Main

Islet transplantation has received considerable attention for decreasing end-stage complications of type I diabetes mellitus. The next vital steps for improving islet transplantation include developing effective means to prevent rejection, and methods of safe and targeted delivery and non-invasive follow-up of engraftment. It has been confirmed that the 'Edmonton protocol'1 for transplantation of naked islets can be reproduced elsewhere with an insulin-independence rate of 90%. Success rates, however, have been as low as 23% (ref. 2) with a 5-year graft survival of only 10–20%. In another multi-institutional trial, 44% out of 36 individuals met the primary endpoint at 2 years after transplantation3. As the underlying differences are poorly understood (but thought to result in part from cytotoxic immunosuppressive regimens), there is an urgent need for non-invasive monitoring of islet fate after transplantation. In particular, a sensitive means of correlating islet long-term function with the anatomical location and route of transplantation is necessary, as are methods to assess successful engraftment and persistence of cellular integrity.

Islet immunoisolation may reduce or avoid immunosuppressive therapy altogether. Microencapsulation, the method most commonly used for immunoisolation, surrounds individual islets with thin alginate membranes that are permeable to insulin and metabolites but not to native antibodies. Small-scale (pre)clinical trials have produced impressive evidence of the potential of microencapsulated cell therapy when administered intraperitoneally4,5. Transhepatic intraportal delivery of naked islets, however, has remained the prevailing method of islet transplantation in individuals with type I diabetes mellitus. Liver grafting through portal vein access, which allows a higher nutrient and oxygen supply, is thought to be the optimal means for transplanting encapsulated cells6,7. Despite the therapeutic success with intraportal delivery, the accuracy of infusion and extent of engraftment of microencapsulated cells, dictating clinical failure or success, cannot be assessed8. To overcome these limitations, we have developed new magnetocapsules that are trackable by MRI and, simultaneously, capable of cellular immunoprotection.

Results

Magnetoencapsulation

The macroscopic appearance of magnetocapsules (Fig. 1a–c) and microscopic appearance of encapsulated cells (Fig. 1d–h) showed uniformity in size (∼350 μm in diameter). Magnetocapsule preparations were stable for at least 18 months. When prepared with alginate containing 20% vol/vol Feridex, the iron concentration was 80.8 ± 4.9 ng of iron per capsule. Magnetocapsules containing human islets (Fig. 1e) showed a characteristic Feridex-like color. Dextran-specific immunostaining (Fig. 1f,g) showed that Feridex labeling was uniform without particle clustering.

Figure 1: Macroscopic and microscopic appearance of magnetocapsules.
figure 1

(a) Unlabeled capsules (without Feridex) feature a transparent appearance of alginate. (b) Unstained Feridex-containing magnetocapsules show a ferric rust-like color originating from the Feridex iron oxide particles. (c) Prussian Blue (Fe3+-specific) staining of Feridex-containing magnetocapsules. (d–f) A single human islet encapsulated without (d) and with (e,f) Feridex. (d,e) Unstained samples. Again, note the ferric rust color from iron oxides in e and the uniform, smooth incorporation of iron particles without clustering or aggregation. (f) Staining with dextran-specific FITC-conjugated antibody showed the presence of the dextran coat of Feridex particles (green); islets are stained with DAPI for cell nuclei (blue). (g) Single magnetocapsule containing encapsulated βTC-6 cells. Dextran-specific (Feridex-specific) immunostaining, green; DAPI, blue. (h) Newport Green and propidium iodide staining of βTC-6 cells 48 h after magnetoencapsulation showed >95% cell viability. Scale bars, 1 mm (a–c); 150 μm (d–h).

Full size image

In vitro comparisons did not reveal differences in capsule permeability, islet cell viability or insulin secretory response. Both unlabeled microcapsules and magnetocapsules were permeable to lectins of ≤75 kDa, but impermeable to lectins of ≥120 kDa (Supplementary Table 1 online), ensuring blockage of antibodies but allowing diffusion of insulin (∼5 kDa) and nutrients. The viability of magnetoencapsulated βTC-6 cells (Fig. 1e) was 94% and 82% at 3 and 6 weeks, respectively, similar to unlabeled capsules (96% and 81% at 3 and 6 weeks, respectively).

