78 Diabetes Volume 67, January 2018

Mitochondrial UCP2 Controls Pancreas Development

Benjamin Broche,1,2,3 Selma Ben Fradj,1,2,3 Esther Aguilar,1,2,3 Tiphaine Sancerni,1,2,3,4 Matthieu Bénard,1,2,3 Fatna Makaci,1,2,3 Claire Berthault,1,2,3 Raphaël Scharfmann,1,2,3 Marie-Clotilde Alves-Guerra,1,2,3 and Bertrand Duvillié1,2,3

Diabetes 2018;67:78–84 | https://doi.org/10.2337/db17-0118

The (UCP) 2 be- belongs to the family of UCPs (2). Despite the well-accepted longs to the family of the UCPs. Despite its name, it is now role of UCP1 as a proton transporter and a UCP in the brown accepted that UCP2 is rather a metabolite transporter adipose tissue, it was shown that UCP2 is a metabolite trans- than a UCP. UCP2 can regulate oxidative stress and/or porter with no or little mitochondrial uncoupling activity (3). energetic metabolism. In rodents, UCP2 is involved in the UCP2 is expressed in the spleen, lungs, stomach, adipose control of a-andb-cell mass as well as insulin and glu- tissue, and pancreas (4,5). Moreover, several studies in- cagon secretion. Our aim was to determine whether the dicate that UCP2 is a repressor of b effects of UCP2 observed on -cell mass have an embry- (ROS)productionindifferentcelltypes(6,7).Inaddition, Ucp2 onic origin. Thus, we used knockout mice. We found UCP2 can regulate the balance between glycolysis and oxi- an increased size of the pancreas in Ucp22/2 fetuses at dative phosphorylation in murine embryonic fibroblasts (7) embryonic day 16.5, associated with a higher number of and in different types of cancer cells (8,9). Recently, Ucp2 a-andb-cells. This phenotype was caused by an increase mutations were discovered in humans and were associated of PDX1+ progenitor cells. Perinatally, an increase in the with congenital hyperinsulinism (10). In mice, the absence ISLET STUDIES proliferation of endocrine cells also participates in their expansion. Next, we analyzed the oxidative stress in the of UCP2 also leads to increased insulin secretion (11), sup- pancreata. We quantified an increased nuclear transloca- porting the observation in humans. The knockout of UCP2 tion of nuclear factor erythroid 2–related factor 2 (NRF2) in induces an increase in the number of endocrine cells, and fi the mutant, suggesting an increased production of reac- this phenotype is ampli ed by a high-fat diet (12,13). tive oxygen species (ROS). Phosphorylation of AKT, an The aim of our study was to determine whether the 2/2 ROS target, was also activated in the Ucp22/2 pancreata. b-cell hyperplasia observed in adult Ucp2 mice has an 2/2 Finally, administration of the antioxidant N-acetyl-L-cysteine embryonic origin. For this, we used Ucp2 mouse em- to Ucp22/2 pregnant mice alleviated the effect of knock- bryos at different stages and we analyzed the development ing out UCP2 on pancreas development. Together, these of the pancreas. data demonstrate that UCP2 controls pancreas develop- ment through the ROS-AKT signaling pathway. RESEARCH DESIGN AND METHODS Animals Experiments were in agreement with the French animal 2 2 During the last decade, the impact of mitochondrial dys- care committee guidelines. Ucp2 / mice (C57Bl6/J back- function in pancreatic islet development and diabetes has ground) were previously described (14). N-acetyl-L-cysteine been widely studied (1). However the underlying mecha- (NAC) (Sigma-Aldrich, Saint-Quentin-Fallavier, France) treat- nisms involving the mitochondria are still not well under- ment was initiated at embryonic day 9.5 (E9.5) until E13.5, stood. The mitochondrial uncoupling protein (UCP) 2 and at E12.5 until E19.5.

