Abstract
Cell identity switches, where terminally-differentiated cells convert into different cell-types when stressed, represent a widespread regenerative strategy in animals, yet they are poorly documented in mammals. In mice, some glucagon-producing pancreatic α-cells and somatostatin-producing δ-cells become insulin expressers upon ablation of insulin-secreting β-cells, promoting diabetes recovery. Whether human islets also display this plasticity, especially in diabetic conditions, remains unknown. Here we show that islet non-β-cells, namely α-cells and PPY-producing γ–cells, obtained from deceased non-diabetic or diabetic human donors, can be lineage-traced and reprogrammed by the transcription factors Pdx1 and MafA to produce and secrete insulin in response to glucose. When transplanted into diabetic mice, converted human α-cells reverse diabetes and remain producing insulin even after 6 months. Surprisingly, insulin-producing α-cells maintain α-cell markers, as seen by deep transcriptomic and proteomic characterization. These observations provide conceptual evidence and a molecular framework for a mechanistic understanding of in situ cell plasticity as a treatment for diabetes and other degenerative diseases.
Fostering cell regeneration in damaged tissue is one of the cornerstones of regenerative medicine. Attempts at reprogramming human fibroblasts, keratinocytes or pancreatic exocrine cells toward insulin production have been unsatisfactory 1–4. In diabetic mice, insulin-producing cells are naturally reconstituted by consistent but rare islet cell-type interconversion events 5,6,7. In human islets, bihormonal cells have been described under certain conditions in vitro 8,9 and in diabetics 10–12. This is compatible with the existence of adaptive cell identity changes that must be confirmed and understood to harness their regenerative potential in diabetes 8,13–15. Indeed, these claims remain speculative without evidence based on cell-tracing analyses 16,17.
Pdx1 and MafA are required for β-cell development, maturation and function 18. We and others have previously shown that overexpression of these transcription factors (TFs) convert embryonic or adult α-cells of diabetic mice into insulin-producers 13,19. In this work we show that human α- and γ-cells may become glucose-dependent insulin secreters.
Generation of human monotypic pseudoislets
We devised a multistep approach to study islet cell plasticity: i) cell-sorting by flow cytometry using cell-surface antibodies 20, ii) adenoviral GFP-labeling of purified islet cells expressing Pdx1, MafA and/or Nkx6.1, iii) reaggregation of labeled cells into monotypic “pseudoislets”, i.e. islet-like 3D-clusters containing only one islet cell-type, and iv) in vitro and in vivo functional, molecular profiling, and immunogenicity analyses (Fig. 1a).
Cell purity after sorting was 99% for α- and β-cells, and 94% for γ-cells (Extended Data Fig. 1a-c), while δ-cell purification was unreliable with available antibodies (Extended Data Fig.1; Supplementary Tables 1 and 2). Purified islet cells were transduced with different TFs and GFP (>99% efficiency; Extended Data Fig. 1d), allowing a traceability equivalent to genetic cell-lineage analyses in mice.
Since dissociated single β-cells fail to secrete insulin in vitro 21, we validated the experimental approach by assessing the secretory function of monotypic pseudoislets of β-cells. We reaggregated the GFP-labeled β-cells either alone or with human mesenchymal stem cells (MSCs) and umbilical vein endothelial cells (HUVECs), to recreate a beneficial niche for cell function 22,23 (Extended Data Fig. 2a,b). β-cell pseudoislets formed within a day in the presence of HUVECs and MSCs (hereafter “HM”) (Extended Data Fig. 2a,c). β-cell pseudoislets showed rare apoptosis and restored glucose-stimulated insulin secretion (GSIS), comparable to that of native islets (Extended Data Fig. 2d–g). The absence of α-, δ- and γ-cells in monotypic β-cell pseudoislets did not affect GSIS ex vivo. Thus, simple reaggregation invigorates the purified human β-cells, likely by reconstituting an islet-like environment 24.
Insulin secretion by transduced human α-cells
Pdx1, MafA and Nkx6.1 are β-cell-enriched TFs spontaneously upregulated in insulin-producing α-cells after total β-cell ablation in mice 6. We thus explored whether human non-β-cells acquire insulin production upon ectopic expression of these factors. We transduced purified human α-cells with bicistronic adenoviral vectors expressing a murine β-cell TF along with GFP (Pdx1-GFP, MafA-GFP and Nkx6.1-GFP), before pseudoislet reaggregation and analysis (a week later; Fig. 1a,b).
Control GFP-transduced α-cells (αGFP pseudoislets) were insulin-negative (99.4±0.4 %), while the Pdx1 and MafA combination (αPM pseudoislets) triggered the highest reprogramming efficiency (38.3±5.0%; Fig. 1c,d and Extended Data Fig. 3a,b). After a week in culture, nearly all insulin-producing α-cells maintained GCG and ARX expression (Fig. 1d and Extended Data Fig. 3c,d; see below the RNA profiling). αPM cells cultured as single-cells displayed a much lower reprogramming frequency (3.9%) ex vivo or in vivo after transplantation (Figure 1e; Extended Data Fig. 3e, and not shown). Similar to β-cells, α-cells aggregated faster into pseudoislets in the presence of HM cells, though reprogramming frequency remained unchanged (Figure 1e; Extended Data Fig. 3f–h). Apoptosis and proliferation were rare (Extended Data Fig. 3i,j). Both αPM and αPM+HM pseudoislets displayed significant GSIS in culture (Fig. 1f), with HM cells further enhancing secretion. Therefore, Pdx1/MafA coexpression engages human α-cells into glucose-dependent insulin secretion.
Insulin secretion by transduced human γ-cells
We observed that PPY-producing γ-cells transduced with PM engage in insulin production as efficiently as α-cells, while maintaining PPY expression (Extended Data Fig. 4a–d). HM cells accelerated reaggregation, yet decreasing reprogramming frequency (Extended Data Fig. 4e–g). γPM pseudoislets secreted insulin upon glucose stimulation, even better than α-cells (Figs. 1f; Extended Data Fig. 4h). This is the first observation of γ-cell plasticity. Combined, these observations represent the first direct evidence for the plasticity of mature human islet non-β-cells.
Diabetes remission by insulin-secreting α-cells
Pseudoislets maintained in culture lose cells steadily, yet insulin mRNA levels increase (Extended Data Fig. 4i,j). This suggests that culture conditions are not optimal but reprogramming nevertheless progresses with time. To evaluate pseudoislet function in vivo, we transferred pseudoislets made of αPM+HM cells (“αPM pseudoislets” hereafter) to immunodeficient NSG mice, in five different experiments (Fig. 2a; Supplementary Table 3). α-cells from non-diabetic donors were transplanted into healthy (Experiment #1: Extended Data Fig. 5) or diabetic host mice (Exps. #2 to #4: Fig. 2, Extended Data Fig. 6), and from type 2 diabetic (T2D) donors to diabetic mice (Exp. #5: Fig. 2, Extended Data Fig. 7).
