could be indefinitely repeated, and would not be confronted to half-life of the labeling agent issues. Unfortunately, this strategy has been hampered by the lack of β-cell specificity of candidate molecules (5). For this reason, we have chosen an ex vivo labeling approach in which cells are labeled prior to transplantation.

The two main clinically relevant technologies for cell transplant monitoring include positron-emission tomography (PET) and MRI. The use of PET with pretransplant cell labeling has thus far been limited by the short half-life of tracers, which prevents accurate monitoring beyond the first hours posttransplant (6,7). In contrast, animal studies have demonstrated that cell grafts labeled with superparamagnetic iron oxide (SPIO) particles could be monitored in vivo by MRI. The signal remained stable in syngeneic models, while it disappeared at the time of rejection in allogeneic models (8–13).

The aim of the present study is to investigate whether MRI monitoring of SPIO-labeled islets could be applied clinically, using the well-established islet transplantation procedure as a model.


This is a pilot study investigating the feasibility and safety of MRI monitoring of iron-labeled islet grafts. It was approved by the Ethical Committee for Clinical Research at the University of Geneva Hospitals, and the off-label use of ferucarbotran (Resovist, Schering, Baar, Switzerland) was approved by the Swiss Federal Office of Public Health. All patients on the waiting list for islet transplantation at the University of Geneva Hospitals could potentially be included. Exclusion criteria included pregnancy, hypersensitivity to components of Resovist and severe claustrophobia, and patients with pacemakers, implanted defibrillators, vascular clips (when implanted for less than 2 weeks) or any other implant or device not allowed for MRI.

Islet culture, labeling and quality assessment

Islet labeling was performed by pooling approximately 20 000 islet equivalents (IEQ) per flask into 2–15 flasks (175 cm2, content: 650 mL/flask, nonadherent; Sarstedt, Svelen, Switzerland) and culturing them in a serum-free Connaught's Medical Research Labs (CMRL) 1066-based medium (30 ml/flask), supplemented with 10 μL/mL (280 μg iron/mL) SPIO (ferucarbotran; Resovist, Schering, Baar, Switzerland). Other components of the medium included glutamax (2 mmol/L, final concentration), 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (25 mmol/L), sodium pyruvate (5 mmol/L), linoleic acid (5.35 mg/L), sodium hydroxide (2.5 mmol/L), vitamin E (0.01 mmol/L), nicotinamide (10 mmol/L), penicillin (112 kU/L), streptomycin (112 mg/L), zinc sulfate (4.8 mg/L), insulin (6.3 mg/L), transferrin (6.3 mg/L), selenium (6.3 μg/L) and human albumin (0.6%). Culture temperature was 37°C for the first 24 h, and 24°C thereafter (when applicable). Culture (labeling) was performed at least overnight and up to 48 h (Table 1). Only the purest islets were selected for labeling. Labeled and nonlabeled islets were pooled for transplant.

Table 1.  Islet transplant characteristics
 Labeled isletsTotal islets
TransplantNb (IEQ)Purity (%)Viability (%)Duration of labeling* (h)Nb (IEQ)Nb/kg (IEQ/kg)Purity (%)
  1. *Corresponding to the culture time from isolation to transplant

Patient 11 284 999543978
250 000809048294 500555635
Patient 2137 1008384Overnight288 615665455
250 0008090Overnight354 333656172
3 330 111594833
Patient 31374 9168785Overnight580 622984144
2100 0008790Overnight316 958537260
Patient 41287 1668590Overnight468 874586066
2306 2508090Overnight422 000510343
Mean 172 2058388 371 224625954
SD ±144 690 ±3±3  ±100 640 ±1442 ±16 

Viability was assessed by propidium iodide and fluorescein diacetate staining. In vitro function of both labeled and nonlabeled islets was evaluated by glucose-stimulated insulin release in static incubation assays run in triplicate. Two hundred IEQ were preincubated at low glucose concentration (2.2 mmol/L) for 60 min. Basal and stimulated insulin concentrations were measured by enzyme-linked immunosorbent assay (ELISA) (Mercodia, Uppsala, Sweden) in the culture medium after further incubation at low and high glucose concentrations (2.2 and 22.2 mmol/L, for two periods of 90 min each). At the end of static incubation, the insulin content was extracted by incubating the islets with acid-ethanol at 4°C for 60 min. Results were expressed as absolute values and as ratios of basal or stimulated secretion over insulin content. Stimulation indexes (SI) were calculated by dividing each stimulated response by each basal insulin secretion. The mean of the indexes was then reported. This in vitro test was performed with islets from seven different isolations.

