Superiority of Small Islets in Human Islet Transplantation

Chemicals were purchased from Fluka (Fluka Chemie, Buchs, Switzerland) and from Invitrogen (Life Technologies/Invitrogen, Basel, Switzerland). Insulin and C-peptide were measured by radioimmunoassays (Insulin-CT, IRMA C-PEP; CIS Bio International, Schering, Baar, Switzerland). The intra- and interassay variations were, respectively, 5.9 and 7.9% for insulin and 4.9 and 5.8% for C-peptide.

Human pancreata were removed from brain-dead donors and stored in cold University of Wisconsin solution. Cold ischemia times ranged from 2 to 10 h. The islet isolation was performed as previously described (18). In brief, the duct was perfused in a controlled manner with 170 ml cold collagenase (Hanks’ balanced salts [Invitrogen], 3 mg/ml collagenase [Liberase HI; Roche], 70 mmol/l HEPES [Invitrogen], and 10 mmol/l CaCl₂ [Fluka]). The islets were separated from the exocrine tissue by enzymatic and mechanical dissociation. The digested pancreatic tissue was top loaded on a continuous Ficoll gradient with a density from 1.077 to 1.100 (Biochrom, Berlin, Germany) and centrifuged in a refrigerated cell separation apheresis system (COBE 2991; Lakewood, CO). The purified islets were stained with dithizone, counted, and measured with an image analysis system (Saisam; Microvision Instruments, Evry, France) to determine islet yield. Islets were cultured in CMRL 1066 medium (5.5 mmol/l glucose) containing 10% FCS, 100 μg/ml streptomycin, 100 units/ml penicillin, 1 mmol/l sodium pyruvate, 2 mmol/l glutamax, and 2 mmol/l HEPES, hereafter referred as CMRL_compl. Static glucose stimulation was performed in triplicate aliquots with 100 islets per sample.

Islets were handpicked according to size into two groups: small islets were defined as islets with a diameter between 50 and 150 μm and large islets with a diameter between 150 and 300 μm. Islets were counted and the diameter of every islet measured by the image analysis system Saisam. Islet volume is always expressed as IEQs. For each sample, the total islet volume (IEQs) (8), islet number, and size distribution were assessed. The isolation index (islet volume as expressed in IEQs per islet number) is an indicator of the average islet size (e.g., an index <1 indicates that the average islet size is <150 μm and vice versa). Histological sections were analyzed with a Zeiss Axiolab microscope, a Zeiss AxioCam HRc camera, and the imaging software AxioVision 3.1 (Carl Zeiss, Feldbach, Switzerland). Area and diameter of islets were measured using the AxioVision 3.1 software.

The technique of perifusion has been described previously (19). In short, two groups of islets with different sizes, as described above, were handpicked. The islets were placed in a chamber of 300 μl, embedded in Sephadex G-10 (Pharmacia, Uppsala, Sweden), and mounted in a perifusion device, preheated to 37°C (Superafusion 1000, model SF-06; Brandel, Gaithersburg, MD). The islets were perfused with Krebs-Ringer-HEPES buffer (131 mmol/l NaCl, 4.8 mmol/l KCl, 1.2 mmol/l CaCl₂*2H₂O, 25 mmol/l HEPES, 1.2 mmol/l KH₂PO₄, and 1.2 mmol/l MgSO₄*7H₂O) containing 0.1% human serum albumin (Sigma) (Krebs-Ringer-HEPES buffer–human serum albumin) and glucose (Fluka Chemie) at a flow rate of 250 μl/min, and samples of 500 μl were collected. During the first 100 min, the islets were perifused with 2.8 mmol/l glucose, from 100 to 160 min with 20 mmol/l glucose, and from 160 to 220 min with 2.8 mmol/l glucose.

Stimulated insulin secretion was determined as integrated insulin secretion over the time period of high glucose. The amount of insulin secreted is calculated as area under the curve per IEQ in order to adjust for differences in islet volume. Area under the curve for insulin was calculated by using the trapezoidal rule. Basal secretion was calculated by multiplying the average insulin secretion during low glucose exposure just before raising the glucose concentration to 20 mmol/l. The insulin secretion stimulation index was calculated by dividing the amount of stimulated insulin by the amount of insulin produced at basal conditions.

