Bromodomain and Extra Terminal Protein Inhibitors Promote Pancreatic Endocrine Cell Fate

During pancreatic development, multipotent endodermal pancreatic progenitors proliferate in response to signals derived from surrounding mesodermal cells and next differentiate into cells with exocrine and endocrine properties, including β-cells (1,2). Chronic failure to reduce high blood glucose levels results in diabetes, which in most patients is due to impaired functional β-cell mass. Recent efforts, therefore, have concentrated on developing in vitro protocols to produce functional β-cells from induced pluripotent stem cells (iPSCs) or embryonic stem cells to replenish the decreasing β-cell mass (3). These protocols are based on our knowledge of the molecular mechanisms underlying in vivo β-cell development (4,5). Despite recent considerable progress, generating a homogeneous population of fully mature, glucose-responsive β-cells remains a challenge (6).

The mouse embryonic pancreas starts developing at approximately embryonic day 9.0 (E9.0), with the dorsal and ventral budding of the foregut endoderm under the influence of surrounding mesodermal structures. At this stage, multipotent proliferating epithelial pancreatic progenitors express a specific set of transcription factors, such as PDX1 and NKX6.1 (7). They can undergo exocrine and endocrine cell fate until approximately E11.5 and then progressively segregate into acinar progenitors or bipotent endocrine/duct progenitors (8,9). Endocrine progenitors then transiently express the basic helix-loop-helix transcription factor neurogenin3 (NEUROG3) to give rise to all pancreatic endocrine cells (10).

Major progress has been made on growth factors, such as FGF7 and FGF10, that signal through FGFR2b (11–15) and activate the proliferation of pancreatic progenitors as well as on small molecules acting through yet-to-be-discovered pathways (16) or upon coculture on a layer of 3T3-J2 feeder cells (17). On the other hand, information regarding signals that modulate the differentiation of pancreatic progenitors into functional β-cells remains scarce, and the objective of the current work was to increase our knowledge on this topic.

The bromodomain and extraterminal (BET) family of proteins comprises four members: BRD2, BRD3, BRD4, and BRDT. The latter is specifically expressed in the testis, whereas the other three are more ubiquitously expressed (18–20). The BET proteins recently emerged as a major class of epigenetic readers and modulators of gene expression by their ability to recognize and bind through their two bromodomains N-acetylated-lysine residues of histone tails (21). They subsequently induce an opened chromatin structure and can tether various transcription factors at target promoters and enhancer regions to promote transcription (22). The BET family of proteins largely has been associated with cancer progression, and recent efforts have concentrated on developing potent and specific inhibitors of BET proteins (BETis), such as (+)-JQ1 and I-BET151 (21,23,24). BET proteins play a crucial role throughout development, as Brd2- and 4-null mice are embryonic lethal (25,26). Recent work has shown that BRD4 participates in adipogenesis and myogenesis by modulating gene transcription at specific enhancer regions (27), yet the role of BET proteins in organogenesis and cell fate remains poorly characterized.

In this study, we have evaluated the effect of BET protein inhibition during pancreatic development. We used two different in vitro models: 1) a culture of rodent fetal pancreas under conditions that replicate the major steps of in vivo pancreatic development (13) and 2) in vitro differentiation of human iPSCs (hiPSCs) into insulin-producing cells (4,28). We report that BET inhibition using either I-BET151 or JQ1 increases the pool of NEUROG3⁺ endocrine progenitors in mouse embryonic pancreatic explants and in pancreatic progenitors derived from hiPSCs. This increased number of endocrine progenitors resulted in enhanced endocrine differentiation from pancreatic explants, whereas GHRL expression was increased in hiPSCs. The current data demonstrate the importance of the BET protein family during pancreatic endocrine lineage differentiation.

Pregnant C57BL/6J mice were purchased from the Janvier Breeding Center and killed by CO₂ asphyxiation according to French Animal Care Committee guidelines. Dorsal pancreatic buds were dissected as previously described (13).