In vitro studies

The viability of magnetocapsule human islets did not differ from that of islets encapsulated without Feridex (Fig. 2a). One day after encapsulation, there was no difference in insulin secretion between magnetocapsules and non-magnetic capsules (Fig. 2b). The glucose responsiveness stimulation index was 3.36 ± 0.21 and 3.50 ± 0.38 for magnetocapsules and unlabeled capsules, respectively (Fig. 2b).

Figure 2: Functionality of magnetocapsule human islets is retained in vitro.
figure 2

(a) Percentage survival of encapsulated human islets with (filled bars) and without (open bars) incorporation of Feridex. No change in survival rate occurred for magnetocapsules as compared with unlabeled capsules (P = NS). (b) Assessment of glucose responsiveness. C-peptide secretion of encapsulated human islets with and without Feridex was measured after 90 min of incubation in a 3.3 mM (filled bars) and 16.7 mM (open bars) glucose solution. C-peptide secretion was not significantly different between magnetocapsules and unlabeled capsules (P = NS). (c) Insulin secretion of encapsulated islets without Feridex (filled bars) and magnetocapsule islets (open bars) in culture were measured (n = 2) over a 15-d period. Except for the initial time point (*P < 0.05), all statistical trials passed the FDA-approved test for bioequivalence (TOST) and showed no significant difference (NS) between the two capsule preparations by Student's t-test.

Full size image

We assessed the insulin secretory response of magnetocapsule human islets over 15 d (Fig. 2c). Using the bioequivalence test (TOST, threshold = 5%, α = 0.05; ref. 9) approved by the US Food and Drug Administration (FDA), we compared insulin secretion from magnetocapsule islets with that from islets encapsulated without Feridex. Except for a small decrease at day 3, magnetocapsule islet insulin secretion was bioequivalent to secretion by islets in unlabeled capsules, ranging from 2 to 2.5 ng of insulin per islet (Fig. 2c). Thus, incorporation of Feridex does not alter capsule 'porosity' and insulin diffusion.

MRI properties of magnetocapsules

With 81 ng of iron per capsule, we obtained a clear MRI depiction of single capsules in agarose phantoms (Fig. 3a–d) and mice (Fig. 3e). Using three-dimensional, inversion-recovery on-resonance (IRON), positive-contrast MRI (ref. 10), we could selectively enhance the capsule surroundings of single magnetocapsules (Fig. 3b). By using conventional MRI sequences, magnetic resonance properties changed substantially after capsule rupture and there was a 72% loss of the hypointense signal (Fig. 3c,d).

Figure 3: MRI appearance of magnetocapsules.
figure 3

(a,b) As magnetocapsules rapidly settle in solution, they were embedded in a 2% agarose phantom at a density of 50 capsules per ml of gel. By using conventional T2*-weighted images (a), individual magnetocapsules can be easily identified as hypointensities. By using the IRON sequence for generating positive contrast (b), individual magnetocapsules appear as a bright signal with depiction of the capsule surface. (c,d) Magnetocapsules before (c) and after (d) rupture using glass bead treatment. After rupture, a considerable loss of hypointensity occurs and the Feridex-induced contrast decreases to a pinpoint double-dipole T2* susceptibility effect. (e) Magnetic resonance image of a mouse after injection of 500 magnetocapsules in the peritoneal cavity. Single capsules are easily identified (arrows).

Full size image

In vivo mouse studies

We transplanted magnetocapsules (n = 6,000, each containing 500 βTC-6 cells; 3 × 106 cells in total) intraperitoneally in streptozotocin-induced diabetic mice (n = 15). Blood glucose levels had returned to normal (∼100 mg/dl) by 1 week of transplantation and remained constant for 8 weeks (Fig. 4a). By contrast, 9 out of 15 non-transplanted mice died, and the surviving mice remained hyperglycemic. Mice transplanted with βTC-6 magnetocapsules, but not untransplanted mice, showed an increasing gain in net weight (Fig. 4b). Mouse insulin levels were significantly (P < 0.005) increased at 4 and 8 weeks in mice with βTC-6 magnetocapsule engraftment, but not in controls (Fig. 4c).