Static glucose stimulation was performed in triplicates with 100 islets per sample. Islets were incubated for 30 min in RPMI_compl (RPMI-1640 without glucose [11879; Life Technologies/Invitrogen]), 10% FCS, 100 μg/ml streptomycin, 2 mmol/l glutamax, 100 units/ml penicillin, 10 mmol/l HEPES (Life Technologies/Invitrogen) with 2.8 mmol/l d-(+)-glucose to equilibrate the islets, for 2 h in RPMI_compl 2.8 mmol/l glucose, for 2 h in RPMI_compl 20 mmol/l glucose, and again for 2 h in RPMI_compl 2.8 mmol/l glucose. The amount of insulin secreted at high glucose (20 mmol/l) was divided by the amount at low glucose (2.8 mmol/l), and the mean insulin stimulation index was calculated.

Human islets were cultivated over night in CMRL_compl at standard cell culture conditions (37°C, 5% CO₂, 20% O₂). Islets of all sizes were cultivated in one dish. They were divided in four equal samples, placed in 35-mm diameter dishes, and incubated for 24 or 48 h either at normoxic (20% O₂) standard cell culture conditions or in a hypoxia chamber (Coy Laboratory Products, Grass Lake, MI) at hypoxic (1% O₂), but otherwise standard, cell culture conditions. Real-time oxygen measurements were performed in a hypoxic chamber (Coy Laboratory Products) at 7.6 mmHg (1%) oxygen tension with an Oxylite fluorescence-based, non–oxygen-consuming device (Oxford Optronix). The oxygen sensor has been placed on the bottom of a 35-mm Petri dish filled with 2 mm medium (Life Technologies CMRL). With 100 human islets, a steady-state oxygen tension of 5.9 ± 1.5 mmHg was reached after 12 h of hypoxic exposure. After incubation, islets were washed once in PBS and fixed and stained as described below.

Human islets were fixed in Bouin solution or Shandon Zinc Formula-Fixx (Anatomical Pathology International, Runcorn, Cheshire, U.K.) for 20 min either directly after isolation or after 1 to several days in culture. The fixed islets were embedded in 2% agarose in PBS, dehydrated, and paraffinated in a tissue processor Leica TP1020 (Leica Microsystems, Glattbrugg, Switzerland). Three-micrometer sections were cut on a microtome (Leica RM2255). Sections were deparaffinated in xylol and rehydrated in a descending series of ethanol (99, 95, 90, 80, 70%, H₂O, PBS), permeabilized for 5 min in methanol, and blocked for 30 min at room temperature with 10% normal goat serum in PBS. The sections were incubated with the first antibody (diluted from 1:50 to 1:1,000 in PBS/1% normal goat serum according to the antibody for 2 h at room temperature) and with the second antibody (diluted from 1:50 to 1:100 in PBS/1% normal goat serum, 10 ng/ml DAPI) for 2 h at room temperature. Antibodies used were monoclonal antibody anti-insulin (clone K36AC10; Sigma-Aldrich, St. Louis, MO) used at a 1:1,000 dilution, rabbit anti-glucagon (A0565; Dako, Carpinteria, CA) at a 1:50 dilution, goat anti-rabbit Cy3 at a 1:100 dilution, and goat anti-mouse fluorescein isothiocyanate at a 1:50 dilution (Jackson ImmunoResearch Laboratories, West Grove, PA). Terminal dUTP nick-end labeling (TUNEL) staining was done using the ApopTag Peroxidase In Situ Apoptosis Detection Kit (S7100-KIT; Qbiogene, Basel, Switzerland), and nuclei were counterstained with hemalaun.

Since both apoptosis and necrosis play a substantial role in the early posttransplant setting, hematoxylin and eosin staining was used to quantify cell death (including apoptosis and necrosis). The islet diameters (effective diffusion distance to islet core) were measured using Zeiss Axiovision software; analysis was performed using Adobe Photoshop software. Every cell on a cross-section of an islet was counted. Cells with nuclear hyperchromaticity, condensation, fragmentation, and/or hypereosinophilic and coarsely granular cytoplasm were scored as dead cells. To calculate a TUNEL index (positive cells/100 cells), all positive and negative nuclei were recorded. Counting was done blinded by two independent observers (R.Z. and P.K.). Islet volume was calculated considering islets mathematically as spheroids.

