All reagents and antibodies used in this study are listed in Supplementary Tables S1–S3.

Animal experiments were performed in accordance with the Geneva veterinary authorities and approved by the Institutional Animal Care and Use Committee of the University of Geneva. Ten-week-old, pregnant female, Lewis rats were purchased from Janvier Laboratory (Le Genest St-Isle, France) and bred in our animal facility at the Geneva University. Fifteen-to 21-week-old male rats were used for pancreatic islet isolation. Six-to 9-week old male B6.129S7‐Rag1^tm1Mom/J (abbreviated NOD–Rag1 ^null bred at Charles River Laboratories, Saint‐Germain‐Nuelles, France) mice were used as transplantation recipients. All animals were kept under conventional housing conditions with free access to water and food.

Studies involving human tissues were approved by the Commission Cantonale d’Ethique de la Recherche (CCER; protocol PB_2017-00101), in compliance with the Swiss Human Research Act (810.30).

Placentas were obtained from women undergoing elective caesarean section of uncomplicated, term pregnancies. Informed, written consent was obtained from each donor prior to tissue collection.

Human umbilical vein endothelial cells (HUVECs) were isolated using a method adapted from a previously published protocol (21). Briefly, the umbilical vein was rinsed, then distended with Collagenase A solution (2 mg/ml) and incubated at 37°C for 12 min. Released cells were then collected by flushing the vein with cold HBSS supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin, 100 mg/ml streptomycin and 0.25 mg/ml amphotericin B. Isolated HUVECs were plated in a 75 cm² flasks and cultured at 37°C, 21% O₂ and 5% CO₂ in M199 medium supplemented with 20% FBS, 100 U/ml Penicillin and 0.1 mg/ml Streptomycin (1% of a L-Glutamin-Penicillin-Streptomycin stock solution), Fungin 0.1%, 30 μg/ml endothelial cell growth supplement and 100 μg/ml heparin. HUVECs from passage 2 to 7 were used in this study.

hAECs were isolated, cultured and characterized as described previously (10, 14). Freshly isolated hAECs were cultured in DMEM/F-12 medium, supplemented with 10% FBS, 2 mmol/l L-Glutamin, 100 U/ml Penicillin, and 0.1 mg/ml Streptomycin (1% of a L-Glutamin-Penicillin-Streptomycin stock solution, 1 mmol/L sodium pyruvate, 1% MEM NEAA 100X, 0.1% fungin, 0.05 mmol/L 2-mercaptoethanol, 10 ng/ml human recombinant epidermal growth factor (EGF). Only cells at passage 1 were used in this study.

Medium was changed every 48 h. Confluent cells were recovered by mild trypsinization and were cryopreserved for later utilization.

Rat islets were isolated by enzymatic digestion (collagenase V) and purified using a discontinuous Ficoll gradient (22–24). Isolated islets were cultured (37°C, 5% CO₂) in DMEM medium supplemented with 10% FBS, 2 mmol/L L-glutamine, 100 U/ml penicillin, 0.1 mg/ml 1 mmol/L sodium pyruvate and 11 mmol/L glucose for 24 h. Islets were then dispersed into single islet cells (ICs) by incubation in 0.05% trypsin-EDTA (16).

HUVECs and hAECs were analyzed for expression of previously reported endothelial cell surface markers or specific amniotic epithelial cell surface markers by flow cytometry.

For analysis, cells (2.5 × 10⁵) were stained by incubation for 30 min with primary or isotype control antibody in 100 µl PBS with 0.2% BSA, washed twice with PBS, and analyzed. Antibodies used for HUVECs were: AlexaFluor 657-conjugated anti-CD144 (1:40 dilution), PE-conjugated anti-CD31 and PerCP-Cy 5.5-conjugated anti-CD45 (1:25 dilution). Antibodies used for hAECs were: FITC-conjugated anti-human CD105 (clone 266), BV421-conjugated anti-human CD326 (clone EBA-1), PerCP-Cy5.5 conjugated anti-SSEA4 (clone MC813-70) (1:50 dilution), PE-Cy7 conjugated anti-human CD90 (clone 5E10; 1:100 dilution), PE-conjugated anti-human HLA-E (clone 3D12) and APC-conjugated anti-human HLA-G (clone 87G; 1:20 dilution).

Flow cytometry analysis was performed on a Gallios cytometer using the Kaluza Analysis software.