1INSERM, U1016, Institut Cochin, Paris, France This article contains Supplementary Data online at http://diabetes 2CNRS, UMR8104, Paris, France .diabetesjournals.org/lookup/suppl/doi:10.2337/db17-0118/-/DC1. 3 Université Paris Descartes, Sorbonne Paris Cité, Paris, France © 2017 by the American Diabetes Association. Readers may use this article as 4 Université Paris Diderot, Sorbonne Paris Cité, Paris, France long as the work is properly cited, the use is educational and not for profit, and the Corresponding authors: Bertrand Duvillié, [email protected], and Marie- work is not altered. More information is available at http://www.diabetesjournals Clotilde Alves-Guerra, [email protected]. .org/content/license. Received 27 January 2017 and accepted 23 October 2017. diabetes.diabetesjournals.org Broche and Associates 79

Immunohistochemistry and anti-CD133 APC (clone 3152C11). For each antibody, Tissues were fixed in 10% formalin and processed for optimal dilution was determined by titration. Cells were in- immunohistochemistry, as described previously (15). The cubated for 15–30 min at 4°C in the dark, washed, and following antibodies were used: mouse anti-insulin suspended in HBSS without calcium/magnesium supple- (1:2,000; Sigma-Aldrich), rabbit anti-glucagon (1:1,000; mented with 20% FCS, and dead cells were excluded with Euromedex, Souffelweyerrsheim,France),rabbitanti-PDX1 propidium iodide (1:4,000; Sigma-Aldrich). Stained cells were (1:1,000), mouse anti-Ki67 (1:50; BD Pharmingen, Le Pont- analyzed and sorted with FACS Aria III (BD Biosciences). de-Claix, France), rabbit anti-amylase (1:300; Sigma-Aldrich), Data were analyzed in FlowJo (Ashland, OR) software. rabbit anti–neurogenin 3 (anti-NGN3; 1:1,000), rabbit anti– nuclear factor erythroid 2–related factor 2 (anti-NRF2; Islet Isolation Ucp22/2 1:1,000; GeneTex, Irvine, CA), rabbit anti-Akt (1:200), and Neonatal islets from wild-type (WT) and mice rabbit anti–phospho Akt (Ser 473) (1:25) (nos. 9272 and were harvested as described previously (19). Freshly dis- 9271; Cell Signaling, Saint-Quentin, France). The fluores- sected whole pancreata were digested with 0.5 mg/mL cent secondary antibodies were fluorescein isothiocyanate collagenase (Sigma-Aldrich) dissolved in HBSS at 37°C. anti-rabbit and Texas Red anti-mouse antibodies (1:200; Tubes were tapped regularly to aid tissue dispersal. Next, Jackson ImmunoResearch, Suffolk, U.K.), and Alexa Fluor lysates were washed with HBSS containing 10% FBS. Then, anti-rabbitantibody(1:400;Biogenex,Fremont,CA).For islets were handpicked under a dissecting stereoscope NGN3, revelation was performed using the vectastain (Leica, Nanterre, France). ABC kit (Vector Laboratories, Peterborough, U.K.). Fluo- Insulin Secretion rescent image acquisition was performed using the Zeiss Insulin secretion was quantified as described previously (20) fl AxioObserver Z1 inverted uorescence microscope coupled using an ultrasensitive mouse insulin ELISA kit (Crystal with the Zeiss Axiocam MRm (Zeiss, Marly-le-Roi, France). Chem,Zaandam,theNetherlands).