We detected circulating human insulin upon glucose stimulation in healthy mice given αPM pseudoislets from non-diabetic donors (Exp. #1: Extended Data Fig. 5a,b; Supplementary Table 4), and then tested whether the insulin-producing human α-cells ameliorate the clinical signs of mice made diabetic. Glucose tolerance and GSIS were improved in mice receiving the maximum amount of αPM pseudoislets that we could generate from one single donor (between 200 and 1,000) (Exp. #2: Extended Data Fig. 6a–k; Supplementary Table 5). Yet hyperglycemia was not normalized, probably because of the suboptimal quantity of engrafted pseudoislets. We therefore transplanted αPM pseudoislets from multiple non-diabetic-donors (Exp. #3: 6,000 αPM pseudoislets from 6 donors and Exp. #4: 4,000 αPM pseudoislets from 3 donors; Fig. 2a-c; Supplementary Table 6). Diabetic mice became normoglycemic (Fig. 2b,c) and body weight loss was curbed, like controls receiving intact islets (4,000 IEQ; Extended Data Fig. 6l). Glucose tolerance and GSIS were similar in mice transplanted with either reprogrammed α-cells or intact islets (Fig. 2d; Extended Data Fig. 6m,n). Hyperglycemia reappeared upon graft removal, proving that recovery was induced by the engrafted insulin-secreting human αPM.
While αGFP pseudoislet grafts displayed no insulin production (<1%; Extended Data Fig. 5c: Exp. #1), most α-cells in αPM grafts contained insulin (67.9±1.9%; Fig. 2f,g) and became monohormonal (91.7±9.7%; Fig. 2f,h), indicating that the in vivo environment fosters cell conversion and maturation (Exp. #1: Extended Data Fig. 5d–g). Transplanted αPM pseudoislets were vascularized and innervated (Extended Data Figs. 5e, 6o), which correlates with their functionality, without cell proliferation (Extended Data Fig. 6p). They contained abundant GFP/insulin-coexpressing cells up to 6 months, the longest period analyzed (68.4%: reprogramming efficiency; Fig. 2i; Exp. #2, Extended Data Fig. 6q,r). In summary, insulin-producing human α-cells retain their phenotype in vivo, restoring normoglycemia in diabetic mice.
α-cells from T2D patients display distinctive transcriptomic signatures 25. We then explored whether T2D α-cells can engage in regulated insulin secretion in culture and in vivo (Exp. #5; Extended Data Fig. 7, Supplementary Table 7). In transplanted mice, random-fed glycemic levels became lower, with complete recovery when the mouse was given anti-glucagon therapy 26,27 to compensate for the insufficient number of transplanted cells (Extended Data Fig. 7f,g). We also observed improved glucose tolerance and GSIS (Fig. 2e; Extended Data Fig. 7h–j). Interestingly, the engrafted T2D αPM pseudoislets retained ARX expression while endogenous PDX1, MAFA and NKX6.1 were upregulated (Extended Data Fig. 7o). At the ultrastructural level, T2D αPM cells contained β-like dense-core granules (Extended Data Fig. 7p). Apoptotic cells were rare (Extended Data Fig. 7q). In conclusion, human α-cells in T2D conditions preserve the plasticity potential.
Hybrid signature of reprogrammed α-cells
To characterize the insulin-producing α-cells (Fig. 1f), we performed bulk mRNA sequencing (RNA-Seq) on sorted α- and β-cells, on αGFP, αPM, and βGFP pseudoislets after 7 days of culture, and on grafted αPM pseudoislets 1 month post-transplantation (Supplementary Table 8). Principal component and correlation analyses revealed that αPM cells display a signature intermediate between α- and β-cells, further shifted toward that of β-cells after transplantation into mice (Fig. 3a,b).
To identify α-cell identity changes linked to β-cell phenotype acquisition, we focused on differentially-expressed genes (DEGs) between sorted α- and β-cells. 587 DEGs were more abundant in sorted β-cells compared to α-cells; we termed them “β-cell-related” genes (Fig. 3c, Supplementary Table 10) and measured how they were modulated in α-cells upon aggregation or Pdx1/MafA coexpression (“aggregation effect” and “PM effect”, respectively). Cell aggregation led to the upregulation in α-cells of 128 β-related genes (out of 587) (Fig. 3d; Supplementary Table 11), indicating that mere aggregation is sufficient to impose functional aspects of the β-cell signature. Some of the upregulated genes are crucial to β-cell function, such as: ABCC8 (SUR1), ENTPD3 28, SYT13 29 and UCHL1 30. Likewise, other gene-sets involved in β-cell function, such as mitochondrial metabolism genes, were upregulated, while stress/inflammatory signaling pathways were downregulated upon α-cell aggregation (Extended Data Fig. 8a–c).
To elucidate the Pdx1/MafA coexpression effect on α-cell reprogramming, we compared αGFP and αPM pseudoislets. 115 β-related genes were upregulated in αPM, comprising important functional β-cell markers like INS, MAFA (endogenous), PSCK1, ADCYAP131, PFKFB232, RBP133 and SIX334 (Fig. 3e, Supplementary Table 12). Lastly, we evaluated the overall effect of combined aggregation and PM effect by comparing αPM pseudoislets with sorted α-cells: in this comparison, 268 β-related genes were upregulated in αPM pseudoislets, including IAPP and GLP1R (Fig. 3f, Supplementary Table 13). Gene-sets such as “hallmark β-cells” and “regulation of insulin secretion” were also enriched (Fig. 3g, Extended Data Fig. 8d). Moreover, some α-cell-enriched genes were downregulated (Extended Data Fig. 8e–h, Supplementary Table 14–17), while others were enhanced. These contrasting changes in α-identity genes suggest the existence of a simultaneous ongoing refractory response to fate conversion 35 (Supplementary Table 18). Consistent with this, ARX expression remained invariable in reprogrammed α-cells, incidentally confirming that GSIS is compatible with ARX activity (Extended data Fig. 3d).
Global proteomic analyses confirmed that many of the upregulated “β-genes” were also upregulated as proteins. 33 β-cell-related proteins were detected in α-cells after aggregation, Pdx1/MafA coexpression, or both (Fig. 3d-f,h). Most (30/33) were more abundant in αPM pseudoislets than in αGFP pseudoislets. Several functional β-cell proteins were upregulated, such as INS, PCSK1, PFKFB2, RBP1, ADCYAP1, UCHL1, ABCC8, ENTPD3 and IAPP (Fig. 3h, Supplementary Table 19). Interestingly, the protein hierarchical clustering grouped αPM pseudoislets with βGFP pseudoislets, underscoring the phenotypic shift (Extended Data Fig. 8i).
We also performed RNA-Seq on 5 different αPM grafts harvested 1 month after transplantation (Supplementary Table 8). Compared to αPM pseudoislets before transplantation, β-cell markers such as IGF2, MEG3 and GLRA1 and pathways like “hormone synthesis”, “secretion” and “innervation” were further upregulated (Extended Data Fig. 8j; Supplementary Table 20). These observations are compatible with the immunofluorescence results, the highly functional phenotype following transplantation (Fig. 2b-h), and the sympathetic innervation of αPM pseudoislets (Extended Data Fig. 6o).
Combined, transcriptomic and proteomic analyses indicate that aggregation drives the upregulation of functional β-cell markers. Subsequent Pdx1/MafA expression further induces additional β-cell genes, leading to GSIS acquisition (Extended Data Fig. 8k).
From glucagon to insulin secretion
To dissect the reprogramming process at single-cell resolution, we performed single-cell RNA sequencing (scRNA-Seq; Extended Data Fig. 9a). We analysed αGFP, βGFP and αPM pseudoislets cultured for 1 week. T-distributed stochastic neighbor embedding (t-SNE) visualization showed the 3 distinct αGFP, αPM and βGFP cell populations (Fig. 4a). Most cells in the αPM cluster were INS/GCG bihormonal, with little heterogeneity (Extended Data Fig. 9b).
