German landrace hybrid piglets served as pancreas donors. NOD-SCID IL2rγ^−/− (NSG) mice, which lack mature T cells, B cells and NK cells, were obtained from The Jackson Laboratory (strain 005557) and housed under standard SPF conditions. All animal experiments were approved by the responsible authority and performed in agreement with the German Animal Welfare Act and Directive 2010/63/EU.

NPICCs were isolated from pancreata (n = 12) of 2–5 days-old piglets by collagenase digestion as described previously (22, 23). Cell clusters were cultured in RPMI 1640 (PAN-Biotech, Aidenbach, Germany), 2% human serum albumin (Takeda, Konstanz, Germany), 10 mM nicotinamide, 20 ng/ml exendin-4 (Merck, Darmstadt, Germany) and 1% antibiotic-antimycotic (Thermo Fisher Scientific, Germering, Germany) (basal islet culture [B-IC] medium). On day 4, NPICCs were harvested and islet equivalents (IEQ) were determined under a stereomicroscope. NPICCs were either re-cultivated in B-IC medium (control group) or washed with PBS and incubated in TrypLE Express solution (Thermo Fisher Scientific) for 8–12 min at 37°C with gentle mixing every 60 s until dispersion into single cells was observed under a stereomicroscope. Then, the cells were filtered through a 40-µm filter (Corning, Wiesbaden, Germany) to remove debris.

Dispersed islet cells were selected at random and seeded on Sphericalplate 5D (Kugelmeiers, Erlenbach, Switzerland), which contain 750 microcavities per well, in 2 ml B-IC medium to yield re-aggregated clusters composed of 750 cells. This cell number was chosen from previous studies reporting on an optimal islet aggregate size of about 100 µm (6, 10, 24, 25). Plates were centrifuged at 250xg for 3 min and incubated in B-IC medium for 3 days. Then, clusters were gently flushed out from the wells, washed with medium or buffer, and used for measurement of cell viability, glucose stimulated insulin secretion (GSIS), and transplantation.

Cell viability was detected by calcein AM (live cells) and propidium iodide (dead cells) dye staining according to the manufacturer´s instruction (Thermo Fisher Scientific). Samples were analyzed under a fluorescent microscope (n > 50 cluster) and by flow cytometry at day 7 after isolation. To determine the cluster architecture, NPICCs and REPIs were embedded in Epredia™ HistoGel™ Specimen Processing Gel (Thermo Fisher Scientific) and stained for insulin (guinea pig anti-insulin, 1:400, Agilent-Dako, Frankfurt, Germany) and glucagon (rabbit anti-glucagon, 1:100, Cell Signaling, Frankfurt, Germany) followed by incubation with FITC-labelled anti-rabbit IgG and Cy3-labeled anti-guinea pig IgG (Thermo Fisher Scientific). DAPI was used to counterstain cell nuclei. Recovery rate was determined by calculation of the ratio IEQ of REPIs to IEQ of native NPICCs. Apoptotic cells were analyzed on day 7 by TUNEL staining of NPICCs and REPIs using the DeadEnd Fluorometric TUNEL System assay according to the manufacturer´s instructions (Promega, Madison, WI, United States).

On day 7 after isolation, 100–150 islet equivalent (IEQ) of NPICCs and REPIs were washed in glucose-free RPMI medium and Krebs-Ringer buffer (KRB) solution and pre-incubated in KRB containing 2.8 mmol/L glucose (low glucose) for 1 h at 37°C and 5% CO₂ followed by incubation in duplicates in KRB with low glucose or high glucose (20.0 mmol/L) for 1 h. Then, supernatants were collected and porcine insulin concentration was measured in duplicates by ELISA (Mercodia, Uppsala, Sweden). The stimulation index (SI) was calculated by dividing the insulin concentration in high glucose by insulin concentration in low glucose (23).

After induction of diabetes by intraperitoneal injection of 180 mg/kg body weight streptozotocin, diabetic NSG mice (blood glucose levels >350 mg/dl) received native NPICCs (3000, 1500, or 750 IEQs/mouse) or REPIs (750 or 1500 IEQs/mouse) under the left kidney capsule as described recently (22, 26). 3000 IEQs represent the standard dose of NPICCs for transplantation to achieve normoglycemia in about 80% of diabetic NSG mice. Blood glucose levels were monitored by FreeStyle Lite blood glucose test strips (Abbott, Wiesbaden, Germany). Diabetic mice with blood glucose levels >300 mg/dl were treated with insulin glargine (0.25–1 IE s.c. daily). Our primary endpoint was normoglycemia. The observation period was set to a maximum of 16 weeks. We used a stringent definition for normoglycemia which was specified as achievement of pretransplant glycemia (persistent random non-fasting blood glucose levels <120 mg/dl). This cut-off is based on the measurement of non-fasting blood glucose levels of untreated NSG mice (n = 107; 94.6 ± 15.2 mg/dl, range 62–128 mg/dl). In this control cohort there was no animal with non-fasting blood glucose levels above 120 mg/dl on two consecutive days. Because other studies often used non-fasting blood glucose levels <180 mg/dl to define normoglycemia after islet transplantation this threshold was also analyzed.