Simultaneous islet-kidney transplantation and steroid-free immunosuppression was performed according to protocol that has been described previously (18). This protocol was approved by the ethics committee of the University Hospital Zurich, and written informed consent of patients was obtained. Since engraftment and complete vascularization of transplanted islets in the portal fields takes 3–6 months (20), we used the maximal stimulated C-peptide value obtained during a mixed-meal tolerance at 3 months posttransplant as the main clinical outcome measure in the multiple regression analysis. Since every islet isolation is composed of islets with different sizes, multiple islet transplantations would make analysis of the effect of islet size on insulin secretion impossible. Therefore, only patients with a single islet transplant and no need for steroid-based rejection therapy during this time period were included (n = 7).

Results are expressed as means ± SE. Data were analyzed with Statistica for Windows software (1997; Statsoft, Tulsa, OK). Groups were compared with the Student’s t test when parameters were normally distributed, with Mann-Whitney U test when parameters were not normally distributed, and a χ² test for categorical variables. Kruskal-Wallis one-way ANOVA on ranks was used for multiple group comparisons versus control subjects for nominal parameters. P < 0.05 was considered significant. Using a linear regression, we analyzed the effect of the islet volume (IEQs), the islet number, and the islet size (isolation index) on the maximally stimulated C-peptide response measured 3 months after simultaneous islet-kidney transplantation.

Islets from 30 human pancreata were used for experiments. Means ± SE of the donor characteristics were age 53.0 ± 2.4 years, BMI 25.6 ± 0.7 kg/m², cold ischemia time 5:52 ± 0:19 h, total islet yield 281,894 ± 42,058 IEQs, islet number 204,700 ± 21,169, isolation index 1.3 ± 0.1, and stimulation index 4.1 ± 0.6. No difference in the donor BMI, age, static stimulation index, and cold ischemia time was seen between transplanted or not transplanted islets and between the islets used for the different groups of experiments. The total islet yield, islet number, and isolation index were significantly higher in islets used for human transplantation compared with nontransplanted islets (IEQs: 528,985 ± 76,229 vs. 188,034 ± 28,253, P < 0.0001; islet number: 316,571 ± 35,849 vs. 170,652 ± 20,927, P = 0.001; and isolation index: 1.8 ± 0.4 vs. 1.1 ± 0.1, P = 0.004).

The distribution of hand-picked islets (purity >90% by dithizone staining) used in the two groups (small and large islets) is given in Fig. 1. A total of 69 and 82% of the islets were in the desired size range. Insulin secretion adjusted for islet volume during perifusion was significantly different between large and small islets (area under the curve; P < 0.02; Fig. 2). Small islets with the same IEQs produced almost twice the amount of insulin compared with large islets in the basal state and during stimulation (stimulation: 1.9 ± 0.3 vs. 1.0 ± 0.1 fmol/[min × IEQs]); therefore, the insulin secretion stimulation index was not significantly different between the two groups (4.1 ± 0.5 vs. 5.3 ± 0.7, P = 0.17).

For normoxic and hypoxic conditions, cross-sections of 459 islets with 19,242 individual cells of eight different islet donors were analyzed, and distribution of glucagon- and insulin-positive cells was quantified by immunohistochemistry. Cell death was evaluated with hematoxylin and eosin staining (Table 1).

The percentage of glucagon- and insulin-positive cells in relation to total cell number in the size range of 0–50 μm was 11 ± 3 and 76 ± 2%, respectively; in the range 50–100 μm 10 ± 1 and 65 ± 2%, respectively; and >100 μm 14 ± 3 and 50 ± 4%, respectively. The percentage of glucagon-positive cells was not significantly different between islet sizes, whereas the percentage of insulin-positive cells was significantly different between the size range 0–50 and >100 μm (P < 0.0001) and between 0–50 and 50–100 μm (P = 0.0004).