Figure 4: Functionality of magnetocapsule insulinoma cells in vivo in mice.
figure 4

(a,b) Blood glucose (a) and net gain or loss in weight (b) in diabetic mice after transplantation of magnetocapsule βTC-6 mouse insulinoma cells at various time points after streptozotocin injection (filled symbols, n = 15 mice). Non-transplanted mice were included as controls (open symbols, n = 6 at day 60; 9 of 15 mice did not survive 8 weeks). (c) Non-fasting mouse C-peptide levels in the same two groups of mice (magnetocapsule insulinoma, filled bars; controls, open bars). Pre-Tx, pre-transplantation. Asterisks in a–c indicate the day on which the difference in the values between the two groups became significant (Student's t-test, P < 0.05). The data show that magnetocapsule insulinoma cells are fully functional and restore normal glycemia in diabetic mice over at least an 8-week period.

Full size image

In vivo swine studies

Using an MR-compatible catheter, we infused 40,000 magnetocapsules into the portal vein of swine to allow real-time monitoring of correct catheter positioning and initial liver engraftment on a 1.5-T clinical scanner (Supplementary Movie 1 online). The needle could be actively tracked as it traversed the inferior vena cava (IVC) toward the portal vein (Fig. 5a). After precise infusion, magnetocapsules were clearly visualized as hypointensities, representing capsule distribution within the whole liver (Fig. 5b–d). Magnetocapsule distribution was predominantly in the liver periphery with central sparing, correlating to normal portal vein flow patterns. Follow-up MRI at 3 weeks showed no changes in magnetic resonance appearance (Fig. 5d,e) or health complications. Even after a larger dose of 140,000 magnetocapsules, blood bilirubin, alkaline phosphatase, aspartate aminotransferase, alanine aminotransferase and platelet counts were within normal values for 4 weeks after transplantation. Portal pressure showed a mild transient increase of up to 27 mm Hg immediately after injection, but it returned to near-baseline values by 30 min after injection, and then normal pre-injection values persisted for at least 4 weeks. Grafted human magnetocapsule islets secreted insulin (Fig. 5f) with circulating C-peptide values of 0.27–0.38 ng/ml for 3 weeks.

Figure 5: Magnetic resonance–guided transplantation of magnetocapsules and functionality in vivo in swine.
figure 5

(a) Conventional magnetic resonance angiography/venography of the mesenteric venous system was performed with Gd-DTPA before any punctures. White arrow, active needle; black arrow, portal vein. Needle is seen in the IVC in the proper orientation for portocaval puncture. The magnetocapsule injection is shown in real time in Supplementary Movie 1. (b,c) In vivo MRI of magnetocapsules before (b) and 5 min after (c) intraportal infusion of magnetocapsules in a swine. Magnetocapsules can be seen distributed throughout the liver as hypointense signal voids created by the magnetocapsules. (d,e) MRI follow up at 3 weeks shows the persistence of magnetocapsule human islets. (f) Magnetocapsule islets retain functionality in vivo, as assessed by a sustained increase in human C-peptide in plasma.

Full size image

After in vivo MRI, we imaged the liver ex vivo using a 3-T clinical scanner. T2*-weighted MRI showed strong hypointensities and confirmed the in vivo findings (Supplementary Fig. 1 online) that magnetocapsules were lodged in the distal microvasculature of the liver. A three-dimensional reconstruction of the in vivo MRI showed distribution of the magnetocapsules in the distal vasculature in all three dimensions throughout the whole liver (Supplementary Fig. 1 and Supplementary Movie 2 online).

Discussion

We have developed clinical grade magnetocapsules that facilitate simultaneous immunoprotection and magnetic resonance tracking of grafted cells. We have shown the feasibility of hepatic delivery of magnetoencapsulated islets in real time in a clinically relevant animal model by using clinical devices and instruments. In vitro comparison of magnetocapsules and conventional, unlabeled capsules did not reveal differences in capsule permeability, cell viability or insulin secretory response. Magnetocapsule islet cells were found to retain their curative properties, to restore euglycemia in diabetic mice and, for a given dose of 40,000 capsules, to produce sustained C-peptide levels in swine close to levels considered functional in humans (0.3 ng/ml; ref. 3).