Determination of Cellular ATP Levels Quantification Detection of ATP levels was assessed using a luminescence- To quantify the absolute surface of PDX1-, insulin-, based assay kit (Roche, Meylan, France). glucagon-, and amylase-expressing and Hoechst-stained cells, 5-mm-thick sections of each pancreas were digitized at RNA Extraction and PCR E13.5 and E16.5. At E19.5 and postnatal day 2, one of the Procedures are described in Hoarau et al. (16). The oligo- five slides of the total pancreas was digitized (17). On every sequences for RT-PCR are available on request. image, the surface of immunostaining was quantified by Western Blot Analysis ImageJ (National Institutes of Health, Bethesda, MD). At For Western blotting analysis, cells were lysed in Laemmli. E16.5, the total number of immunopositive cells for NGN3 (20 mg) were separated by SDS-PAGE and electro- was counted on all sections of a complete pancreas. Statisti- phoretically transferred onto polyvinylidene fluoride mem- cal significance was determined by Student t test. To mea- brane (Bio-Rad, Marnes-la-Coquette, France). After blocking sure proliferation of early progenitors, we counted the with milk, membranes were probed with mouse anti– frequency of Ki67+ nuclei among 1,000 PDX1+ cells. At phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204; Cell least three rudiments per condition were analyzed. Statis- Signaling), mouse anti–b-actin (Sigma-Aldrich), mouse tical significance was determined using Student t test. anti–a-tubulin (Sigma-Aldrich), rabbit anti-Akt, and rabbit anti–phospho Akt (Ser 473) (nos. 9272 and 9271; Cell Sig- RESULTS naling). Immunoreactive bands were visualized with the Increased Pancreas Growth in the Ucp2 Knockout Mice SuperSignalSystem(Pierce,FisherScientific, Illkirch, France). First, the expression pattern of UCP2 was analyzed. E12.5 pancreatic epithelial and mesenchymal cells were separated Protein Oxidation Ucp2 Protein oxidation of total pancreas homogenates was by FACS (17). expression was enriched in the epithe- measured by assaying the amount of carbonyl groups on lial fraction containing the progenitors (Supplementary Fig. 1A). At E16.5, we separated mesenchymal, acinar, NGN3+, proteins using the OxyBlot kit (Protein Oxidation Detection Ucp2 Kit; Millipore, Molsheim, France). and endocrine cells (18). expression was found pref- erentially in endocrine cells and in a lesser extent in other Cell Suspension and Cell Sorting cell types (Supplementary Fig. 1B). To investigate its role, 2 The procedures were described previously (17,18). In brief, Ucp2+/ mice were intercrossed. The weight, islet insulin cell suspensions were stained in Hanks’ balanced salt solu- secretion, and glycemia of the homozygous neonates were tion (HBSS) without calcium/magnesium supplemented all similar to the controls (Supplementary Fig. 2). During with 20% FCS with the following anti-mouse antibodies the embryonic and fetal periods, the overall external mor- 2 2 purchased from BD Biosciences (Le Pont-de-Claix, France): phology of Ucp2 / animals was normal (Supplementary anti-CD45 PercpCy5.5 (clone 30F11), anti-CD31 PercpCy5.5 Fig. 3). As shown by the Hoechst staining, the size of the 2 2 (clone MEC13.3), anti-TER119 PercpCy5.5 (clone TER119), Ucp2 / pancreas at E16.5, E19.5, and PN2 was increased anti-EpCam BV421 (clone G8.8), anti-CD49f PE (clone GoH3), by nearly twofold compared with controls (Figs. 1, 2, and 4). 80 UCP2 and b-Cell Development Diabetes Volume 67, January 2018

Figure 1—UCP2 deficiency increases pancreas development. WT and Ucp22/2 pancreata were analyzed at postnatal day 2. The white dotted lines demarcate the limits of the pancreas. A: Nuclei were stained with Hoechst 33342 (blue). Islets were detected using anti-insulin (red) (B)and anti-glucagon (green) (C) antibodies. Exocrine cells were detected using anti-amylase antibody (green) (D). The absolute surface areas occupied by Hoechst-positive cells, insulin-positive cells, glucagon-positive cells, and amylase-positive cells were quantified. Each point represents the mean 6 SEM of n $ 3 individual pancreata. *P , 0.05. Scale bar = 50 mm.