In hematoxylin and eosin staining at normoxic conditions, the number of pycnotic nuclei clearly increased with islet size (Figs. 3 and 4). In the size range of 0–50 μm, 3.9 ± 0.6% of cells were pycnotic compared with 6.3 ± 0.5% of cells in the range 50–100 μm (P = 0.0023) and 7.8 ± 1.2% of cells in the range >100 μm (P = 0.0044 vs. islets 50–100 μm). Apoptosis as assessed by TUNEL staining contributes to cell death, with 12.8–23.6% of the cells being TUNEL positive independent of islet size. The percentage of insulin-positive cells undergoing apoptosis as assessed by TUNEL and insulin costaining was 18.5 ± 2.9%, whereas the number of insulin-negative cells undergoing apoptosis was 12.1 ± 2.1% (P = 0.078) independent from islet size.

The percentage of glucagon- and insulin-positive cells in the size range of 0–50 μm was 14 ± 2% and 70 ± 3%, respectively; in the range 50–100 μm, 10 ± 1% and 53 ± 3%, respectively; and >100 μm, 11 ± 2% and 38 ± 8%, respectively. The percentage of glucagon-positive cells was not significantly different between the islet sizes, whereas the percentage of insulin-positive cells was significantly different between the size range 0–50 and >100 μm (P < 0.002) and between 0–50 and 50–100 μm (P < 0.0001).

The number of pycnotic nuclei clearly increased under hypoxic conditions: 8.8 ± 1.1% of nuclei were pycnotic in the size range of 0–50 μm compared with 11.7 ± 0.9% in the size range 50–100 μm (P = 0.038) and 19.1 ± 3.5% in islets >100 μm (P = 0.015 vs. islets 50–100 μm) (Fig. 3). A total of 14.1–24.5% of cells were TUNEL positive, with no difference being observed according to islet size. The percentage of insulin-positive cells undergoing apoptosis as assessed by TUNEL and insulin costaining was 26.6 ± 5.0%, whereas the number of insulin-negative cells undergoing apoptosis was 24.8 ± 3.5% (NS) independent from islet size.

After 24 h of hypoxia 21% islets were lost compared with normoxic culture conditions, and after 48 h of hypoxia 29% of islets were lost compared with 48 h of normoxic culture conditions. Thus, there was a mean 25% loss of islet number due to hypoxia.

The islet volume after 24 h of normoxia as assessed by hematoxylin and eosin staining was 4,803,786 ± 1,214,077 μm³. This value was used as reference (100%) and compared with the value at 48 h of normoxia and at 24 and 48 h of hypoxia. The islet volume after 48 h of normoxia decreased to 3,947,276 ± 930,960 μm³. This decrease was dramatically pronounced by hypoxia. Islet volume decreased after a 48-h culture in hypoxic conditions to 1,312,861 ± 479,415 μm³ compared with 2,171,424 ± 412,254 μm³ after 24 h hypoxia (Fig. 5). The difference between hypoxic and normoxic conditions at 24 h (P = 0.017) and 48 h (P = 0.0026) was highly significant. The percentage of islets with a diameter of >100 μm did not change from 24 to 48 h of normoxia (14.4 vs. 14.2%), but there was a massive decrease in these large islets under hypoxic conditions from 7.3% after 24 h to 3.4% after 48 h of hypoxia, whereas the number of small (<50 μm) and medium (50–100 μm) islets did not change significantly. Hence, we assume that a large proportion of the islet volume loss can predominantly be attributed to large islets and not that large islets are fragmented into small islets.

The donor and recipient characteristics as well as the transplanted islet volume, islet number, and isolation index are given in Table 2. All recipients had end-stage renal failure and were listed for a simultaneous islet-kidney transplant.

The mean islet size or the isolation index (IEQs/islet number) proved to be the best predictor for the C-peptide response per islet volume per kilogram of recipient weight (Fig. 6C). Most of the variability (89%) could be accounted for by this predictor (R² = 0.89). The IEQ per kilogram of recipient weight was not able to predict stimulated C-peptide response (Fig. 6A), whereas the islet number per kilogram of recipient weight explained 84% of the stimulated C-peptide response (Fig. 6B).