Using the pseudotemporal ordering 36 algorithm, we reconstructed the sequence of gene expression profiles and the evolution of each cell without prior knowledge of the genes defining progression (Extended Data Fig. 9a). We found one main path, with few minor branches, enabling the allocation of 3 pseudotime-dependent progression states for αPM cells: “early”, “mid” and “late” (Fig. 4b and Extended Data Fig. 9c,d).
β-cell-related genes were upregulated in “late” cells (INS, UCHL1 and PCSK1; Fig.4c-e), while many α-cell-related genes were downregulated (GCG and TM4SF4). Several of these genes were not previously detected by bulk RNA-Seq (e.g. β-cell genes like HIST3H2A, NEFL, NPTX2 and SUSD4, and the α-cell genes FAP, EGFL7 and USH1C). Some α-/β-cell-related genes were not modulated upon pseudotime progression (ARX; Extended Data Fig. 9g) while others changed in the opposite direction (NR4A2; Fig. 4c). This confirms the persistence of some resistance to reprogramming, which was also observed by bulk RNA-Seq and in mouse studies 35.
When superimposing pseudotime categories on cells in t-SNE mapping, we found a pseudotemporal transition along the α-to-β-cell progression (Fig. 4f). Collectively, this suggests that most α-cells in αPM pseudoislets engage as INS-expressers, without alternative cell-fate allocations. Therefore, human α-cells are suited to direct reprogramming.
Eventually, signaling pathway analyses also revealed changes in the different reprogramming stages: oxidative phosphorylation was more active in “late” than in “early” cells, and RICTOR was downregulated as an upstream regulator (Supplementary Table 21). Interestingly, opposite changes (i.e. impaired oxidative phosphorylation and activated RICTOR) have recently been linked to β-cell “dedifferentiation” during β-cell failure 37. This suggests that human reprogrammed α-cells are healthy and different from T2D-diseased β-cells.
Insulin-secreting α-cells are hypoimmunogenic
The hybrid character of human insulin-producing α-cells, with their curative properties when transplanted to diabetic mice yet maintaining a robust α-cell identity, led us to inquire whether they would be autoimmune targets in type 1 Diabetes (T1D). We performed cytotoxic T lymphocyte (CTL) killing assays using HLA-A2-restricted β-cell-specific CD8+ T-cell clones derived from patients with recent T1D onset 38,39 (Extended Data Fig. 10a). As target cells, βGFP, αGFP and αPM pseudoislets devoid of HM cells were generated from 7 different HLA-A2 haplotype donors (Supplementary Table 22). We validated the specificities and cytotoxic properties of effector cells (CTL clones) in each independent experiment (Extended Data Fig. 10b,c) and on β-cells obtained from βGFP pseudoislets. Preproinsulin (PPI)-directed CTLs killed βGFP cells 39, yet the stressed β-cell-specific anti-DRiP (defective ribosomal product) CTLs did not, implying that β-cells in pseudoislets are not ER-stressed and have no erroneous INS mRNA translation 38 (Fig. 5a). This contrasts with studies reporting that up to 40% β-cells from dispersed human islets are lysed by DRiP CTLs 38, and suggests that pseudoislets confer protection against ER-stress.
Glucagon-producing α-cells were killed by the β-cell-specific CTLs anti-PPI and anti-DRiP only if loaded with their cognate PPI or DRiP peptide epitopes (Fig. 5b), which confirms CTLs’ specificity. Insulin-producing α-cells were sensitive to PPI-directed CTLs, similar to β-cells, but not to DRiP-specific T-cells (Fig. 5c), suggesting that αPM-cells are less stressed than dispersed β-cells38. When loaded with PPI or DRiP peptide epitopes, αPM-cells were more sensitive to the respective CTLs, but with lower death rates as compared to PPI-loaded αGFPs.
In conclusion, glucose-responsive insulin-secreting human reprogrammed α-cells display reduced immunogenicity for T1D autoreactive T-cells. Modified α-cells produce and process insulin, and present its leader/signal peptide on their surface, like native β-cells. The intrinsic β-cell processing steps are operative, but appear to be less prone to ER stress and translational errors than in native β-cells.
Discussion
Breakthroughs in islet cell biology have recently revealed that cell identity and maturity are flexible states. Here we provide conceptual evidence for human islet plasticity, given applicability concerns of mouse data to the clinic 40. It remains to be seen if diabetes therapies modulating islet cell-type interconversion without side-effects are possible through targeted delivery of TFs into islets, or pharmacologically. The streamlined culture system described here should allow the identification of a gene-set required to confer GSIS to islet non-β-cells, perhaps involving small-molecule screening like that adapted for β-like cells differentiated from human pluripotent-cells.
The human fetal β-cell line EndoC-βH1, hES-derived surrogate β-cells 41,42, and reprogrammed murine α-cells 13 are also hypoimmunogenic. Future studies should involve testing the immunogenicity of modified α-cells in humanized mice modeling responses toward islet grafts in vivo.
Here we did not analyze human δ-cells, but in addition to human α- and γ-cells, Pdx1/MafA also reprogram mouse δ-cells (not shown). Interestingly, Nkx6.1 seems dispensable for reprogramming, but might be required for deeper β-cell-like maturation 43. Combined, this islet non-β-cell plasticity leads to a strategic paradigm shift from an “α-cell-specific” to a broader “islet-specific” targeting approach in vivo. Notably, islet non-β-cells (i) are developmentally and epigenetically close to β-cells 8,44, (ii) occupy the same environmental niche, (iii) spontaneously engage in insulin production 6,7 more rapidly and efficiently than other cell-types 1–4,45, even under disease conditions (α-cells from T2D donors), (iv) are abundant in T1 and T2 diabetic patients, and also (v) a massive α-cell loss has no major physiological effect 26,46. For these reasons, islet non-β-cells are optimal targets for β-cell regeneration. The development of islet cell-type interconversion therapies will imply that insulin-producing cells are replenished without the need of remaining β-cells, ex vivo ES/iPS cell differentiation, or other invasive procedures.
Beyond diabetes, the adaptive cell identity changes are probably a feature of different cell types in many organs. Therefore, reconstitution of missing cell populations by fostering the innate in situ adaptive cell plasticity arises as a promising prospective to treat degenerative diseases.
Methods
Human samples.
All studies involving human samples were approved by ethical committee in University of Geneva. Human pancreatic islets were obtained from ECIT program (University Hospital of Geneva or Diabetes Research Institute in Milan) or the NIDDK-funded Integrated Islets Distribution Program (IIDP) at City of Hope. Subject details are described in Supplementary Table 1. The number of islets was determined by material availability, especially in T2D donor samples. Human umbilical vein endothelial cells (HUVECs) and human bone marrow–derived mesenchymal stem cells (MSCs) were purchased from Lonza (catalog number: C2519A, PT-2501) and cultured according to manufacturer’s instructions.
Isolation of human islet cell-types.
Dissociation of human islets and staining with cell-surface antibodies were described previously 8,20. Stained cells were sorted on a FACSAria2 (BD Biosciences) or Moflo Astrios (Beckman Coulter) system. Single viable islet cells were gated by forward scatter, side scatter and pulse-width parameters and by negative staining for DAPI (D1306, Invitrogen) or DRAQ7 (B25595, BD Biosciences) to remove doublets and dead cells.
Purity of sorted cells.