Non-human islets have shown promise in non-human primate diabetic models. As porcine insulin differs from human insulin by only one amino acid residue, porcine islet xenotransplantation is being explored11. At present, stringent immunosuppression is the only means to prevent xenogenic graft rejection. Further potential clinical translation may benefit from microencapsulation, providing full immunoprotection, of porcine islets, which represent a relatively inexhaustible source. Despite the potential of microencapsulation therapy, basic issues such as ideal transplantation site, best mode of delivery, and long-term graft survival are largely unresolved. If microcapsules could be imaged non-invasively, these issues could be better addressed. MRI has excellent soft-tissue contrast, high resolution, whole-body imaging capability, and the ability to track magnetically labeled cells in vivo12. Indeed, MRI cell tracking has been introduced into the clinic13 and has proved to be superior to ultrasound and radionuclide imaging in determining the accuracy of delivery and migration to nearby tissues.

The clinically approved Feridex particles are biodegradable and have limited toxic side-effects, although one report has shown that they inhibit stem cell differentiation pathways14. On the one hand, some in vitro glucose stimulation experiments have shown no differences in insulin secretion between directly labeled and unlabeled islets15,16,17. On the other hand, other in vitro glucose stimulation experiments resulted in a significant (50%) reduction in insulin secretion by labeled cells as compared with unlabeled cells18. With direct Feridex labeling, loss of islet detectability may occur when cells dislodge from transplanted islets and escape into the circulation or surrounding liver tissue. It has been reported that indirect islet labeling can occur through particle uptake by macrophages present in islet preparations16, whereas others have shown that islets can be directly labeled and successfully followed by MRI after liver engraftment17. When using Feridex, however, a high variability in labeled cells has been reported, ranging from 10 to 70% efficiency19. We have also labeled human cadaveric islets with Feridex directly, but were unable to obtain a uniform, consistent amount of labeling. When using magnetocapsules, however, the variability and uncertainty associated with direct islet cell labeling are circumvented.

In standard alginate-poly-L-lysine (PLL)-alginate (APA) microcapsules, the positively charged amino group of the lysine molecule of PLL interacts with the negatively charged carboxyl and hydroxyl groups of the uronic acid in the alginate. For direct magnetic cell labeling, PLL is used as a cationic transfection agent by electrostatic complexing with the anionic Feridex20. We therefore exploited the presence of PLL in APA capsules to complex Feridex. Magnetocapsules were found to have an iron content of 80.8 ± 4.9 ng per capsule, which is more than three orders of magnitude higher than the typical iron content of Feridex-labeled cells, which varies from 10 to 20 pg of iron per cell20. This higher content facilitated the magnetic resonance visualization of single capsules (Fig. 3) by a clinical scanner using a conventional pulse sequence without dedicated gradient inserts. At 1.5 T, with direct labeling of islets without capsules, the minimum number of islets detectable by MRI has been reported to be 200 (ref. 21). An alternative solution to imaging single cells may be the use of iron oxide microspheres22. These particles, however, are not clinically approved and it is not known whether they can be taken up by islets through phagocytosis.

Because our magnetocapsules are synthesized by direct mixing of performed Feridex particles, they differ from the iron oxide–containing capsules described previously23. Our magnetocapsules are clinical grade formulations without de novo synthesis of iron oxides from ferric iron. We use clinically applicable formulations of alginate-based magnetocapsules that combine an FDA-approved ferumoxide formulation (Feridex) with a clinical, pharmaceutical grade alginate formulation of Protanal and Keltone. Both Protanal and Keltone are used as coating agents in oral medications and in food products for human consumption. Although being used off label, our clinically applicable approach is expected to facilitate clinical translation. The first clinical study on magnetically labeled cells13 also used the FDA-approved iron oxide formulation Feridex (labeled Endorem in Europe) in an off-label application. With 40,000 magnetocapsules and 80 ng of iron per capsule, the total dose of iron injected was about 3 mg for swine weighing 40–45 kg. Our total Feridex dose is about 14% of the current FDA-approved intravenous dosage given to humans as a liver agent (39 mg of iron for a 70-kg individual). We found no alterations in functional liver test parameters, and intraportal pressures showed only a slight transient increase immediately after transplantation. These findings show that administering a magnetocapsule volume of 6 ml in total is clinically safe, in agreement with a theoretical calculation that a 20-ml total capsule load can be safely injected into one liver lobe24.