Many factors, such as better identification of optimal donor characteristics (5), superior organ retrieval techniques (6,7), minimization of cold ischemia times, improved isolation technique, and avoidance of immunosuppressive drugs, which impair islet function (4), have been identified that influence the outcome of islet isolation and clinical islet transplantation (3,5,21–23). The fact that the functional capacity of transplanted islets in patients with type 1 diabetes is only about a third of the capacity of a person without diabetes indicates that two-thirds of transplanted islets are either lost or show a reduced function. To determine the amount of transplanted β-cells, IEQs are calculated depending on the number and diameter of the islets; one IEQ corresponding to a virtual islet of 150 μm diameter (8).

By reporting islet transplantation results in IEQs, one important factor for posttransplant islet survival and clinical outcome has been neglected, namely the size of the islets. Here, we provide in vitro evidence, both at the functional and at the morphological level, as well as clinical data that islet size may be a crucial determinant of islet function and survival. As shown by the perifusion experiments, insulin secretion, but not the stimulation index, is greatly influenced by islet size. When normalized to IEQs, basal and stimulated insulin secretion was twice as high in small islets compared with large islets. Indeed, MacGregor et al. (24) found that in rat islets the insulin secretion is proportional to the oxygen consumption and that in smaller islets the oxygen consumption and insulin secretion are significantly higher.

In culture and after transplantation, the oxygen supply of islets and nutrients depends solely on diffusion. The pO₂ level may be sufficiently high throughout the transplanted islets to maintain cell viability, but there is insufficient oxygen to support substantial insulin production and secretion (16). Papas et al. (25) showed that insulin secretion was gradually affected with pO₂ levels <7mmHg. In perifused cultured pancreatic islets, insulin secretion decreased by 50%, when pO₂ was lowered to 10 mmHg in the perifusate (26). If oxygen supply depends on diffusion, then the islet radius (i.e., the diffusion distance) is the determinant factor for pO₂ levels in the center of the islets. Not only oxygen supply but also a reduced nutrient supply might be a limiting factor and contribution to central necrosis (27). It has been shown that glucose concentration is lower in the center of large compared with small islets (28). Thus, due to a shorter diffusion distance, oxygen and nutrient supply is better in smaller islets, which allows for a higher insulin production. The functional data that small islets perform better than large islets are further supported by our morphological findings. At normoxic conditions small islets contain 26% more functional insulin-positive cells compared with large islets, whereas the percentage of glucagon-positive cells is not different between small and large islets. This might explain a part of the difference in insulin production between small and large human islets.

Islet cell death was strongly dependent on islet size and oxygen tension. The percentage of cell death as assessed by hematoxylin and eosin morphology doubled during hypoxic compared with normoxic culture conditions and doubled in islets with a diameter of >100 μm compared with islets <50 μm (Fig. 3). Whereas the percentage of glucagon-positive cells did not decrease in islets during hypoxia, there was a tendency toward a larger decrease of insulin-positive cells in large compared with small islets (−12 vs. −6%).

The overall 25% loss of islet number between normoxic and hypoxic culture conditions gives only a hint at the real islet volume loss, because the number of large islets (>100 μm) is almost identical after 24 and 48 h of normoxia (14.4 and 14.2%), whereas after 24 h of hypoxia there is only half the number of large islets (7.3%) left and after 48 h of hypoxia only a quarter of the number (3.4%). Considering that 19% of islet number in this size category (>100 μm) comprises 78% of the total islet volume of the pancreas (9), this 11% decrease of the large islets is responsible for a massive drop in the total islet volume (Fig. 5).

The fact that up to 16% of cells in islets were TUNEL positive during both normoxic and hypoxic conditions indicates that apoptosis is one of the mechanisms involved, possibly by hypoxia-inducible factor-1α signaling (29). However, as there was no correlation between TUNEL positivity and islet volume, mechanisms other than apoptosis, such as necrosis, must additionally be involved (27).