For evaluation of cell purity, sorted islet cell fractions were immunostained for insulin, glucagon, somatostatin, pancreatic polypeptide and ghrelin (see Supplementary Data 1–3). Stained cells were examined with a confocal microscope (Leica TCS SPE). Only batches with high purity (> 90%) were used to analyze in following experiments (see Extended Data Fig. 1b and Supplementary Table 2). Evaluation of purity was also performed by qPCR and RNA-seq. The purity (Extended Data Fig. 1b,c) was calculated by the method as previously reported 8.
Cell-labeling.
Adenoviral vectors were produced and purified as described 47. Sorted cells were transduced with adenoviral vectors encoding either GFP only, Pdx1-IRES-GFP, MafA-IRES-GFP, Nkx6.1-IRES-GFP or in combination at 37°C in 5% CO2 incubator for 12 hours. After 3-times wash, cells were resuspended in BN-medium: Advanced DMEM/F12 (12634–010) supplemented with penicillin/streptomycin, 10 mM HEPES, 2 mM GlutaMAX (35050–038), 1x B27 (17504044), 1x N2 (17502048, Life Technologies), 10 mM Nicotinamide (N0636), and 1 mM N-acetyl-L-cysteine (A9165, Sigma).
Reconstitution of pseudoislets.
For reaggregation into pseudoislets, labeled islet cells (α-/β-/γ-cells) were seeded on 96-well ultra-low adherent culture plates for 7 days (1,000 cells/well). Every other day culture medium was changed. For reaggregation with HUVECs/MSCs (HM), both 400 HUVECs and 100 MSCs were seeded per well together with 1,000 islet cells in mixed culture-medium: 50% BN-medium + 50% EGM-2 medium (CC-3162, Lonza) at 37°C in 5% CO2 incubator.
Morphometric analyses of pseudoislets.
Seven days after culture, aggregates were handpicked and their diameter was measured using Leica M205FA binocular equipped with a Leica DFC360FX camera.
Single islet cell culture.
Sorted islet cells were seeded at single-cell density on chamber-slide wells (Sigma) coated Matrigel (356231, Corning) and cultured in BN-medium at 37°C in 5% CO2 incubator, and then assayed as described below.
In vitro time-lapse imaging.
For live imaging of cultured cells, images of cultured cells in plate-wells were captured manually at indicated time-points using Nikon Eclipse TE300 microscope (Nikon).
Evaluation for transduction efficiency.
One or two weeks after reaggregation culture, pseudoislets were dissociated again into single cells using Accutase (A1110501, Life Technologies) and FACS-analyzed to count GFP+ cells. Doublets and dead cells were removed as mentioned above.
In vitro glucose-stimulated C-peptide secretion test.
Pseudoislets or size-matched native islets were handpicked for each assay replicate and washed by incubation for 30 min at 37°C in Krebs-Ringer Bicarbonate buffer (KRB) containing no glucose and then equilibrated by incubation for 1 hour at 37°C in basal KRB containing 3 mM glucose (Sigma). Samples were then transferred into fresh KRB containing 3 mM (Low) glucose for 1 hour followed by incubation for another hour in KRB containing 20 mM (Hi) glucose at 37°C. Medium was collected after 1-hour incubation at each glucose concentration and stored at −80°C for subsequent analyses. Human C-peptide concentration was quantified using human Ultrasensitive C-peptide ELISA kit (10–1141-01, Mercodia).
Total RNA extraction and quantitative RT-PCR.
For all samples, the total RNA was extracted using the Qiagen RNeasy Micro kit, and treated with RNase-free DNaseΙ (Qiagen). cDNA was synthesized using the QuantiTect RT kit (Qiagen). qPCR reactions and analyses were performed as described previously 6. Each individual sample was run in triplicate. Expression levels were normalized to RN18S, ACTB or GFP. Primer sequences are shown in Supplementary Table 23.
Streptozotocin-mediated diabetes model.
130 mg/kg Streptozotocin (S0130, Sigma) was administrated by intraperitoneal injection into male NSG mice (2–4-month-old). 7 days after injection, mice exhibiting hyperglycemia (> 20 mM) were used in subsequent experiments.
RIP-DTR mice.
5 ng/20g body weight diphtheria toxin (DT) (D0564, Sigma) was administrated by intraperitoneal single injection into male, and 3-times injections into female NSG RIP-DTR mice (2–4-month-old). 7–14 days after injection, mice exhibiting hyperglycemia (> 20 mM) were used in subsequent experiments.
Transplantations into the kidney.
Islet transplantations under the kidney capsule were performed as described 48. Pseudoislets or native islets were handpicked and mixed with Matrigel, and then transplanted under the capsule of kidney in immunodeficient NSG or NSG RIP-DTR mice, using a Leica M205FA stereomicroscope. The left kidney was selected in the transplantation of Exp. #1, #2, #4 and #5, but both sides of kidneys were used in transplantation of Exp. #3. Control non-grafted animals underwent sham-operations.
Glucose tolerance tests.
Mice were fasted for 16 hours before starting experiments. Intraperitoneal glucose tolerance test (ipGTT) was performed as described 6.
Nephrectomy.
For graft removal, the kidney with graft was ligated at the renal hilum using 3–0 silk (B. Braun), and then resected. Removed grafts were processed for analyses. Nephrectomy was also performed in control animals.
Anti-glucagon receptor antibody (GCGR-Ab) treatment.
As described previously 26, anti-GCGR monoclonal antibody A-9 was generated at Eli-Lilly and Company (Yan H, Hu S-FS, Boone TC, Lindberg RA, inventors; Amgen Inc., assignee. Compositions and methods relating to glucagon receptor antibodies. United States patent US 8158759 B2, 2012 Apr 17). It was delivered using a subcutaneous implanted osmotic pump (model 2002, Alzet) containing 3 mg/ml of anti-GCGR mAb in PBS for 2 weeks.
Immunofluorescence analyses.
Samples processes for immunofluorescence were performed as described previously 6. Frozen sections were cut at 10 μm-thick. Primary antibodies used were: guinea pig anti-porcine insulin (A0564, DAKO, 1:600), chicken anti-insulin (GW10064F, Sigma, 1:1000), mouse anti-glucagon (G2654, Sigma, 1:1000), rabbit anti-glucagon (A0565, DAKO, 1:600), rabbit anti-somatostatin (A0566, DAKO, 1:600), rabbit anti-pancreatic polypeptide (T-4088, PenLabs, 1:750), goat anti-ghrelin (sc-10368, SantaCruz, 1:200), rabbit anti-GFP (Life Technologies, 1:500), chicken anti-GFP (ab-13970, Abcam, 1:500), guinea pig anti-Pdx1 (gift from C. Wright, 1:1000), rabbit anti-MafA (A300–611A, Bethyl, 1:500), rabbit anti-Nkx6.1 (BCBC, 1:400), rabbit anti-pHH3 (06–570, Upstate, 1:500), rabbit anti-CD31 (ab28364, abcam, 1:50), rabbit anti-Vimentin (ab92547, abcam, 1:100), rabbit anti-Tyrosine hydroxylase (ab152, chemicon, 1:1000), guinea pig anti-ARX(AB2834, BCBC, 1:100), rabbit anti-Synaptophysin (A0010, DAKO, 1:50). Secondary antibodies were coupled to Alexa 405, 488, 568, 647 (Life Technologies), FITC, Cy3, Cy5 (Jackson Immunoresearch), or TRITC (Southern Biotech). Sections were counterstained with DAPI. All sections were examined with a confocal microscope (Leica TCS SPE). In Fig. 2i, confocal tile-scan images were merged as a maximum projection.
TUNEL staining.