This difference was not observed at E13.5 (Fig. 3C). More- However, a nonsignificant decrease of ATP per cell was 2 2 2 over, the absolute surfaces of insulin, glucagon, and amylase quantified at E16.5 (6.93 3 10 4 pmol in Ucp2 / vs. 2 2 2 were also increased in the Ucp2 / pups and fetuses (Figs. 9.83 3 10 4 pmol in WT, P = 0.25). The second hypothesis 1 and 2). Using an antibody directed against NGN3, we is that UCP2 is involved in the regulation of the production showed that the number of endocrine precursors increased of ROS (5,6). To examine this possibility, we performed proportionally to the pancreas size at E16.5 in the mutants immunofluorescence experiments to visualize the nuclear (Supplementary Fig. 4). To investigate the mechanism re- translocation of the ROS-sensitive factor NRF2. In the ab- 2/2 sponsible for the increased growth of the Ucp2 pancreata, sence of oxidative stress, NRF2 is associated with the pro- fi we quanti ed progenitor proliferation using anti-PDX1 and tein Keap1, which promotes the degradation of NRF2 by anti-Ki67 antibodies. At E13.5, we found an increased pro- the ubiquitin proteasome pathway. Also, oxidants can mod- + A B liferation of PDX1 progenitor cells (Fig. 3 and ), but not ify the cysteine residues of Keap1, leading to nuclear trans- at E12.5 (Supplementary Fig. 5). Together, these data dem- location of NRF2. Interestingly, we observed NRF2 only at Ucp2 onstrate that deletion induces an overgrowth of the the periphery of the nuclei at E13.5 in the WT pancreata pancreas due to an increased proliferation of progenitor cells. whereas nuclear translocation of NRF2 was observed in the UCP2 Controls Oxidative Stress and AKT Signaling mutants (Supplementary Fig. 7). At E16.5, nuclear trans- 2/2 in the Ucp22/2 Fetal Pancreas location was found both in WT and Ucp2 pancreata, in Twomainmechanismshavebeendescribedtoexplainthe an area containing b-cells, and this event was increased in biological effects of UCP2. First, UCP2 can modulate the the mutants (Supplementary Fig. 7). These results suggest energetic metabolism by controlling the balance between that ROS are involved in endocrine development. We also glycolysis and oxidative phosphorylation (8). To examine the quantified protein oxidation levels using the OxyBlot assay 2 2 energetic status of Ucp2 / pancreas, we quantified the (Supplementary Fig. 7). We found a nearly twofold increase 2 2 ATP content in Ucp2 / and WT pancreata. No difference of the protein oxidation levels in the mutant pancreata. was found at the pancreatic level (Supplementary Fig. 6). Together, these results indicate that oxidative stress is diabetes.diabetesjournals.org Broche and Associates 81

Figure 2—Pancreas growth is enhanced in the Ucp22/2 E16.5 fetuses. WT and Ucp22/2 pancreata were analyzed at E16.5. The white dotted lines demarcate the limits of the pancreas. A: Photographs of the WT and Ucp22/2 pancreata. Scale bar = 200 mm. B: Nuclei were stained with Hoechst 33342 (blue). Endocrine development was investigated using anti-insulin (red) (C) and anti-glucagon (green) (D) antibodies. The absolute surface areas occupied by Hoechst-positive cells, insulin-positive cells, and glucagon-positive cells were quantified. Each point represents the mean 6 SEM of n $ 3 individual pancreata. *P , 0.05. Scale bar = 50 mm.

2 2 2 2 higher in the Ucp2 / pancreata. To further investigate effect was abolished when Ucp2 / fetuses received signaling pathways involved in the pancreatic phenotype NAC treatment (Fig. 4). Interestingly, a-andb-cell prolif- 2 2 2 2 of Ucp2 / fetuses, we first analyzed the ERK1/2 pathway. eration was increased in E19.5 Ucp2 / pancreata versus No difference was found between mutants and controls controls (Supplementary Figs. 12 and 13). This effect was (Supplementary Fig. 8). Second, using immunofluorescence, abrogated when an NAC treatment was administrated. we analyzed the AKT signaling pathway, sensitive to ROS Thus, the knockout of Ucp2 controls the proliferation of levels (21). The total AKT level was slightly increased in the endocrine cells in an ROS-dependent manner. It leads to a mutants at E16.5 but not at E13.5. At both stages in the nonsignificant increased fraction of endocrine cells at PN2 mutants, we found an increased ratio of phospho-AKT to (Supplementary Fig. 14) but not at E19.5. Finally, we treated 2 2 total AKT, confirmed by Western blot at E16.5 (Supplemen- Ucp2 / and control mice with NAC from E9.5 to E13.5. tary Figs. 9 and 10). Thus, these data suggest that the Such treatment reduced progenitor proliferation induced activation of the ROS-AKT signaling pathway is involved by the deletion of Ucp2 (Fig. 3B). Altogether, these data 2 2 in the growth of Ucp2 / mouse pancreas. demonstrate that increased oxidative stress caused by the lack of UCP2 is responsible for the increased fetal pancreata NAC Treatment Reverses the Pancreatic Phenotype of the Ucp22/2 Fetuses growth. To further analyze the implication of ROS, we treated pregnant mice with the antioxidant NAC between E12.5 DISCUSSION 2 2 and E19.5. In Ucp2 / pancreata, the number of NRF2+ Our main findingisthatUCP2isanegativeregulatorof cells decreased when treated with NAC, validating its anti- pancreas development. Indeed, the absence of UCP2 in- oxidant effect (Supplementary Fig. 11). Pancreatic weight duces an increase in cell proliferation and a larger pancreas. and b-cell and a-cell masses were increased in untreated Moreover, this effect is induced by oxidative signals, through 2 2 Ucp2 / fetuses, compared with controls (Fig. 4). This the activation of the AKT pathway. 82 UCP2 and b-Cell Development Diabetes Volume 67, January 2018