In vivo function of both labeled and nonlabeled human islets was also assessed by transplanting 2000 IEQ under the kidney capsule of athymic nu/nu (nude) mice (Janvier, Le Genest, France; 11 animals). This experiment was performed in accordance with a protocol approved by the Institutional Animal Care and Use Committee. Animals were kept in the animal facility of the Faculty of Medicine at the University of Geneva. They had free access to food and water at all times. Diabetes was induced by a single injection of 220 mg/kg of streptozotocin (Sigma, St. Louis, MO). Only diabetic animals with blood glucose levels higher than 20 mmol/L were used as recipients. Time to reverse diabetes (blood glucose <11 mmol/L) was recorded as the endpoint.

Magnetic Resonance Imaging (MRI)

MRI studies were performed prior to and at 5 days, 6 weeks and 6 months after each transplant and whenever a significant metabolic event occurred.

Images were acquired on a 1.5T Achieva MR system R1.5 (Philips Medical Systems, Best, The Netherlands) using the standard abdominal flex coil and a sequence 2D T2*-weighted breath hold gradient echo (fast field echo [FFE], repetition time [TR]= 220 ms, echo time [TE]= 18 ms, acquired voxel size: 1.8 × 2.7 × 5.0 mm, reconstructed voxel size: 1.31 × 1.31 × 5.00 mm, flip angle: 25°, clear).

Analyses were performed blindly by four different investigators, studying all sections acquired over the liver. Iron-labeled islets appeared as dark signal spots on T2*-weighted images. Images were analyzed in a dynamic fashion (on computer) in order to differentiate islet dark spots identified in one image, and vessels appearing tubular in several consecutive images. They were counted manually, and results were provided as an average of the four counts. The coefficients of variation (100*mean/standard deviation [SD]) among the four counts were also calculated for each time point, and the mean of all coefficients was reported.

Liver (L) and paraspinous muscle (M) signal intensities were obtained in operator-defined regions of interest on T2*-weighted images. The L/M ratio was calculated by dividing mean intensities (14).

Statistical analysis

Results were provided as mean ± SD. Continuous variables were compared with the Student's t-test. Reversal of diabetes in nude mice was analyzed by the Kaplan-Meier method, and differences among groups were tested by the log-rank test. All tests were conducted using the standard alpha level of 0.05. Analyses were done using the statistical software (StatSoft, Tulsa, OK).


Patients, islets and transplants

Four patients were recruited (one female and three males, mean age of 52 ± 9 years). Types of transplant were: islet after kidney (IAK) for patient 1, simultaneous islet kidney (SIK) for patient 2 and islet transplant alone (ITA) for patients 3 and 4. They received a total of nine islet infusions (Table 1).

Islets from seven isolations were labeled with SPIO (one infusion was performed before this study started, patient 2 demonstrated high, spontaneous iron load and the last infusion was not labeled). Only the purest portions (83 ± 3% purity) of the preparations were selected for labeling (39 ± 26% of total number of islets).

Posttransplant immunosuppression consisted of anti-CD25 antibody induction (daclizumab; Zenapax, Roche, Basel, Switzerland) for ITA and IAK recipients, and antithymocyte globulins (Thymoglobulin, Genzyme, Baar, Switzerland) for the SIK recipient. Maintenance immunosuppression consisted of sirolimus (Rapamune, Wyeth, Zug, Switzerland) and low-dose tacrolimus (Prograf, Astellas Pharma, Villars-sur-Glâne, Switzerland).

Labeled and nonlabeled islet assessment

After labeling, islet viability was 88 ± 4%. Islets demonstrated similar function as nonlabeled controls. In vitro basal insulin secretions were 108 ± 106 and 282 ± 324 mU/L (p = 0.2) for labeled and nonlabeled islets, respectively, stimulated secretions were 266 ± 224 and 395 ± 454 mU/L (p = 0.5) and stimulation indexes 3.2 ± 2.2 and 2 ± 1.3 (p = 0.2), respectively. When normalizing results for the islet insulin content, basal ratios were 0.04 ± 0.03 and 0.16 ± 0.20 (p = 0.2) for labeled and nonlabeled islets, respectively, and stimulated ratios were 0.13 ± 0.14 and 0.2 ± 0.2 (p = 0.4), respectively. Two thousand human IEQ reversed diabetes in 8 ± 4 and 6 ± 2 days in nude mice of the labeled and control groups (Log-rank test, p = 0.93).