TUNEL staining was performed to evaluate apoptosis using DeadEnd Fluorometric TUNEL System, according to manufacturer’s instructions (G3250, Promega).
Electron microscopy.
Small portion of engrafted islets were fixed with 2.5% glutaraldehyde and 4% paraformaldehyde in PBS, and processed as described 6. Morphometric analyses were performed using Philips/FEI Tecnai 20 transmission electron microscope.
Global transcriptomics analysis.
Preparation of libraries, RNA sequencing and the quality controls were performed within the Genomics Core Facility of the University of Geneva. In brief, extracted RNA samples were assessed for the quality by Agilent bioanalyzer prior to library generation. Reverse transcription and cDNA amplification were performed using the SMARTer Ultra Low RNA kit (Clontech). cDNA libraries were prepared using Nextera XT DNA Sample Preparation kit (Illumina), multiplexed and sequenced on an Illumina HiSeq4000 platform with single-end 110-bp reads. The sequencing quality control was done with FASTQC v.0.11.2, followed by sequence alignment to the human reference genome (hg38) using the TopHat v2.0.13 (default parameters). Biological quality control and summarization were done with the PicardTools v1.80. Finally, 13 million reads per samples in average were used for differential expression analysis.
Transcriptomic data analyses:
The normalization and differential expression analysis was performed with the R/Bioconductor package TCC v.1.16.0 using DEG elimination strategy 49. DEGs between sorted α- and β-cells were selected (FC>2 FDR<0.05) and their expression levels in the different other comparisons was considered for further analyses. PCA sample correlation analysis was performed with the 1,000 most variable genes. The Pearson correlation coefficients were calculated for log2 transformed ratios. The output data are displayed graphically as either a PCA-plot, heatmap, dendrogram, volcano-plots or Venn diagrams.
Pathway analysis:
The pathway analyses were performed with gene set enrichment analysis (GSEA, http://software.broadinstitute.org/gsea/index.jsp) or Ingenuity Pathway Analysis (IPA®, QIAGEN Redwood City, www.qiagen.com/ingenuity). Gene sets with significant enrichment in GSEA were identified among C1, C2 or C5 of Molecular Signatures Database v6.0; “HALLMARK PANCREAS BETA CELL” (http://software.broadinstitute.org/gsea/msigdb/cards/HALLMARK_PANCREAS_BETA_CELLS.html) and “REACTOME REGULATION OF INSULIN SECRETION” (www.reactome.org/content/detail/422356) in Figure 4, “GO OXIDATIVE PHOSHPRYLATION”(GO:0006119) and “GO RESPIRATORY CHAIN” (GO:0070469) in Extended Data Fig. 8a,b. IPA were performed with the following settings: Expression Value Type (Exp Log Ration), Reference set (Ingenuity Knowledge Base), Relationships to consider (Direct and Indirect Relationships), Interaction networks (70 molecules/network; 25 networks/analysis or 30 molecules/network; 25 or 10 network/analysis), Data Source (all), Confidence (Experimentally Observed), Species (Human, Mouse, Rat), Tissue & Cell Lines (all), Mutations (all).
Global proteomics analysis
Cell lysis and protein digestion:
Cells were washed in DPBS and lysed in buffer containing 4% SDS, and boiled at 95°C for 7 min on a shaker, and sonicated (three rounds a 30 sec, 30% power). The protein concentration was determined using a BCA protein assay kit. Samples were pooled (Supplementary Table 12), and dry aliquots containing an estimated amount of 50 μg of proteins were further processed using Filter-Aided Sample Preparation50, and desalted using C18 Oasis™ Elution plates (Waters, Milford, MA).
Tandem Mass Tag (TMT) 11-plex labelling:
TMT reagents were re-suspended in ACN. Desalted peptides were re-suspended in 50 μL of 200 mM EPPS pH 8.5, 15 μL of ACN, and 5 μL of the TMT reagents were added to the respective peptide samples, gently vortexed, and incubated for 1.5 h at RT. To prevent unwanted labelling, the reaction was quenched by adding 5 μL of 5% hydroxylamine and incubated for 15 min at RT. Equal amounts of the TMT-labelled samples were combined and concentrated to near dryness, followed by desalting via C18 solid phase extraction and passage over a Pierce detergent removal spin column (Thermo Fisher Scientific).
Off-line basic pH reversed phase fractionation:
The combined labelled peptide samples were pre-fractionated by basic pH reversed phase HPLC as described previously51, using an Agilent (P/N 770995–902) 300Extend-C18, 5 μm, 250 mm x 4.6 mm id column, connected to an Agilent Technology off-line LC-system. Solvent A was 5% ACN, 10 mM NH4HCO3 pH8, and solvent B was 90% ACN, NH4HCO3 pH 8. The samples were re-suspended in 500 μL solvent A and loaded onto the column. Column flow was set to 0.8 mL/min and the gradient length was 70 min, as follows: from 0–35 min solvent 50% A/ 50% B, and from 35–50 min 100% B, and from 50–70 min 100% A. The labelled peptides were fractionated into 96 fractions, and further combined into a total of 12 fractions. Each fraction was acidified with 1% formic acid, concentrated by vacuum centrifugation to near dryness, and desalted by StageTip. Each fraction was dissolved in 5% ACN/ 5% formic acid for LC-MS/MS analysis.
LC-MS3 analysis:
From each of the 12 fractions, ∼3 μg was dissolved in 1% aqueous formic acid (FA) prior to LC-MS/MS analysis on an Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific, San Jose, CA) coupled to a Proxeon EASY-nLC 1000 liquid chromatography (LC) pump (Thermo Fisher Scientific). Peptides were fractionated on a 75-μm inner diameter microcapillary column packed with ∼35 cm of Accucore resin (2.6 μm, 150 Å, Thermo Fisher Scientific, San Jose, CA). For each analysis, we loaded ~1 μg onto the column. Peptides were separated using a 2.5 hr gradient of 2 to 25% ACN in 0.125% formic acid at a flow rate of ∼350 nL/min. Each analysis used the multi-notch MS3-based TMT method52 on an Orbitrap Fusion mass spectrometer, which has been shown to reduce ion interference compared to MS2 quantification53. The scan sequence began with an MS1 spectrum (Orbitrap analysis; resolution 120,000; mass range 400−1400 m/z; automatic gain control (AGC) target 5 × 105; maximum injection time 100 ms). Precursors for MS2/MS3 analysis were selected using a Top10 method. MS2 analysis consisted of collision-induced dissociation (quadrupole ion trap analysis; AGC 2 × 104; normalized collision energy (NCE) 35; maximum injection time 200 ms). Following acquisition of each MS2 spectrum, we collected an MS3 spectrum using our recently described method52 in which multiple MS2 fragment ions were captured in the MS3 precursor population using isolation waveforms with multiple frequency notches. MS3 precursors were fragmented by high-energy collision-induced dissociation (HCD) and analysed using the Orbitrap (NCE 65; AGC 2 × 105; maximum injection time 300 ms, resolution was 50,000 at 400 Th).
Protein analysis:
Mass spectra were processed using a Sequest-based in-house software pipeline54, and spectra were converted to mzXML using a modified version of ReAdW.exe. Database searching included all entries from the human uniprot database (March 11, 2014). This database was concatenated with one composed of all protein sequences in the reversed order. Searches were performed using a 50 ppm precursor ion tolerance for total protein level analysis. The product ion tolerance was set to 0.9 Da. These wide mass tolerance windows were chosen to maximize sensitivity in conjunction with Sequest searches and linear discriminant analysis54,55. TMT tags on lysine residues and peptide N termini (+229.163 Da) and carbamidomethylation of cysteine residues (+57.021 Da) were set as static modifications, while oxidation of methionine residues (+15.995 Da) was set as a variable modification.