Figure 3—The proliferation of the progenitor cells is increased in the Ucp22/2 E13.5 pancreata. A: The proliferation of the PDX1+ progenitor cells was analyzed using anti-PDX1 (green) and anti-Ki67 (red) antibodies. B: Proliferation percentage was also quantified. Each point represents the mean 6 SEM of n $ 3 individual pancreata. *P , 0.05. Scale bar = 50 mm.Forhighermagnification, scale bar = 10 mm. C: E13.5 pancreata were analyzed using anti-PDX1 antibodies (green). Nuclei were stained with Hoechst 3342 (blue). The white dotted lines demarcate the limits of the pancreas. The absolute areas of Hoechst-positive cells and PDX1-positive cells were quantified. Each point represents the mean 6 SEM of n $ 3 individual pancreata. Scale bar = 50 mm.

The Roles of UCP2 in Physiological and Pathological repressor of stem cell differentiation (22). Moreover, in mu- Processes rine embryonic fibroblasts, UCP2 was shown to negatively Here we show that the deletion of Ucp2 increases progen- control their proliferation (7). Finally, the roles of UCP2 itor and endocrine cell proliferation, two cell types that were investigated in different pathologies. In cancer cell normally express Ucp2 (Supplementary Fig. 1). This sug- lines expressing low levels of UCP2, its overexpression de- gests a potential autocrine effect of UCP2, but we do not creases cell proliferation through metabolic changes and in exclude other paracrine effects. For example, the mesen- consequence represses the malignant phenotype. Moreover, chyme that expresses lower levels of Ucp2 also controls in diabetes, the involvement of UCP2 is still controversial 2 2 progenitor proliferation (15). Moreover, several recent stud- (23). Indeed, Emre et al. (23) treated WT and Ucp2 / mice ies indicate that UCP2 plays a crucial role in the development with low doses of streptozotocin to generate an experimental of several cell types. Indeed, during human stem cell differ- model of diabetes. They found that autoimmune diabetes 2 2 entiation, UCP2 expression decreases, suggesting its role as a was accelerated in Ucp2 / mice, with the presence of an diabetes.diabetesjournals.org Broche and Associates 83

Figure 4—The antioxidant NAC normalizes the phenotype of the Ucp22/2 embryonic pancreata. Pregnant WT and Ucp22/2 mice were treated with 10 mmol/L NAC (drinking water) from E12.5 to E19.5 days postcoitum. Fetal pancreata were analyzed at E19.5. The white dotted lines demarcate the limits of the pancreas. A: Nuclei were detected using Hoechst staining (blue) and a-andb-cells were detected with anti-insulin (red) and anti-glucagon (green) antibodies. Scale bar = 50 mm. B: The pancreas weights, the a-cell mass, and the b-cellmasswerethen calculated. Each point represents the mean 6 SEM of three individual pancreata. *P , 0.05.