E11.5 pancreatic buds were cultured as previously described (13). The medium was changed daily and supplemented with either 0.1% DMSO, 500 nmol/L I-BET151 (Sigma-Aldrich), or 100 nmol/L JQ1 (Abcam). For the cell proliferation assay, 10 μmol/L BrdU (Sigma-Aldrich) was added during the last hour of culture.

Wild-type SBAD03-01 and SBAD03-04 hiPSCs were obtained through the Innovative Medicines Initiative/European Union–sponsored StemBANCC consortium through the Human Biomaterials Resource Centre, Birmingham University (www.birmingham.ac.uk/facilities/hbrc). Cells were cultured and differentiated as previously described (29); BETi or DMSO control was added daily to culture medium during stage 4 (pancreatic endoderm) or 5 (endocrine progenitors) of the differentiation. Data presented in this article are from the SBAD03-01 line; similar data were obtained with SBAD03-04 hiPSCs (Supplementary Fig. 1).

Total RNA was extracted as previously described (29,30). Real-time PCR was performed with a QuantStudio 3 or OneStep Plus Real-Time PCR system (Applied Biosystems) or an Mx3005P quantitative PCR system (Stratagene). Each reaction consisted of either a mix of Power SYBR Green PCR Master Mix (Applied Biosystems) with a specific pair of designed primers or a mix of Taqman Universal PCR Master Mix with a specific labeled probe (Applied Biosystems). Data are presented relative to cyclophylin A (for rodent samples) or HPRT1 and ACTB (for hiPSC samples).

Mouse fetal pancreata were processed for immunohistochemistry, as previously described (30). All primary antibodies and dilutions are described in the Supplementary Data. The fluorescent secondary antibodies were purchased from Jackson ImmunoResearch (1/400). The biotin-labeled secondary antibodies were purchased from Vector Laboratories (1/200). NEUROG3 and MAFA were detected using the VECTASTAIN Elite ABC Kit (Vector Laboratories). The nuclei were stained using the Hoechst 33342 fluorescent stain (0.3 mg/mL) (Invitrogen). Surface area quantifications were performed on one out of three consecutive sections (i.e., sections separated by 12 μm) to avoid counting the same cell twice. The signal was quantified using ImageJ software (National Institutes of Health) and summed to obtain the surface area per explant (expressed in mm²). NEUROG3⁺ and MAFA⁺ nuclei were manually counted on one out of three consecutive sections with ImageJ and then summed to obtain the number of positive nuclei per explant.

hiPSCs were fixed with 4% paraformaldehyde at room temperature for 30 min and then processed for immunocytochemistry as previously described (31).

Cells were sorted as previously described (32,33). hiPSCs were harvested to a single-cell solution using TrypLE Select and subsequently fixed and stained as previously described (28). Additional information on antibodies is provided in the Supplementary Data.

Confluent Min6 cells were treated with either 0.1% DMSO or 1 μmol/L JQ1 for 24 h at 37°C in a humidified 95% air/5% CO₂ gas mixture. Chromatin immunoprecipitation (ChIP) was performed as described by Cotney et al. (34) using 5 μg of BRD2, 3, and 4 rabbit polyclonal antibodies (A302-583A; A302-368A; A301-985150; Bethyl Laboratories). Immunoprecipitated chromatin was analyzed by real-time PCR. Primer sequences are detailed in the Supplementary Data.

Pools of at least five pancreatic buds were incubated overnight in culture medium containing 2.8 mmol/L glucose (Sigma-Aldrich) and 2% FCS (Eurobio), then incubated in Krebs-Ringer buffer (125 mmol/L NaCl, 4.7 mmol/L KCl, 2.5 mmol/L CaCl₂, 1.2 mmol/L MgSO₄, 1.2 mmol/L KH₂PO₄, 25 mmol/L NaHCO₃, pH 7.4) supplemented with 0.2% BSA (Sigma-Aldrich) and 2.8 mmol/L glucose for 60 min. Insulin secretion was assessed by sequential static incubations of 60 min in Krebs-Ringer buffer containing 45 μmol/L 3-isobutyl-1-methylxanthine (Sigma-Aldrich) and increasing concentrations of glucose (2.8–22.4 mmol/L) and finally 30 mmol/L KCl without 3-isobutyl-1-methylxanthine.