Peptide-spectrum matches (PSMs) were adjusted to a 1% false discovery rate (FDR)56,57. PSM filtering was performed using a linear discriminant analysis, as described previously54, while considering the following parameters: XCorr, ΔCn, missed cleavages, peptide length, charge state, and precursor mass accuracy. For TMT-based reporter ion quantitation, we extracted the summed signal-to-noise (S/N) ratio for each TMT channel and found the closest matching centroid to the expected mass of the TMT reporter ion.
The search space for each reporter ion was limited to a range of 0.003 Th to prevent overlap between the isobaric reporter ions. For protein-level comparisons, PSMs were identified, quantified, and collapsed to a 1% peptide FDR and then collapsed further to a final protein-level FDR of 1%. Moreover, protein assembly was guided by principles of parsimony to produce the smallest set of proteins necessary to account for all observed peptides.
Proteins were quantified by summing reporter ion counts across all matching PSMs using in-house software, as described previously54. PSMs with poor quality, MS3 spectra with more than eight TMT reporter ion channels missing, MS3 spectra with TMT reporter summed signal-to-noise ratio that is less than 100, or no MS3 spectra were excluded from quantitation58. Protein quantitation values were exported for further analysis in Microsoft Excel. Each reporter ion channel was summed across all quantified proteins and normalized assuming equal protein loading of all 11 samples.
Proteomic data analysis:
Quantitative analyses of protein expression was performed based on normalized TMT ratios.
Single-cell RNA-Seq.
Pseudoislets of αGFP, αPM and βGFP cultured for 1 week were dissociated into single-cells, and single, viable and GFP+ cells were sorted on a Moflo Astrios (Beckman Coulter) system. Sorted single-cells from αGFP, αPM and βGFP were loaded as separated samples for single-cell RNA-seq using the Chromium Controller (10xGenomics) and the Single Cell 3′ Library Kit v2 (PN-120236, PN-120237, PN-120262) according to the manufacturer’s protocol. Amplified cDNA and subsequent libraries were assessed for quantity and quality on an Agilent 2100 Bioanalyzer (Agilent Technologies). Libraries were sequenced as 100-bp paired-end reads on a HiSeq 4000 platform (Ilumina). The Cell Ranger software pipeline (ver.2.1, 10xGenomics) was used to demultiplex cellular barcodes, map reads to the transcriptome using the STAR aligner to produce a sparse cell/gene matrix.
Further analyses were performed using the R package Seurat (ver.2.3.2) 59. As QC steps, cells were used for analysis if they passed a total count threshold of 5,000 (αGFP), 10,000 (βGFP) or 12,000 (αPM) counts, an expressed genes threshold of 2,000 genes and had a lower percentage of mitochondrial counts than 18% (αPM and βGFP) or 20% (αGFP). Finally, cells were removed that expressed more than 100 counts of PPY. Genes were kept if they were expressed with at least 4 counts in at least 2 cells. After applying these QC criteria, 532 single cells and 5,092 genes in total remained and were included in downstream analyses. Differential expression analysis was performed using the MAST test.
To perform pseudotime analysis of αPM cells, we used Monocle2 60. Sparse matrix files were imported directly into Monocle using the R package cellrangerRkit, and QC steps were also performed in Monocle to remove low-quality cells and genes: cells were kept if they had at least 13,500 counts and 3,000 genes. Cells were removed if they expressed more than 250 counts PPY. Genes were kept if they were expressed. To generate pseudotime trajectories, the ‘DDRTree’ reduction method was used in Monocle2 with the default parameters. Cells ordered in the pseudotime created 10 distinct states (Extended Data Fig. 9). Cells belonging to state 1 were designated as “early αPM”, cells belonging to state 9 were designated as “late αPM” and all other cells were designated as “mid αPM”. Cellular identities belonging to each state were recovered from Monocle and “early αPM”, “mid αPM” and “late αPM” states were added to the corresponding cells as metadata in Seurat. Differential expression between states, αGFP and βGFP were calculated in Seurat.
Immunogenicity tests.
Human islets from non-diabetic donors with matching HLA-A2 haplotype were obtained from ECIT or IIDP (see Supplementary Tables 1, 2 and 22). FACS-sorting, transduction and reaggregation were performed as mentioned above. Monotypic pseudoislets were harvested 1–2 weeks after aggregation in Geneva. Then we brought the pseudoislets to LUMC in Netherland to perform the CTL killing assay, as previously described 38,39. Briefly, after dispersion with Accutase, pseudoislet cells were labeled with 51Cr and then cocultured with the following specific CTL clones: CMVpp65 (“CMV”, a T-cell clone recognizing cytomegalovirus-specific antigen as negative control), alloreactive T-cell clone JS132 (“HLA-A2”, as positive control), INS-DRiP1–9 (“DRiP”, a T-cell clone killing stressed β-cells producing INS-DRiP1–9 peptide, a defective ribosomal product derived from aberrant insulin transcript translation38) or PPI15–24 (“PPI”, a T-cell clone that kills β-cells presenting preproinsulin signal peptide15–24 in physiologic conditions. This unconventionally processed β-cell epitope is more prominently presented in hyperglycemic conditions and T1D patients 39). Target cell lysis was determined by measuring 51Cr release with γ-counter. The specificities and cytotoxic properties of CTL clones were evaluated in each independent experiment (Extended Data Fig. 10). In some control conditions, islet cells were loaded with either INS-DRiP1–9 (DRiP) or preproinsulin (PPI) peptide epitopes.
Animal experiments.
NOD.Cg-PrkdcscidII2rgtm1Wj/SzJ (abbreviated as NSG) mice were obtained from Charles River. NSG RIP-DTR mice were generated in the Jackson Laboratory by backcrossing the RIP-DTR allele from B6-background animals (Hprttm1(Ins2-HBEGF)Herr)6. All mice were treated in accordance with the guidelines and regulation of the Direction Générale de la Santé, state of Geneva (license number GE/103/14).
Statistical analyses.
No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment. All statistical analyses were performed using GraphPad Prism 7 software. All error bars in figures represent standard error of mean (s.e.m.) or standard deviation (SD) as indicated in the legends. p-values and statistical methods were described in figure legends. The number of samples in this study was limited by the availability of human donor samples, especially T2D donors. The Healthy NSG mice used in Extended Data Fig. 5 were of both genders, but only male NSG mice were used in STZ-diabetic model in Extended Data Fig. 6–8.
Extended Data
Supplementary Material
Acknowledgements
We are grateful to Roland Stein for carefully reading the manuscript, and constructive comments and suggestions. We thank C. Gysler for technical help, J-P. Aubry-Lachainaye for FACS assistance, C. Delucinge-Vivier and M. Docquier for RNA-seq. We thank Q. Zhou for viral vectors, R. Millican and P. Cain for anti-GCGR antibody, and R.Nano and L.Piemonti for human donor samples. Human islets were provided through the JDRF award 31–2008-416 (ECIT Islet for Basic Research program) or the NIDDK-funded Integrated Islet Distribution Program (IIDP) at City of Hope, National Institutes of Health (NIH) Grant no. DK098085. This work was funded with grants from the Research Council of Norway (NFR 247577) and the Novo Nordisk Foundation (NNF15OC0015054) to S.C.; NIH/NIDDK grant DK098285 to J.A.P.; Bergen Forskningsstiftelse (BFS2014REK02) and the Western Norway Regional Health Authority (Bergen Stem Cell Consortium) and the Novo Nordisk Foundation (NNF17OC0027258) to H.R.; NIH/NIDDK (Human Islet Research Network, DK104209 and DK108132), the Juvenile Diabetes Research Foundation (SRA-2015–67-Q-R), the Fondation Privée des HUG – Confirm Award, the Fondation Aclon, and the Swiss National Science Foundation (NRP63 no. 406340–128056, no. 310030_152965 and the Bonus of Excellence grant no. 310030B_173319) to P.L.H.