increased lymphocytic infiltration. On the contrary, using different doses of hydrogen peroxide were added to the similar experiments, Lee et al. (12) found that treatment of medium. We found that ROS stimulate endocrine differen- 2 2 WT and Ucp2 / mice with low doses of streptozotocin tiation by increasing the expression of NGN3. Moreover, resulted in hyperglycemia that was much less severe in this effect was ERK dependent. Despite similarities with the 2/2 Ucp2 mice than controls. The difference between these current study, some of these ROS effects are different from 2 2 two studies was suggested to be connected to the genetic the in vivo Ucp2 / model. Indeed, ROS-induced endocrine background of the mice. Moreover, in humans, another re- development in vitro is mainly due to an increased differ- Ucp2 cent illustration is that variants of the are asso- entiation, whereas here in vivo, an increased proliferation ciated with diabetes and diabetic retinopathy in a Chinese 2/2 of the progenitor cells mainly occurs in Ucp2 pancreata population (24). The exact mechanism responsible for di- prior to differentiation. We hypothesize that this difference abetes in these patients still needs to be elucidated. may be associated with the ROS levels in these two models. UCP2 and Oxidative Stress Moreover, in other cell types, such as embryonic stem cells, Previously, we used a culture model to analyze the effects of induced pluripotent stem cells, adipocytes, and neural pro- ROS on endocrine pancreas development (16). Embryonic genitors, oxidative stress was shown to stimulate either pancreata were cultured at the air/medium interface, and cell proliferation or cell differentiation, or both (16). Thus, 84 UCP2 and b-Cell Development Diabetes Volume 67, January 2018 these observations indicate thattheeffectsofROSarehighly 7. Pecqueur C, Bui T, Gelly C, et al. Uncoupling protein-2 controls proliferation by dependent on the cellular context. Moreover, downstream of promoting fatty acid oxidation and limiting glycolysis-derived pyruvate utilization. 2 2 ROS production, we found an activation of AKT in the Ucp2 / FASEB J 2008;22:9–18 pancreata. This link between ROS and AKT is similar to Le Belle 8. Esteves P, Pecqueur C, Alves-Guerra M-C. UCP2 induces metabolic re- programming to inhibit proliferation of cancer cells. Mol Cell Oncol 2014;2: et al. (21), which established that proliferative neural stem e975024 cells have high endogenous ROS levels that regulate both self- 9. Esteves P, Pecqueur C, Ransy C, et al. Mitochondrial retrograde signaling renewal and neurogenesis in a PI3K/AKT-dependent manner. mediated by UCP2 inhibits cancer cell proliferation and tumorigenesis. Cancer Res 2014;74:3971–3982 Conclusions fi 10. González-Barroso MM, Giurgea I, Bouillaud F, et al. Mutations in UCP2 in Our study demonstrates that UCP2 de ciency enhances the congenital hyperinsulinism reveal a role for regulation of insulin secretion. PLoS One growth of the pancreas during embryogenesis and the 2008;3:e3850 perinatal period. This effect is mediated by an activation 11. Zhang CY, Baffy G, Perret P, et al. Uncoupling protein-2 negatively regulates of the ROS-AKT signaling pathway. These mechanisms are insulin secretion and is a major link between obesity, beta cell dysfunction, and important to better understand congenital hyperinsulinism type 2 diabetes. Cell 2001;105:745–755 observed in children. 12. Lee SC, Robson-Doucette CA, Wheeler MB. Uncoupling protein 2 regulates reactive oxygen species formation in islets and influences susceptibility to di- abetogenic action of streptozotocin. J Endocrinol 2009;203:33–43 Acknowledgments. The authors thank Latif Rachdi (INSERM, U1016, Institut 13. Joseph JW, Koshkin V, Zhang C-Y, et al. Uncoupling protein 2 knockout mice Cochin; CNRS, UMR8104; Université Paris Descartes, Sorbonne Paris Cité) for helping have enhanced insulin secretory capacity after a high-fat diet. Diabetes 2002;51: to isolate the mouse neonatal islets. The authors thank Diane Girard (INSERM, 3211–3219 U1016, Institut Cochin; CNRS, UMR8104; Université Paris Descartes, Sorbonne Paris 14. Blanc J, Alves-Guerra MC, Esposito B, et al. Protective role of uncoupling Cité) for the English editing of the manuscript. The research leading to these results protein 2 in atherosclerosis. Circulation 2003;107:388–390 received support from Société Francophone du Diabéte–Boehringer Ingelheim-Lilly. 