MRI of magnetically labeled islets after intraportal infusion has, until now, been shown only in small animal models17,18. As a proof-of-principle, magnetocapsules were disintegrated in vitro and showed a 72% signal reduction in magnetic resonance appearance after capsule rupture (Fig. 3c,d), suggesting that it may be possible to detect capsule rupture in vivo. It is not yet known how long intact capsules persist in individuals. Capsule rupture should precede immunorejection and islet graft dysfunction. MRI used as a non-invasive method to provide information on the status of immunoprotection would be extremely valuable as a decision-making endpoint either for initiating immunosuppressive therapy or to regraft magnetocapsule islets. The excellent surface contrast of magnetocapsules provided by the IRON imaging10 (or potentially alternative methods25) could be particularly useful for this purpose.

Islet transplantation has been very successful in the concentrated direct delivery of islets into target organs, such as the spleen or liver. At present, X-ray fluoroscopy is used to puncture the portal vein blindly, and then a non-directed, large volume of islets is infused directly into the portal vein over 10 min (ref. 26). This technique has a high rate of complications (15–20%), including portal vein hypertension and thrombosis, and has several other limitations including traversal of the hepatic capsule, poor soft-tissue resolution, and the inability to visualize islet delivery and engraftment. In addition, a transhepatic approach requires embolization of the hepatic parenchymal track, which carries the risk of inadvertent injury to the hepatic artery.

As a result of the increased speed of MRI acquisition and available magnetic resonance catheter-tracking technology, MRI is now being explored as an alternative for various vascular27 and cell delivery procedures28. Magnetic resonance–guided islet transplantation could potentially reduce the complication rates encountered when using X-ray fluoroscopy. Using only magnetic resonance guidance and an active intravascular needle, we created a temporary bridge between the IVC and the portal vein, facilitating catheterization of the mesenteric venous circulation, thereby providing several advantages over conventional X-ray fluoroscopy. Our technique accesses the portal vein through a transfemoral IVC approach rather than a transhepatic approach. The retroperitoneum provides a safe space that is capable of providing a seal for vascular punctures, as has long been shown with transcaval aortography and transcaval placement of catheters. Accessing the portal vein from a transcaval approach under MRI also allows easy navigation to either the right or the left portal vein, and facilitates high-resolution liver MRI with intravascular magnetic resonance guide wires and targeted delivery to the liver segment of choice.

In conclusion, we have developed methods for the immunoisolation, magnetic resonance tracking and magnetic resonance–guided targeted intraportal delivery of human cadaveric islets in swine, an animal whose larger vasculature closely resembles that of humans. Exploiting magnetic resonance–guided delivery and imaging of magnetocapsules may overcome many of the hurdles to characterizing and increasing islet transplantation efficacy in human clinical trials. The use of magnetocapsules may also lead to applications in other encapsulated cell therapies.

Methods

Cell culture.

Fresh human cadaveric islets were provided by the National Islet Cell Resource Center. Average purity and viability were 90% and 85%, respectively. For microencapsulated cells, we cultured groups of 100 microcapsules each containing approximately one islet in multiwell plates. We grew mouse βTC-6 insulinoma cells (ATCC) in medium containing 5.5 mM glucose.

Magnetoencapsulation.

Magnetocapsule synthesis is based on a one-step modification (that is, Feridex addition) of the Lim-Sun method29. Our modification uses an electrostatic (van de Graaff) droplet generator, which produces smaller, stronger and more uniform capsules as compared with the older air-jet technique. Before encapsulation, human cadaveric islets were passed through a 20-g needle. We suspended cells, adjusted to 400 islet equivalents per ml or 1.5 × 107 cells/ml (βTC-6), in 2% (wt/vol) ultrapurified sodium Protanal-HF alginate (FMC Biopolymers) and 20% (vol/vol) Feridex (Berlex Laboratories). We passed this solution through a needle at 200 μl/min using a nanoinjector pump. We collected droplets, representing islet cells surrounded by the first layer of alginate, in a Petri dish containing 100 mM CaCl2 in 10 mM HEPES, and washed them three times. We suspended gelled droplets in 0.05% poly-L-lysine (22–24 kDa; Sigma) for 5 min to crosslink alginate and Feridex. We washed and resuspended droplets in 0.15% Keltone HVCR alginate (Monsanto) for 5 min, and then washed them again. For capsule rupture, we manually agitated magnetocapsules in a 50-ml conical tube filled with 1-mm glass beads.