Footnotes
Data availability
RNA-Seq data that support the findings of this study have been deposited in NCBI’s Gene Expression Omnibus and accessible through GEO series accession code: GSE117454 (bulk RNA-Seq) and GSE123844 (scRNA-Seq). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD011933.
Competing interests Oregon Health & Science University has commercially licensed one of the antibodies described herein (HIC1–2B4/HPi2); C.D. and M.G. are inventors of this antibody. This potential conflict of interest has been reviewed and managed by OHSU.
References
- 1.Mauda-Havakuk M et al. Ectopic PDX-1 expression directly reprograms human keratinocytes along pancreatic insulin-producing cells fate. PLoS One 6, e26298, doi: 10.1371/journal.pone.0026298 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Zhu S et al. Human pancreatic beta-like cells converted from fibroblasts. Nat Commun 7, 10080, doi: 10.1038/ncomms10080 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Lima MJ et al. Suppression of epithelial-to-mesenchymal transitioning enhances ex vivo reprogramming of human exocrine pancreatic tissue toward functional insulin-producing beta-like cells. Diabetes 62, 2821–2833, doi: 10.2337/db12-1256 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Lee J et al. Expansion and conversion of human pancreatic ductal cells into insulin-secreting endocrine cells. Elife 2, e00940, doi: 10.7554/eLife.00940 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Chera S & Herrera PL Regeneration of pancreatic insulin-producing cells by in situ adaptive cell conversion. Curr Opin Genet Dev 40, 1–10, doi: 10.1016/j.gde.2016.05.010 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Thorel F et al. Conversion of adult pancreatic alpha-cells to beta-cells after extreme beta-cell loss. Nature 464, 1149–1154, doi: 10.1038/nature08894 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Chera S et al. Diabetes recovery by age-dependent conversion of pancreatic delta-cells into insulin producers. Nature 514, 503–507, doi: 10.1038/nature13633 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bramswig NC et al. Epigenomic plasticity enables human pancreatic alpha to beta cell reprogramming. J Clin Invest 123, 1275–1284, doi: 10.1172/JCI66514 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Spijker HS et al. Conversion of mature human beta-cells into glucagon-producing alpha-cells. Diabetes 62, 2471–2480, doi: 10.2337/db12-1001 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Yoneda S et al. Predominance of beta-cell neogenesis rather than replication in humans with an impaired glucose tolerance and newly diagnosed diabetes. J Clin Endocrinol Metab 98, 2053–2061, doi: 10.1210/jc.2012-3832 (2013). [DOI] [PubMed] [Google Scholar]
- 11.White MG et al. Expression of mesenchymal and alpha-cell phenotypic markers in islet beta-cells in recently diagnosed diabetes. Diabetes Care 36, 3818–3820, doi: 10.2337/dc13-0705 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Butler AE et al. Marked expansion of exocrine and endocrine pancreas with incretin therapy in humans with increased exocrine pancreas dysplasia and the potential for glucagon-producing neuroendocrine tumors. Diabetes 62, 2595–2604, doi: 10.2337/db12-1686 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Xiao X et al. Endogenous Reprogramming of Alpha Cells into Beta Cells, Induced by Viral Gene Therapy, Reverses Autoimmune Diabetes. Cell Stem Cell 22, 78–90 e74, doi: 10.1016/j.stem.2017.11.020 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Ben-Othman N et al. Long-Term GABA Administration Induces Alpha Cell-Mediated Beta-like Cell Neogenesis. Cell 168, 73–85 e11, doi: 10.1016/j.cell.2016.11.002 (2017). [DOI] [PubMed] [Google Scholar]
- 15.Li J et al. Artemisinins Target GABAA Receptor Signaling and Impair alpha Cell Identity. Cell 168, 86–100 e115, doi: 10.1016/j.cell.2016.11.010 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Zhou Q & Melton DA Pancreas regeneration. Nature 557, 351–358, doi: 10.1038/s41586-018-0088-0 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.van der Meulen T et al. Artemether Does Not Turn alpha Cells into beta Cells. Cell Metab 27, 218–225 e214, doi: 10.1016/j.cmet.2017.10.002 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Zhu Y, Liu Q, Zhou Z & Ikeda Y PDX1, Neurogenin-3, and MAFA: critical transcription regulators for beta cell development and regeneration. Stem Cell Res Ther 8, 240, doi: 10.1186/s13287-017-0694-z (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Matsuoka TA et al. Mafa Enables Pdx1 to Effectively Convert Pancreatic Islet Progenitors and Committed Islet alpha-Cells Into beta-Cells In Vivo. Diabetes 66, 1293–1300, doi: 10.2337/db16-0887 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Dorrell C et al. Human islets contain four distinct subtypes of beta cells. Nat Commun 7, 11756, doi: 10.1038/ncomms11756 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Bosco D, Orci L & Meda P Homologous but Not Heterologous Contact Increases the Insulin-Secretion of Individual Pancreatic B-Cells. Experimental Cell Research 184, 72–80, doi:Doi 10.1016/0014-4827(89)90365-0 (1989). [DOI] [PubMed] [Google Scholar]
- 22.Takebe T et al. Vascularized and functional human liver from an iPSC-derived organ bud transplant. Nature 499, 481–484, doi: 10.1038/nature12271 (2013). [DOI] [PubMed] [Google Scholar]
- 23.Hematti P, Kim J, Stein AP & Kaufman D Potential role of mesenchymal stromal cells in pancreatic islet transplantation. Transplant Rev (Orlando) 27, 21–29, doi: 10.1016/j.trre.2012.11.003 (2013). [DOI] [PubMed] [Google Scholar]
- 24.Hilderink J et al. Controlled aggregation of primary human pancreatic islet cells leads to glucose-responsive pseudoislets comparable to native islets. J Cell Mol Med 19, 1836–1846, doi: 10.1111/jcmm.12555 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Segerstolpe A et al. Single-Cell Transcriptome Profiling of Human Pancreatic Islets in Health and Type 2 Diabetes. Cell Metab 24, 593–607, doi: 10.1016/j.cmet.2016.08.020 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Damond N et al. Blockade of glucagon signaling prevents or reverses diabetes onset only if residual beta-cells persist. Elife 5, doi: 10.7554/eLife.13828 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Neumann UH, Denroche HC, Mojibian M, Covey SD & Kieffer TJ Insulin Knockout Mice Have Extended Survival but Volatile Blood Glucose Levels on Leptin Therapy. Endocrinology 157, 1007–1012, doi: 10.1210/en.2015-1890 (2016). [DOI] [PubMed] [Google Scholar]
- 28.Syed SK et al. Ectonucleotidase NTPDase3 is abundant in pancreatic beta-cells and regulates glucose-induced insulin secretion. Am J Physiol Endocrinol Metab 305, E1319–1326, doi: 10.1152/ajpendo.00328.2013 (2013). [DOI] [PubMed] [Google Scholar]
- 29.Andersson SA et al. Reduced insulin secretion correlates with decreased expression of exocytotic genes in pancreatic islets from patients with type 2 diabetes. Mol Cell Endocrinol 364, 36–45, doi: 10.1016/j.mce.2012.08.009 (2012). [DOI] [PubMed] [Google Scholar]