Several mechanisms might explain why large islets are more prone to cell death than small islets. After the isolation process, islets are no longer attached to the vascular system until revascularization takes place 7–10 days after transplantation (30). As reported by Carlsson et al. (16), transplanted islets have a pO₂ level of only 3–8 mmHg independent of the implantation site even after full revascularization. To simulate the oxygen tension in the liver, which is ∼3–5 mmHg (16), we incubated the human islets in a hypoxic chamber with 1% oxygen concentration, which corresponds to the above-mentioned oxygen tension in the liver (as demonstrated in experiments).

To assess a possible role of islet size in the clinical setting of islet transplantation we compared islet function as defined by C-peptide production in relation to islet volume in a series of islet-transplanted patients. In these patients, there was a large variance of stimulated C-peptide response in relation to the transplanted islet volume (IEQs) (Fig. 6A). By using the islet number, however, 84% of the stimulated C-peptide response could be explained. Correction for islet size by use of the isolation index (IEQs/islet number) proved to be the best predictor for the C-peptide response per islet volume (Fig. 6C), with 89% of the variability being accounted for by this parameter. Although the number of patients used for the multiple regression analysis is low, the correlation between isolation index and C-peptide response seems to be quite robust. Our findings have to be confirmed in larger series of islet transplantations, preferably in patients who received islets from a single pancreas donor.

The patient with the second smallest islets (i.e., isolation index of 0.80) received only 3,351 IEQs/kg recipient body wt and 4,181 islets/kg recipient body wt, respectively. This patient had the identical maximal C-peptide response (850 pmol/l) as a patient who received much larger islets (isolation index of 3.3: 11,544 IEQs/kg and 3,497 islets/kg recipient body wt and a C-peptide response of 870 pmol/l). In other words, with an isolation index of 3.3 a 2.6-times higher transplanted islet volume (IEQs) with large islets was required to achieve the same insulin secretion as obtained by transplantation of small islets (Fig. 6C).

Our data are in line with recent findings in a rat model. MacGregor et al. (24) demonstrated that when a marginal islet volume of large islets was transplanted into diabetic rats, all recipients failed to achieve normoglycemia, whereas 80% of the transplantations using small islets were successful. Furthermore, single-donor insulin independence has been achieved successfully with younger pancreas donors (31), and isolated islets of younger donor pancreata showed a better insulin secretion in vivo and in vitro (32). In accordance with our results, these transplanted islets have been small (isolation index <1.0) (32). The necessary number of IEQs per kilogram of recipient weight to achieve insulin independence shows a large variation in clinical practice. An isolation with a small isolation index has most likely a smaller islet volume (IEQs) than an isolation with larger islets, because small islets contribute only very little to the total islet volume (volume ∼ diameter³). It is known that 80% of the total islet number is <100 μm in the human pancreas, but these islets comprise only 22% of the total volume, whereas only 5% of islets are >150 μm, but these islets comprise 47% of the total volume (9).

Our data suggest that in the setting of human islet transplantation, islet size may be a key factor that determines transplantation outcome. We therefore propose to use the isolation index as a correction factor for the traditionally reported islet volume in IEQs by using the following equation: stimulated C-peptide = IEQs/kg × 0.2248 × isolation index^−0.9627 as shown in Fig. 6C. Alternatively, and more simply, one could rely more on islet number (islet size >50 μm) instead of IEQs, with only a minimal loss in prediction power. Based on our small patient series, a minimal islet number of ∼7,000 islets/kg recipient wt would be required to become insulin independent.

In the future, more emphasis has to be put on transplanting preferentially smaller islets. Based on our experiments and diffusion experiments in tumor cells (17), we suggest an optimal islet size of 50–100 μm, or more generally spoken, the smaller the islet the higher the chance for survival in a hypoxic milieu. By using islet isolations with an islet volume, which so far has been considered insufficient for a good clinical outcome, but with an adequate islet number, more islet isolations could now be used for transplantation. This would be a contribution to reduce the burden of organ donor shortage.

In conclusion, implementation of our findings might result in higher survival rate of islets, achievement of better islet transplantation outcomes, and economic savings by increasing the percentage of transplantable islet isolations.

We gratefully acknowledge the funding of this study by the Hans-Paul Wälchli Foundation.

We thank Heidi Seiler and Marie-Therèse Achermann for the technical assistance in the islet isolation and patient care. We also thank Dr. Vladimir Kaplan for the help in the statistical analysis.