15. Attali M, Stetsyuk V, Basmaciogullari A, et al. Control of beta-cell differentiation Duality of Interest. No other potential conflicts of interest relevant to this by the pancreatic mesenchyme. Diabetes 2007;56:1248–1258 article were reported. 16.HoarauE,ChandraV,RustinP,ScharfmannR,DuvillieB.Pro-oxidant/ Author Contributions. B.B., S.B.F., E.A., and C.B. designed and performed antioxidant balance controls pancreatic b-cell differentiation through the ERK1/2 experiments and analyzed the data. T.S., M.B., and F.M. performed experiments. pathway. Cell Death Dis 2014;5:e1487 R.S. contributed to discussion and wrote the manuscript. M.-C.A.-G. and B.D. 17. Ramond C, Glaser N, Berthault C, et al. Reconstructing human pancreatic designed research experiments, performed experiments, analyzed data, and wrote differentiation by mapping specific cell populations during development. eLife 2017; the manuscript. B.D. and M.-C.A.-G. are the guarantors of this work and, as such, 6:e27564 had full access to all the data in the study and take responsibility for the integrity of 18. Sugiyama T, Rodriguez RT, McLean GW, Kim SK. Conserved markers of fetal thedataandtheaccuracyofthedataanalysis. pancreatic epithelium permit prospective isolation of islet progenitor cells by FACS. Proc Natl Acad Sci U S A 2007;104:175–180 References 19. Huang C, Gu G. Effective isolation of functional islets from neonatal mouse 1. Simmons RA. Developmental origins of diabetes: the role of oxidative stress. pancreas. J Vis Exp 2017;119 Free Radic Biol Med 2006;40:917–922 20. Lecomte MJ, Pechberty S, Machado C, et al. Aggregation of engineered human 2. Fleury C, Neverova M, Collins S, et al. Uncoupling protein-2: a novel gene linked b-cells into pseudoislets: insulin secretion and gene expression profile in normoxic to obesity and hyperinsulinemia. Nat Genet 1997;15:269–272 and hypoxic milieu. Cell Med 2016;8:99–112 3. Vozza A, Parisi G, De Leonardis F, et al. UCP2 transports C4 metabolites out of 21. Le Belle JE, Orozco NM, Paucar AA, et al. Proliferative neural stem cells have mitochondria, regulating glucose and glutamine oxidation. Proc Natl Acad Sci U S A high endogenous ROS levels that regulate self-renewal and neurogenesis in a 2014;111:960–965 PI3K/Akt-dependant manner. Cell Stem Cell 2011;8:59–71 4. Pecqueur C, Alves-Guerra MC, Gelly C, et al. Uncoupling protein 2, in vivo 22. Zhang J, Khvorostov I, Hong JS, et al. UCP2 regulates energy metabolism distribution, induction upon oxidative stress, and evidence for translational regulation. and differentiation potential of human pluripotent stem cells. EMBO J 2011;30: J Biol Chem 2001;276:8705–8712 4860–4873 5. Robson-Doucette CA, Sultan S, Allister EM, et al. Beta-cell uncoupling protein 23. Emre Y, Hurtaud C, Karaca M, Nubel T, Zavala F, Ricquier D. Role of uncoupling 2 regulates reactive oxygen species production, which influences both insulin and protein UCP2 in cell-mediated immunity: how macrophage-mediated insulitis is glucagon secretion. Diabetes 2011;60:2710–2719 accelerated in a model of autoimmune diabetes. Proc Natl Acad Sci U S A 2007;104: 6. Arsenijevic D, Onuma H, Pecqueur C, et al. Disruption of the uncoupling protein-2 19085–19090 gene in mice reveals a role in immunity and reactive oxygen species production. Nat 24. Shen Y, Wen Z, Wang N, et al. Investigation of variants in UCP2 in Chinese Genet 2000;26:435–439 type 2 diabetes and diabetic retinopathy. PLoS One 2014;9:e112670