HUVECs were further characterized by immunostaining. Immunofluorescent assessment was performed on the cells cultured on gelatine-coated glass coverslips. Fixed cells were washed, permeabilized and stained with the following primary antibodies: mouse anti-CD31 (1:50 dilution), rabbit anti-von Willebrand factor (1:100 dilution) and mouse anti-vimentin (1:50 dilution). Cells were then incubated with corresponding Alexa Fluor and FITC-conjugated secondary antibodies. For nuclear counterstaining samples were mounted with aqueous solution containing 4,6 diamidino-2-phenylindole (DAPI).

The tube formation assay was performed according to manufacturer’s protocols of Corning® Matrigel® Matrix. Briefly, Matrigel thawed overnight at 4°C was mixed with VEGF (200 ng/ml) and 250 μl of matrix was added to each well osf 24-well plates. After 1 h of incubation at 37°C, cells (8 × 10⁴) were seeded onto the Matrigel and tube formation of HUVECs was observed and photographed using an inverted phase-contrast microscope during 6 h.

Lentiviral vector carrying the green fluorescent protein (GFP) under the control of an endothelial specific promoter Vascular endothelial cadherin (VEC/Cdh5) (LV-VEC.GFP) was provided by Prof. A. Follenzi (Università del Piemonte Orientale). HUVECs were transduced with LV-VEC.GFP at passage 3 using a multiplicity of infection (MOI) of 10 (MOI = 10). Transduction efficiency was assessed by fluorescent microscopy and flow cytometry and considered successful when at least 80% of cells showed expression of GFP.

Pre-vascularized islet organoids (PIO) were generated on AggreWell™400 24-well plates by seeding mixture of ICs, HUVECs and hAECs at a ratio of 5:4:1 (800 cells/organoid). Undissociated native islets (NI), ICs spheroids (400 ICs/spheroid), hereafter referred to as pseudo-islets (PI), and IC:HUVEC spheroids (ratio 1:1, 800 cells/spheroid), hereafter referred to as IC + HUVEC served as controls. PIO, PI and IC + HUVEC were cultured for 4 days to allow cell aggregation at 37°C, 21% O₂ and 5% CO₂.

Culture medium for PIO was prepared by mixing equal volumes of complete DMEM, DMEM/F12 and M199 medium, hereafter referred to as organoid medium. IC + HUVEC were cultured in the mixture of complete DMEM and M199 medium at the ratio 1:1. Finally, PI and NI were cultured in complete DMEM medium. Culture medium was changed every other day. Mean diameter of NI, PIO and PI were calculated on the images taken on light microscope using ImageJ software.

In order to observe PIO composition and cell distribution during culture, fluorescent carbocyanine dyes CM-DiL (red) prelabeled hAECs and GFP transduced HUVECs were used. Pictures were taken using an epifluorescent microscope (DMi8 manual microscope).

PIO, PI and NI were collected fixed in formalin and embedded in paraffin. Serial sections of 5 μm were cut and processed for immunofluorescent staining. Slides were stained with the following primary antibodies: guinea pig anti-insulin (1:100), chicken anti-GFP (1:500), and rabbit anti-CK-7 (1:100). The following secondary antibodies were then applied: donkey anti–guinea pig Alexa 555 Fluor-conjugated (1:300), donkey anti–guinea pig FITC‐conjugated (1:200), donkey anti-mouse AMCA-conjugated (1:50), goat anti-chicken Alexa Fluor 488 (1:500).

One hundred PIO were resuspended in a collagen solution, transferred into prewarmed 24‐well plates and allowed to gelify for 30 min. Next, 0.1 ml organoid medium supplemented with VEGF-A at the concentration of 200 ng/ml was pipetted on top of each hydrogel containing PIO. The hydrogels were cultured for 24 h at 37°C, 5% CO₂, and 100% humidity. As control, one hundred IC + HUVEC spheroids and PI were cultured in the same way in the hydrogel.

To assess functional capacity, 300 NI and an equivalent number of PIO and PI, were incubated in duplicates for 1 h at 37°C in Krebs–Ringer solution containing low glucose (2.8 mmol/L) in order to equilibrate the samples. After a change of medium, islets and aggregates were incubated at 37°C for another hour in Krebs–Ringer solution containing low glucose (2.8 mmol/L), followed by 1 h at high glucose (16.7 mmol/L). Supernatants were collected and stored at −20°C. Insulin concentration in supernatants was measured using a rat insulin ELISA kit and normalized to the total insulin content. Results are expressed as the ratio between insulin secreted in high glucose to low glucose, referred to as stimulation index (SI). In addition, total insulin content per IC was measured by dividing the total insulin content by the number of ICs present in the NI, PI and PIO.