In vitro characterization of magnetocapsules.

We assessed the presence of Feridex in magnetocapsules with Prussian Blue staining and a spectrophotometric, Ferrozin-based iron assay of acid-digested samples12. We carried out immunostaining with a dextran-specific antibody (Stemcell Technologies) to visualize dextran-coated Feridex particles within magnetocapsules30.

After magnetoencapsulation, cell viability was determined by a microfluorometric assay. We incubated encaspulated cells with 10 mM Newport Green (Sigma) for 30 min and 5 mM propidium iodide (Sigma) for 10 min. We randomly selected seven representative microcapsules from three independent preparations each (21 in total). For capsule permeability measurements, we incubated magnetocapsules with one of four fluorescently labeled lectins of varying molecular weight (Supplementary Methods online).

We used a static incubation assay to assess the insulin secretion response (Supplementary Methods). For assessment of magnetic resonance contrast, we suspended magnetocapsules in 2% agarose at a density of 50 capsules per ml of gel, and performed phantom imaging at 3 T (Supplementary Methods).

Mouse studies.

Mouse studies were approved by our institutional animal care and use committee. For the induction of diabetes, we gave 30 C57/BL mice (Charles River) streptozotocin intravenously at 185 mg per kg (body weight). Mice were considered diabetic if they had three consecutive, non-fasting blood glucose levels of ≥20 mM, as measured by a glucometer (Lifescan, Johnson and Johnson). We transplanted 15 mice with magnetocapsule βTC-6 cells into the peritoneal cavity; the other 15 mice received empty magnetocapsules (no cells). Under isoflurane anesthesia, mice were given a single intraperitoneal transplant of 6,000 empty magnetocapsules or 6,000 magnetocapsules containing 500 cells each (3 × 106 cells in total). Every 2–3 d, we measured body weight and took blood samples for blood glucose measurements. To assess magnetic resonance detectability of magnetocapsules in mice, we transplanted 500 magnetocapsules intraperitoneally. Immediately after injection, we performed MRI at 9.4 T (Supplementary Methods).

Swine studies.

We used ten healthy swine (40–45 kg). Using ultrasound guidance, we achieved percutaneous access into the right femoral vein with an 11F sheath. We transferred swine to the magnetic resonance laboratory and advanced a sheath with a magnetic resonance–visible nitinol marker into the IVC. We performed an intravascular puncture of the portal vein using a custom-built, magnetic resonance–trackable needle28. We made an access puncture from the IVC to the portal vein below the splenic vein using real-time magnetic resonance guidance. We advanced a 0.038 nitinol guide wire (Nitrex) into the portal vein, and exchanged the puncture needle for an 8F catheter with a nitinol marker on the distal tip to allow for magnetic resonance visualization. We advanced the 8F catheter under magnetic resonance fluoroscopy into the portal vein for infusion of 40,000 magnetocapsules. We performed MRI immediately and at 3 weeks after magnetocapsule transplantation. We gave two swine a larger dose of 140,000 magnetocapsules in a volume of 30 ml saline (with a packed capsule volume of 6 ml), and obtained liver function (blood) tests and portal pressure measurements (pressure transducer) over 4 weeks (Supplementary Methods). In one swine, we injected 40,000 human magnetocapsule islets and drew blood before and at 1, 2 and 3 weeks after transplantation. We measured specific human C-peptide levels in plasma by ELISA (Alpco Diagnostics) using replicate samples.

Ex vivo imaging and histological correlation.

After humanely putting the swine to death, we collected the liver, fixed it in 4% paraformaldehyde, and suspended it in a styrofoam box filled with 3% (wt/vol) gelatin. We performed MRI at 3 T (Supplementary Methods). We sliced the liver into 1-cm transverse sections and stained them with Prussian Blue.

Statistical analysis.

We used Student's t-test with a significance level of P < 0.05. We also analyzed data by the bioequivalence test, with a two one-sided t-test approach (TOST)9. We performed all analyses using the software R.

Note: Supplementary information is available on the Nature Medicine website.