Pancreatic explants were homogenized in radioimmunoprecipitation assay buffer, and protein concentration was assayed using a Bicinchoninic Acid Protein Assay Kit (Pierce). Secreted insulin and content were measured in duplicate by Mouse Ultra Sensitive ELISA Kit (Crystal Chem) according to the manufacturer’s instructions.

All quantitative data are presented as the mean ± SD. Statistical significance was set at 5% and determined using ordinary one-way ANOVA with Dunett post hoc test except for ChIP data, which were determined using repeated-measures one-way ANOVA.

At E11.5, we separated the EpCAM⁺ fraction that contains epithelial progenitors from EpCAM⁻ mesenchymal fraction by FACS and performed real-time PCR. Brd2, 3, and 4, but not Brdt, mRNAs were detected in both the epithelial and the mesenchymal fractions (Fig. 1A). At E16, EpCAM⁺ cells were separated from EpCAM⁻ cells by FACS and further divided into three fractions that contain exocrine cells, endocrine progenitors, and hormone-producing cells (33). Brd2, 3, and 4 were detected in the mesenchymal as well as in all three EpCAM⁺ fractions, with the highest expression in the hormone-producing cells (Fig. 1B). We next cultured E11.5 pancreatic dorsal buds for 1, 3, 5, or 7 days under culture conditions that successfully replicated in vivo development of exocrine and endocrine lineages, with day 3 (d3) being comparable to E15.5 and d7 to E17.5–18.5 (35). We observed a sustained expression of Brd2, 3, and 4 during the 7-day culture period (Fig. 1B). These results hence show that Brd2, 3, and 4 are expressed as early as E11.5 and that their expression is maintained during pancreas development.

We treated mouse pancreatic explants with I-BET151 and analyzed by real-time PCR the expression of Neurog3, a specific marker of pancreatic endocrine progenitors (10). In control explants, Neurog3 mRNA level peaked after 3 days of culture when endocrine progenitors develop and decreased thereafter as they further differentiate into hormone-expressing cells (10,35). I-BET151 treatment enhanced Neurog3 expression after 3 days of culture with a 10-fold induction over control conditions at d5 and d7 (Fig. 2A). Similar induction of Neurog3 levels was obtained with another BETi, JQ1 (Fig. 2A). Immunohistochemical analysis at d5 showed an increase of NEUROG3⁺ cells upon treatment with either I-BET151 or JQ1 (Fig. 2B and C for quantification). Finally, both inhibitors enhanced the expression of NeuroD and Fev, two downstream targets of NEUROG3 (36,37) (Fig. 2D). The surge of NEUROG3⁺ cells was not the result of an increased proliferation of multipotent pancreatic progenitors during the early stages of explant development: 1) There was no major variations of multipotent progenitor markers Pdx1, Sox9, and Ptf1a after 24 h of exposure to BETis (Supplementary Fig. 2), and 2) immunohistochemical analyses of BrdU incorporation indicated that the proliferation of PDX1⁺ pancreatic progenitors did not increase. In fact, BrdU incorporation by PDX1⁺ cells upon BETi treatment was decreased by 10% (Supplementary Fig. 3A and B). The NEUROG3 surge did not result from an increased proliferation of endocrine progenitors. Indeed, BrdU incorporation remained nearly undetectable in NEUROG3⁺ cells from pancreatic buds cultured with I-BET151 or JQ1 (Supplementary Fig. 3C). TUNEL staining revealed no variations in apoptotic events (Supplementary Fig. 3D and E).