Three days before transplantation mice were subjected to intraperitoneal injection of STZ (180 mg/kg). Non-fasting blood glucose levels were then checked daily using a portable glucometer. Only mice with blood glucose levels over 18 mmol/L for 3 consecutive days were used in this study. Glycemia readings over 28 mmol/L, indicated as “high” on glucometer, were recorded as 30 mmol/L.

A marginal mass of 300 islet equivalents (IEQ) for NI and 1200 PIO, PI and IC + HUVEC were transplanted. Number of organoids was based on the average number of islet cells per IEQ, previously estimated as 1,560 ICs/IEQ (25).

At the day of transplantation, NI and engineered constructs were recovered from culture, packed in PE50 tubing and transplanted into the epididymal fat pad (EFP) of diabetic mice. Non-fasting glucose was assessed daily during the first week and 3 times per week thereafter. Normoglycemia was defined as two consecutive blood glucose levels under 11.1 mmol/L.

Graft capacity to clear glucose in vivo was assessed dynamically by intraperitoneal glucose tolerance test (IPGTT) at 30 days after transplantation. Mice were fasted for 6 h and intraperitoneally injected with 2 g of glucose/kg. Blood glucose measurements were taken at 0, 15, 30, 45, 60 and 120 min.

Functional graft vasculature was assessed by infusing DyLight 594-conjugated Lycopersicon Esculentum (Tomato) lectin into the beating left ventricle of mice hearts. Mice were injected with 100 μl of undiluted lectin. Lectin was allowed to circulate for 1 min. Then, the right ventricle was cut to allow blood flow decompression and a volume of 3 ml of PBS was injected into the left ventricle, followed by 1 ml of 4% PFA. The graft bearing EFPs were collected and fixed overnight in 4% PFA at 4°C. They were then maintained in 30% sucrose at 4°C until used for histology.

Grafts were recovered, fixed in formalin and embedded in paraffin. Serial sections of 5 μm were cut and processed for immunofluorescent staining. Tissue samples were permeabilized with 0.5% Triton X-100/PBS for 30 min, followed by 1-h incubation in 0.5% BSA/PBS at room temperature to block unspecific sites. Slides were then incubated with the following primary antibodies: guinea pig anti-insulin (1:100), rabbit anti-CD34 (1:2,000), chicken anti-GFP (1:500), and rabbit anti-VEGF (1:100). The following secondary antibodies were then applied: donkey anti–guinea pig Alexa 555 Fluor-conjugated (1:300), donkey anti–guinea pig FITC‐conjugated (1:200), donkey anti‐rabbit Alexa 555 Fluor-conjugated (1:300) and goat anti-chicken Alexa Fluor 488 (1:500). Both primary and secondary antibodies were diluted in PBS-0.5% BSA. Finally, slides were mounted with aqueous mounting medium containing DAPI for nuclear staining. Slides were processed on a Zeiss Axioscan.Z1 slide scanner and a Zeiss Axiocam. To analyse vascularization, six pictures per condition were taken and the number of CD34⁺ cells were counted and normalized by the graft area.

Morphometric analysis was performed using Zen 2.3 Blue Edition software.

Graft bearing EFPs recovered at 3 and 30 days after transplantation were processed for PCR analysis. RNA was extracted using the RNeasy minikit and reverse transcribed with a High Capacity cDNA Reverse transcription kit. Gene amplification was performed by RT-PCR using TaqMan Fast Advance Master Mix. Primers used for amplification are listed in Supplementary Table S4. RPLP1 was used as a housekeeping gene to normalize gene expression values. Data were calculated using the comparative cycle threshold Ct method (2^−ΔCt method) and are expressed in arbitrary units.

Continuous variables are expressed as mean ± SEM. Multiple comparisons were analyzed using one-way ANOVA followed by Dunnett multiple comparisons test while two-way comparisons were analyzed using the Student’s t-test. Cumulative number of animals reaching normoglycemia was compared using the log‐rank (Mantel‐Cox) test. A p value ≤0.05 was considered statistically significant. All statistical analyses were performed with the Prism software 8.0.

HUVECs reached 80% confluence within 5 days with initial seeding density of 6,000 cells/cm². Morphologically, cells displayed typical elliptic shape (Figure 1A) and were positive for von Willebrand factor and CD31 (Figure 1B). Endothelial origin of the cells was additionally confirmed by flow cytometry. Cells were positive for CD31 and CD144 (97.8% ± 0.7 and 98.1% ± 0.6, respectively) and negative for CD45 (95.8%) (Figure 1C).