Transplant outcome

All patients reached insulin independence (Figure 1), with normalized hemoglobin A1c (HbA1c) under 6.5%. Patient 1 had to restart insulin injections 15.5 months after the first transplant following a severe urinary sepsis. Some degree of autoimmunity may also have occurred, as his anti-glutamic acid decarboxylase (GAD) antibody titre increased from 5000 U/L pretransplant to 29 000 U/L (normal range <9.5 U/L). He was, at the time of writing, stable on 9 U/day (32 U/day prior to transplant).

Figure 1.

Clinical outcome. Clinical outcome after transplantation of iron-labeled (black bar) and nonlabeled (striped bar) islets. Grey histograms show periods of partial islet graft function (insulin dependence with detectable C-peptide), and white histograms show periods of full graft function (insulin independence). All patients underwent a baseline MRI prior to transplant; posttransplant MRI studies are shown by black triangles. Numbers located over the triangles refer to the average of black spots observed within the liver by four different investigators on each specific MRI; ‘-’refers to the fact that no spot could be observed due to diffuse hypointense background.

Two patients experienced significant anemia, which required treatment with erythropoietin (patient 1, 2) and/or oral iron supplementation (patient 1). Patient 1 further required i.v. infusions of iron sucrose (Venofer, Vifor, St Gallen, Switzerland) from 17 to 19 months, and from 24 to 26 months after initial transplant (Figure 1).

Pretransplant ferritin levels were within normal ranges (from 36 μg/L to 195 μg/L), except in patient 2, who had a spontaneously high level (326 μg/L; normal ranges being 11–137 μg/L in females and 26–417 μg/L in males). Subsequent ferritin levels remained overall stable, except in patient 1, who maintained high levels following i.v. iron supplementation (400, 477, 333 and 368 μg/L at 11, 13, 16 and 30 months after his first islet transplant).

MRI islet monitoring

Prior to transplant, all livers appeared normointense on T2*-weighted images with L/M ratios between 1.3 and 1.5, except for patient 2, who had a diffusely hypointense liver, with L/M ratios between 0.13 and 0.26 during the whole follow-up (Figure 2).

Figure 2.

MRI images. MRI images of patients 1–4 at various time points prior to and after transplant. The arrows indicate ‘black spots’ induced by iron-labeled islets.

Because of the persisting spontaneous liver hypointensity, islet-induced hypointense dots could not be identified in further images performed in patient 2. In contrast, all other three patients demonstrated dark signal voids in posttransplant liver images. They were visible 5 days after transplant already despite a diffusely decreased signal of the overall liver parenchyma compared to baseline (L/M, 5 days: 1 ± 0.15 vs. pretransplant: 1.4 ± 0.12, p = 0.01), reflecting the injection of extracellular iron together with the labeled islets or the poor engraftment of some islets. Six weeks after transplantation, the diffuse, decreased signal of the liver parenchyma partially resolved (L/M: 1.1 ± 0.1), and spots could still be identified in subsequent images. These spots were located throughout the liver in two patients (patients 1 and 4), but only in the left lobe in patient 3 (Figure 2).

The number of spots identified within the liver ranged from 3 to 138 in the first imaging after transplant (Figure 1). There was no correlation between the number of injected islets and the number of spots, but the only recipient (patient 1) of islets labeled during 48 h demonstrated higher counts. While the sharpness of the spots tended to decrease over time, the number of dark signal voids remained stable in all patients. However, they could not be identified starting from the third posttransplant MRI of patient 1 because of the overall decreased intensity of the liver parenchyma (L/M between 0.09 and 0.51; Figure 2). This correlated to the i.v. iron therapy and to an increased ferritin level, as described above. While the patient experienced a decrease in graft function, during the same time period, no clear correlation could be established between MRI signal and islet function.