- 30.Chu KY, Li H, Wada K & Johnson JD Ubiquitin C-terminal hydrolase L1 is required for pancreatic beta cell survival and function in lipotoxic conditions. Diabetologia 55, 128–140, doi: 10.1007/s00125-011-2323-1 (2012). [DOI] [PubMed] [Google Scholar]
- 31.Inagaki N, Kuromi H & Seino S PACAP/VIP receptors in pancreatic beta-cells: their roles in insulin secretion. Ann N Y Acad Sci 805, 44–51; discussion 52–43 (1996). [DOI] [PubMed] [Google Scholar]
- 32.Muller YL et al. A cis-eQTL in PFKFB2 is associated with diabetic nephropathy, adiposity and insulin secretion in American Indians. Hum Mol Genet 24, 2985–2996, doi: 10.1093/hmg/ddv040 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Kane MA et al. CrbpI modulates glucose homeostasis and pancreas 9-cis-retinoic acid concentrations. Mol Cell Biol 31, 3277–3285, doi: 10.1128/MCB.05516-11 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Arda HE et al. Age-Dependent Pancreatic Gene Regulation Reveals Mechanisms Governing Human beta Cell Function. Cell Metab 23, 909–920, doi: 10.1016/j.cmet.2016.04.002 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Cigliola V et al. Pancreatic islet-autonomous insulin and smoothened-mediated signalling modulate identity changes of glucagon(+) alpha-cells. Nat Cell Biol 20, 1267–1277, doi: 10.1038/s41556-018-0216-y (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Qiu X et al. Single-cell mRNA quantification and differential analysis with Census. Nat Methods 14, 309–315, doi: 10.1038/nmeth.4150 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Kim-Muller JY et al. Aldehyde dehydrogenase 1a3 defines a subset of failing pancreatic beta cells in diabetic mice. Nat Commun 7, 12631, doi: 10.1038/ncomms12631 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Kracht MJ et al. Autoimmunity against a defective ribosomal insulin gene product in type 1 diabetes. Nat Med 23, 501–507, doi: 10.1038/nm.4289 (2017). [DOI] [PubMed] [Google Scholar]
- 39.Skowera A et al. CTLs are targeted to kill beta cells in patients with type 1 diabetes through recognition of a glucose-regulated preproinsulin epitope. J Clin Invest 118, 3390–3402, doi: 10.1172/JCI35449 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Kulkarni RN & Stewart AF Summary of the Keystone islet workshop (April 2014): the increasing demand for human islet availability in diabetes research. Diabetes 63, 3979–3981, doi: 10.2337/db14-1303 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.van der Torren CR et al. Immunogenicity of human embryonic stem cell-derived beta cells. Diabetologia 60, 126–133, doi: 10.1007/s00125-016-4125-y (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.van der Torren C et al. Innate and adaptive immunity to human beta cell lines: implications for beta cell therapy. Diabetologia 59, 170–175, doi: 10.1007/s00125-015-3779-1 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Schaffer AE et al. Nkx6.1 controls a gene regulatory network required for establishing and maintaining pancreatic Beta cell identity. PLoS Genet 9, e1003274, doi: 10.1371/journal.pgen.1003274 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Herrera PL Adult insulin- and glucagon-producing cells differentiate from two independent cell lineages. Development 127, 2317–2322 (2000). [DOI] [PubMed] [Google Scholar]
- 45.Rezania A et al. Reversal of diabetes with insulin-producing cells derived in vitro from human pluripotent stem cells. Nat Biotechnol 32, 1121–1133, doi: 10.1038/nbt.3033 (2014). [DOI] [PubMed] [Google Scholar]
- 46.Thorel F et al. Normal glucagon signaling and beta-cell function after near-total alpha-cell ablation in adult mice. Diabetes 60, 2872–2882, doi: 10.2337/db11-0876 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Zhou Q, Brown J, Kanarek A, Rajagopal J & Melton DA In vivo reprogramming of adult pancreatic exocrine cells to beta-cells. Nature 455, 627–632, doi: 10.1038/nature07314 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Mathe Z et al. Tetracycline-regulated expression of VEGF-A in beta cells induces angiogenesis: improvement of engraftment following transplantation. Cell Transplant 15, 621–636 (2006). [DOI] [PubMed] [Google Scholar]
- 49.Sun J, Nishiyama T, Shimizu K & Kadota K TCC: an R package for comparing tag count data with robust normalization strategies. BMC Bioinformatics 14, 219, doi: 10.1186/1471-2105-14-219 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Wisniewski JR, Zougman A, Nagaraj N & Mann M Universal sample preparation method for proteome analysis. Nature methods 6, 359–362, doi: 10.1038/nmeth.1322 (2009). [DOI] [PubMed] [Google Scholar]
- 51.Paulo JA, O’Connell JD & Gygi SP A Triple Knockout (TKO) Proteomics Standard for Diagnosing Ion Interference in Isobaric Labeling Experiments. J Am Soc Mass Spectrom 27, 1620–1625, doi: 10.1007/s13361-016-1434-9 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.McAlister GC et al. MultiNotch MS3 Enables Accurate, Sensitive, and Multiplexed Detection of Differential Expression across Cancer Cell Line Proteomes. Analytical chemistry 86, 7150–7158, doi: 10.1021/ac502040v (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Paulo JA & Gygi SP A comprehensive proteomic and phosphoproteomic analysis of yeast deletion mutants of 14–3-3 orthologs and associated effects of rapamycin. Proteomics 15, 474–486, doi: 10.1002/pmic.201400155 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Huttlin EL et al. A tissue-specific atlas of mouse protein phosphorylation and expression. Cell 143, 1174–1189, doi: 10.1016/j.cell.2010.12.001 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Beausoleil SA, Villen J, Gerber SA, Rush J & Gygi SP A probability-based approach for high-throughput protein phosphorylation analysis and site localization. Nature biotechnology 24, 1285–1292, doi: 10.1038/nbt1240 (2006). [DOI] [PubMed] [Google Scholar]
- 56.Elias JE & Gygi SP Target-decoy search strategy for mass spectrometry-based proteomics. Methods Mol Biol 604, 55–71, doi: 10.1007/978-1-60761-444-9_5 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Elias JE & Gygi SP Target-decoy search strategy for increased confidence in large-scale protein identifications by mass spectrometry. Nat Methods 4, 207–214, doi: 10.1038/nmeth1019 (2007). [DOI] [PubMed] [Google Scholar]
- 58.McAlister GC et al. Increasing the multiplexing capacity of TMTs using reporter ion isotopologues with isobaric masses. Analytical chemistry 84, 7469–7478, doi: 10.1021/ac301572t (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Butler A, Hoffman P, Smibert P, Papalexi E & Satija R Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol 36, 411–420, doi: 10.1038/nbt.4096 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Qiu X et al. Reversed graph embedding resolves complex single-cell trajectories. Nat Methods 14, 979–982, doi: 10.1038/nmeth.4402 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.Masini M et al. Ultrastructural morphometric analysis of insulin secretory granules in human type 2 diabetes. Acta Diabetol 49 Suppl 1, S247–252, doi: 10.1007/s00592-012-0446-6 (2012). [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.