Glucose tolerance in transplanted mice was determined by intraperitoneal glucose tolerance test (IPGTT) 10–14 days after development of normoglycemia using 2 g glucose/kg body weight. Blood samples were obtained from the tail vein at 0 and 10 min to measure porcine insulin in duplicates by ELISA (Mercodia) that had no cross-reactivity with mouse insulin. To provide evidence that normoglycemia was mediated by the grafted tissue and not by pancreatic islet regeneration, graft bearing kidneys were removed in three transplanted animals followed by daily blood glucose measurements for 3 days. When the post uninephrectomy blood glucose level was >400 mg/dl, the achieved normoglycemia was considered as graft-dependent.

Paraffin sections of the graft bearing kidney were stained with the following antibodies (Ab): guinea pig anti-insulin (1:400, Agilent-Dako), rabbit anti-glucagon (1:100, Cell Signaling), rabbit anti-somatostatin (1:50, Agilent-Dako), rabbit anti-pancreatic polypeptide (PP) (1:5000, Proteintech) and rabbit anti-CD31 (1:100, Cell Signaling). Secondary antibodies used were HRP- or alkaline phosphatase-conjugated anti-guinea pig IgG (Agilent-Dako) and anti-rabbit IgG (Vector Laboratories, California, United States). Fuchsin + substrate chromogen (Agilent-Dako) or 3,39-diaminobenzidine (Kem-En-Tec Nordic A/S, Uppsala, Sweden) were used as chromogen. To visualize glucagon and PP, the ImmPRESS® HRP horse anti-rabbit IgG polymer detection kit (Vector Laboratories) was used.

Cellular composition of NPICCs and REPIs were quantified by QuPath software (version 0.3.2) using pictures scanned by uScope MXII slide scanner (Microscope International, Dallas, United States). The numbers of cells that stained positive for glucagon, somatostatin or PP were expressed as number of positive cells per 100 insulin positive beta cells. To assess vascularization, the area of CD31 positive cells (endothelial cell marker) were detected and normalized to the islet area.

Accuri C6 flow cytometer (BD Biosciences, Heidelberg, Germany) was used to determine the number of insulin-, glucagon-, and somatostatin-positive cells in NPICCs and REPIs. Single cells were prepared by digesting cell clusters with TrypLE solution (Thermo Fisher Scientific), washed with PBS +10% fetal calf serum (FCS), and filtered through a 30 µm pre-separation filter (Miltenyi, Bergisch-Gladbach, Germany). Then, cells were fixed/permeabilized with an intracellular staining buffer set (Thermo Fisher Scientific) and incubated with Fc-Block (anti-mouse CD16/CD32) for 10 min at room temperature. Thereafter, cells were stained with fluorochrome-labeled antibodies against insulin (anti-insulin-AF647, clone T56-706), glucagon (anti-glucagon-PE, clone U16-850), and somatostatin (anti-somatostatin-AF488, clone U24-354) (BD Biosciences). All antibodies were pretested for appropriate dilution and specificity using isotype control antibodies. Antibodies were incubated at 4°C for 30 min, washed two times with permeabilization buffer and analyzed on a flow cytometer with FlowJo software version 10.4 (TreeStar, Ashland, United States).

Data are presented as means and standard deviations (SD). Statistical differences between two groups were analyzed with Shapiro-Wilk test for normality and F test for homogeneity of variance and examined with the Mann-Whitney U test. Time to normoglycemia was compared with log-rank test. The area under the curve (AUC) was calculated using trapezoidal rules. p values <0.05 were considered statistically significant. Statistical analyses were performed using GraphPad Prism (version 9.2, GraphPad, San Diego, United States).

On day 3 after isolation, NIPPCs were dissociated by mild TypLE treatment resulting in single cells with a viability of 88.2 ± 5.6%. After 3–4 days of seeding in 3D-cluture plates, clusters composed of 750 re-aggregated cells showed a uniform size with a mean diameter of 113.3 ± 10.5 µm (range 94.8–130.4, median 113.4 µm) whereas the diameter of native NPICCs was much more heterogeneous (range 50.1–289.6 µm, median 95.2 µm) (Figures 1A–C). The size variation of native NPICCs was significantly higher as compared to REPIs (p < 0.01). As expected, there was cell loss during the cell dissociation and re-aggregation process. The recovery rate, defined as IEQ of REPIs compared to control NPICCs on day 7 after isolation was 63.5 ± 18.9% (Figure 1D).