Of note, counting results demonstrated a high interobserver variability, with a mean coefficient variation of 68 ± 34%.


The present study demonstrates the feasibility and safety of MRI-based monitoring after islet transplant. The technique could be extended to all areas of cell transplantation, including hepatocyte, cardiac, neural and stem cell transplants.

Along with previous animal results (8–11), our results suggest that iron labeling is harmless to human islets in in vitro static incubation and viability tests or in vivo tests after transplantation in diabetic animals. As a further confirmation, all transplanted patients did achieve insulin independence.

The labeling agent (ferucarbotran) is broadly used for clinical MR liver imaging. Labeling is the result of endocytosis. This process is not specific for β-cells, and SPIO could be found in all islet cell types (11). As a consequence, we elected to label the purest transplanted islets only. All had 80% or over purity and represented about 40% of islet grafts. This allowed a better specificity of the labeling.

Importantly, the SPIO have been demonstrated to remain stable within cells, both in culture and after transplantation (10,15). This represents one of the main advantages of the described MRI-based technique, and is in opposition to PET tracers, which have a low cell retention rate (5,6). In the present study, islet-induced spots could still be identified 6 months after transplantation.

The main limitation of the described technique is linked to the specific intraportal location of the islets, and is due to the possible presence of iron overload and background within the liver. This can be either spontaneous (patient 2) or following i.v. iron supplementation therapy (patient 1). Patients with pretransplant MRI L/M ratios of <1, suggesting high intrahepatic iron content (14), should be excluded. High ferritin levels are also suggestive of similar patterns.

While the described technique represents the first report of long-term clinical islet graft imaging, some questions remain only partly answered. The monitoring modality should ideally be able to correlate with the number of transplanted cells. While this was demonstrated in animal models (11,12), in the present study, no correlation could be found between the number of transplanted islets and the number of spots within the liver. Also, the number of spots was low in regard to the number of transplanted islets. This likely reflects the fact that we only observed islets when several of them had engrafted at the same location, grouping iron particles together.

Counts were performed manually, and demonstrated a high variability from one investigator to the other (mean coefficient of variation: 68 ± 34%). This clearly emphasizes the need for an accurate and automated counting technique.

A useful monitoring tool should also be able to detect damages early enough to allow for appropriate intervention to salvage the graft. In our series, only one patient experienced a loss of islet function (patient 1), and unfortunately required i.v. iron treatment around the same time period. Despite repeated images, the liver background remained high up to 5 months after treatment. We were, therefore, unable to determine whether the subsequent absence of spots within the liver was only linked to the background or also to the loss of islets. Of note, animal studies have demonstrated that the iron contained in rejected cells is cleared from the liver, and MRI can detect a decrease in the number of spots at the time of rejection (9,11). In addition, when injected i.v., Resovist is cleared from the liver within 10 days (earliest time point tested), with normalized MRI imaging (16,17).

We believe that this study represents an important proof of feasibility of the MRI-based islet graft-imaging concept. It opens the field to new studies aiming at optimizing MRI contrast agents (18), increasing islet labeling efficiency, improving image resolution and quality and developing quantification methods to better correlate signal and islet mass.


The authors thank the fellows and the staff of the Cell Isolation and Transplantation Center (Dr. Mathieu Armanet, Dr. Reto Baertschiger, Solange Charvier, Raymond Mage, David Matthey, Florentina Naville, Nadine Pernin, Corinne Sinigalia and Dr. Anne Wojtusciszyn), the transplant coordinators, the radiology team, the immunology team, the diabetology team, the nephrology team and all Swiss and French pancreas-recovering teams. The study was supported in part by grant 3200B0–113899 from the Swiss National Science Foundation (to TB, CT, JPV, DB), grant R01 AI 74225–01 from NIH/NIDDK (to TB, DB, CT, JPV), a grant from the Research and Development Foundation at Geneva University Hospitals (to CT, TB, PM) and a grant from the ‘Fondation pour la lutte contre le cancer et pour des recherches medico-biologiques’ (to TB, CT, PM). CT is recipient of grant 118593/1 of the Swiss National Science Foundation and of a grant of the Alberta Heritage Foundation for Medical Research; JPV is supported by the grant FNS PP00B-68778 and FS is recipient of grant no. MZO 00023001 from the Ministry of Health of the Czech Republic.