When cultured on Matrigel, HUVECs formed well-shaped vascular-like structures over a period of 6 h (Figure 1D).

To track HUVECs within organoids both in vitro and in vivo, cells were transduced with LVs carrying green fluorescent protein (GFP) gene under the control of the VEC promotor. HUVEC positivity for GFP was observed during culture and confirmed by flow cytometry 3 days after transduction with 86.6% of GFP+ cells (Figure 1E right and left panel, respectively).

hAECs used in this study were isolated from six different placentas. Flow cytometry analysis demonstrated strong positivity of hAECs for the embryonic cell surface marker SSEA-4 (88.4 ± 5.0%) and the epithelial cell adhesion molecule (CD326; 95.9 ± 1.3%). HLA-E and HLA-G were expressed in 16.9 ± 4.7% and 48.6 ± 12.3% of the cells, respectively. Finally, expression of CD105 and CD90 by hAECs were 17.6 ± 5.6%, 50.1 ± 7.1, respectively. The results of each hAEC preparation are described in Supplementary Figure S1.

Generation of PIO and PI is described in Figure 2A. Aggregation and incorporation of the different cell types occurred within 4 days (Figures 2B,C). Mean diameter of NI, PI and PIO was 144.4 ± 6.6, 105.8 ± 1.2 and 134.3 ± 2.3 μm, respectively (Figure 2D). NI showed the biggest heterogeneity in size. PI exhibited a significantly smaller mean diameter in comparison with PIO (p < 0.0001), due to fewer cellular content. Cellular composition observed by fluorescent microscopy showed that all 3 cell types were present in the PIO (Figure 2E). The functional capacity of the constructs was evaluated by glucose-stimulated insulin secretion (GSIS) assay. PI and PIO demonstrated significantly improved insulin secretion in response of glucose stimulation (SI = 7.8 ± 1.5 and 7.7 ± 1.2), compared to NI (SI = 2.0 ± 0.5, p = 0.013 and p = 0.014, respectively). No significant difference was observed between PI and PIO (Figure 2F). In addition, total insulin content/IC was measured and compared between the three groups. PI and PIO demonstrated an increased insulin content/IC (0.01 ± 0.003 and 0.008 ± 0.002 pmol/L, respectively) in comparison with NI (0.002 ± 0.0004 pmol/L). These dramatic enhancement of static GSIS secretion in our constructs compared to unmodified native islets indicate that better oxygen and nutrient access, and improved transport of glucose and insulin, enhanced survival and function of PI and PIO. Our findings are consistent with previous reports on better in vitro performance of smaller pseudoislets (26, 27).

To investigate the angiogenic potential of the PIO, collagen-based sprouting assays were performed. Our results demonstrated that PIO showed more extensive sprouting in surrounding matrix compared to IC + HUVEC (Supplementary Figure S2). In contrast, no sprouting was observed from PI (data not shown). Furthermore, immunofluorescence revealed GFP positive cells confirming their endothelial nature.

To assess whether incorporation of hAECs and HUVECs into the islet organoids could promote engraftment and function in vivo, diabetic NOD–Rag1 ^null mice were transplanted with a marginal mass of PIO (n = 14), NI (n = 13) and PI (n = 9). Mice transplanted with PIO demonstrated significant improvement of glycaemic control compared to both controls. Average blood glucose levels were significantly lower in the PIO group compared to NI and PI (Figure 3A). Normoglycemia was reached in 78.6% of animals (11/14) in the PIO group, in comparison with 55.6% (5/9) and 46.2% (6/13) for the PI and NI groups, respectively (Figure 3B). Median time to achieve normoglycemia was 6 days in the PIO group, 21 days in the PI group and >30 days in the NI group. To investigate secretory function of the graft, IPGTT was performed at 30 days post-transplantation. Mice transplanted with PIO and non-diabetic controls (NDC) showed lower blood glucose levels when compared to animals transplanted with PI and NI (Figure 3C). This is illustrated by the increasing area under the curve (AUC) of the different groups, with PIO (966.8 ± 113.7), PI (1783 ± 351.1, p = 0.05 vs. PIO) and NI (1856 ± 294.5, p = 0.014 vs. PIO; Figure 3D).