Fluorescence microscopy revealed that insulin-positive cells were more uniformly dispersed and their relative abundance was increased in REPIs as compared to NPICCs (Figure 1E). This was confirmed by flow cytometry where slightly higher percentage of insulin- and glucagon-positive cells were detected. The proportion of somatostatin-positive cells was not altered. The percentage of endocrine cells (sum of alpha, beta and delta cells) was significantly increased in re-aggregated clusters (55.9 ± 2.2% vs. 50.0 ± 2.2% in NPICCs) (p < 0.05) (Figure 1F).

Glucose-dependent insulin secretion was assessed in four independent experiments with handpicked NPICCs and REPIs. REPIs exhibited a significantly higher GSIS (1.9-fold) as compared to native NPICCs (1.4-fold) (p < 0.05) (Figure 1G). Taken together, compared to conventional NPICCs, REPIs had a higher homogeneity in size, contained higher proportions of endocrine cells, and could serve as a suitable source for in vivo transplantation.

We next validated the in vivo function of REPIs in diabetic NSG mice by transplanting 750 and 1500 IEQ REPIs. We included groups transplanted with 750, 1500, and 3000 IEQ native NPICCs as controls. All animals remained diabetic after transplantation of 750 native NPICCs or 750 REPIs (Figure 2A). Prevalence of non-fasting blood glucose levels <120 mg/dl was 85.7% in mice transplanted with 1500 IEQ REPIs (n = 6 of 7) (median diabetes reversal time 63 days) as compared to around 33% (n = 2 of 6) in those transplanted with 1500 native NPICCs (median diabetes reversal time 98 days) (p < 0.05) (Figure 2A). Notably, the reversal rate and median diabetes reversal time of mice transplanted with 1500 REPIs (85.7% 63 days) were comparable to those of mice with 3000 native NPICCs (77.2%; 66 days), indicating a higher efficacy per IEQ unit for REPIs.

Similar results were obtained by using blood glucose levels <180 mg/dl as the threshold for normoglycemia. Percentage of mice below this cut-off was significantly higher in animals transplanted with 1500 IEQ REPIs (85.7%, median time 59 days) versus 1500 native NPICC (50%, median time 90 days) (p < 0.05) (Supplementary Figure S1).

We further performed IPGTT in mice developing normoglycemia. As shown in Figure 2C, the glucose clearance in the group that received 1500 REPIs (AUC 6260 ± 305.3) was significantly better than that in the group with 3000 NPICCs (AUC 8073 ± 536.2) (p < 0.01). Plasma insulin levels at 0 and 10 min during IPGTT were similar in all groups (Figure 2D). These data suggested that the re-aggregation of clusters from dispersed NPICCs resulted in a significant improvement of in vivo function.

To characterize the composition of endocrine cells and the density of vascularization following transplantation with NPICCs and REPIs, immunohistochemistry was performed at the end of the post-transplantation period. Three grafts per group were stained to detect endocrine cell components including insulin, glucagon, somatostatin and pancreatic polypeptide (PP) positive cells. Although there was a random rearrangement of endocrine cells in the first days after re-aggregation, this picture changed during in vivo maturation. As shown in Figure 3, the grafts derived from REPIs and from NPICCs consisted of a core of insulin-positive cells, surrounded by glucagon-, somatostatin- and PP-positive cells, which mainly located in the islet mantle (Figures 3A,B), suggesting the spatial reorganisation of endocrine cells in vivo during the post-transplant period.

There were small differences in the morphology between REPI and NPICC grafts. The relative proportion of insulin-positive cells in REPIs was slightly higher than in NPICC grafts (89.1 ± 3.4% vs. 82.6 ± 3.2%) . Conversely, glucagon- (10.9 ± 3.3% vs. 17.3 ± 3.2%) and somatostatin- (0.2 ± 0.1% vs. 2.3 ± 0.1%) positive cells were slightly decreased while the relative proportion of PP-positive cells was significantly reduced in REPI grafts (0.2 ± 0.1% vs. 1.2 ± 0.2%) as compared to NPICC grafts from the same pig donors (p < 0.01). (Figure 3B). Quantification of CD31 staining within the grafts revealed no difference in endothelial cell (EC) area in both groups (Figures 3C,D).