We further investigated whether the improved glycemic control in the PIO group was associated with insulin production from the transplanted β cells. Remarkable upregulation of rat insulin mRNA levels in the graft was found in the PIO group in comparison to controls (PIO vs. PI, p = 0.013, PIO vs. NI, p = 0.013; Figure 3E). These results were supported by insulin measurements in the serum taken from the same mice (Figure 3F). Although a statistical significance wasn’t achieved, a ten-fold increase in insulin levels was detected in the PIO group (1,259 ± 521 pmol/L), in comparison to both controls (NI: 140.6 ± 22.1 pmol/L, PI: 159.8 ± 14.4 pmol/L, p = ns).

Glp-1r, pdx1 are known to be critical for promoting insulin secretion (28–31). Therefore, we investigated whether these genes were involved in the improved secretory outcomes of PIO. Gene expression analyses revealed upregulation of genes involved in β-cell function (pdx1, pcsk1, pcsk2 and glp-1r) in PIO at 30 days post-transplantation, compared to controls (pdx1: PIO vs. PI, p = 0.0009, PIO vs. native islet, p = 0.0009; glp-1r: PIO vs. PI, p = 0.002, PIO vs. native islet, p = 0.002; pcsk1: PIO vs. PI, p = 0.02, PIO vs native islet p = 0.021 and pcsk2: PIO vs. PI, p = 0.0005, PIO vs. native islet, p = 0.0006; Figure 3G). Interestingly, at an earlier time points (3 days), a similar increase in gene expression was observed in PI and PIO in comparison with NI group, although without reaching statistical differences (Supplementary Figure S3). These data indicate that incorporation of accessory cells into the organoids supports long term secretory function of β cells.

To evaluate engraftment and revascularization, graft-bearing EFPs were removed at 30 days post-transplantation and processed for histology. Immunohistochemical staining for CD34, a marker for endothelial cells, showed that vessel density was significantly higher in the PIO samples (22.6 ± 3.5 CD34 + cells/cm²) than in the NI samples (7.6 ± 0.9, p = 0.002; Figures 4A,B). Furthermore, in the PIO group, vessels were observed not only around graft, but mainly within β-cell positive area.

To investigate whether the blood vessels formed within the engrafted tissue constructs become functional and contribute to graft perfusion, we used intravascular injection of fluorescently labeled Lectin. Histological assessment of the Lectin-perfused grafts demonstrated the presence of functional Lectin positive vascular network within the PIO, in contrast only few vessels were present within NI (Figure 4C).

Next, we examined the mechanisms by which supportive cells (HUVECs and hAECs) contributed to rapid neovascularization of the graft. To this end, we investigated whether these cells might induce the production of angiogenic factors, such as vegf-a (Figure 4D). We observed, that rat vegf-a mRNA expression was significantly higher in PIO group (0.365 ± 0.033 AU) compared to NI (0.038 ± 0.005 AU; p = 0.0006) group. This finding was further confirmed by immunohistochemical staining for vegf-a of recovered samples, demonstrating higher fluorescent intensity in the PIO compared to NI (Figure 4E). These data indicate that incorporation of HUVEC and hAEC into PIO contribute to graft revascularization.

Finally, we evaluated whether incorporation of hAECs into the organoids was essential for the engraftment and vascularization of the PIO. To this end, we added an additional group of mice transplanted with spheroids composed of IC: HUVEC (1:1 ratio) to the three existing groups.

Figure 5 summarizes the results obtained with this group. Blood glucose control was significantly lower in the IC + HUVEC group in comparison to the PIO group (Figure 5A). The IPGTT performed at 30 days post-transplantation demonstrated a poor glucose clearance in the IC + HUVEC group (Figure 5B). Response to increased blood glucose levels was significantly lower than for the PIO group as demonstrated by the AUC (2044 ± 578.1 vs. 966.8 ± 113.7, p = 0.008, respectively; Figure 5C).

After demonstrating that incorporation of supportive cells into the PIO improved graft revascularization, we investigated the degree to which these cells were contributing to new vessel development in the graft. To easily identify donor-derived new vessels, GFP-transduced HUVECs were incorporated into the PIO. Graft-bearing EFPs were recovered at 30 days post-transplantation and processed for immunohistological analysis. Interestingly, GFP positive cells were found inside the graft in the PIO group, while none was found in the IC + HUVEC group (Figure 5D). Both human and mouse vessels were positively stained by anti-CD34 confirming the establishment of anastomoses between donor derived HUVECs and mouse blood vessels. Furthermore, GFP/CD34 double positive endothelial cells were found at the graft periphery, inside capillaries containing erythrocytes, indicating that HUVECs were able to migrate and merge with a murine vascular system, forming functionally perfused blood vessels, as shown in Figure 5E. These data indicate that hAECs support HUVECs inside the organoids and thus contribute to accelerated revascularization.