We next examined the effect of BETis on human endocrine progenitors derived from hiPSCs. They were first differentiated toward multipotent pancreatic progenitors and further cultured in the presence of I-BET151 and JQ1 (Fig. 3A). Both inhibitors increased NEUROG3 expression (Fig. 3B). Immunohistochemistry and flow cytometry analysis further confirmed enhanced NEUROG3 expression (Fig. 3C and D), with an almost doubling of the NEUROG3⁺ population (Fig. 3E). The increase of NEUROG3 expression followed a dose-response curve, with no decline in cell viability (Supplementary Fig. 4). These results indicate that BET bromodomain inhibition increases the pool of endocrine progenitors both in mouse embryonic pancreatic explants and in a model of multipotent pancreatic progenitors derived from hiPSCs.

We next assessed whether the increased pool of NEUROG3⁺ cells that developed upon BETi treatment gives birth to more hormone-producing cells in mouse pancreatic buds. Somatostatin (Sst) and glucagon (Gcg), markers of δ- and α-cells, respectively, were significantly increased by I-BET151 and JQ1 treatments (Fig. 4A). Expression of ghrelin (Ghrl), a marker of ε-lineage expressed at low levels in control conditions, was markedly increased, peaking at d5 of culture with a 25- and a 47-fold increase with I-BET151 and JQ1, respectively (Fig. 4A). Immunohistological analysis indicated a sharp increase in the number of GHRL⁺ cells at d7 with both inhibitors (Fig. 4B and C for quantification). Most GHRL⁺ cells stained negative for GCG, and all GHRL⁺ cells stained negative for PDX1, INS, and SST (Fig. 4B and Supplementary Fig. 5). Altogether, these results show that BET bromodomain inhibition induces an increase of α-, δ-, and ε-cell markers. The expression of acinar markers also was evaluated, and we observed a strong decrease in Amy expression, whereas Cpa1 and CelA were upregulated (Fig. 4D). This finding suggests that BET inhibition enhances acinar development but that in the meantime, BET activity is required for amylase expression.

We next determined whether I-BET151 and JQ1 treatments influence β-cell development in mouse pancreatic buds. As expected, Pcsk1/3, MafA, and Iapp expression increased during the culture period in control conditions, whereas Nkx6.1, which encodes a transcription factor first expressed in all early multipotent pancreatic progenitors and next restricted to β-cells, decreased (Fig. 5A, white bars). BETis sharply increased the expression of all four β-cell markers (Fig. 5A). The increase in MafA mRNA was further confirmed by immunohistochemistry at d7 of culture. I-BET151 and JQ1 treatments increased the number of MAFA⁺ cells by 2.5 ± 0.6- and 2.8 ± 0.6-fold, respectively (Fig. 5B and C for quantification). These results demonstrate that β-cell development is induced upon BETi treatment.

Of note, however, we observed a major reduction of Ins1 expression upon BETi treatment. Ins1 mRNA levels were reduced by 47- and 20-fold after 7 days of culture with I-BET151 and JQ1, respectively (Fig. 5D). This was further confirmed by immunohistochemistry (Fig. 5E and F for quantification) and by ELISA (Fig. 5G). On the other hand, Ins2 mRNA levels were reduced only mildly after I-BET151 and JQ1 treatment (Fig. 5H). This difference between Ins1 and Ins2 expression after BETi treatment was next analyzed by immunohistochemistry. We used antibodies against either C-peptide 1 or C-peptide 2 as surrogate markers of INS1 and INS2 expression. We observed an almost complete loss of C-peptide (C-PEP) 1 signal with both inhibitors, whereas many cells remained positive for C-PEP 2 (Supplementary Fig. 6). Such a differential effect of BETis on Ins1 and Ins2 expression also was observed in the mouse insulinoma cell line MIN6 (Supplementary Fig. 7A) and in primary mouse islets (Supplementary Fig. 7B). To determine whether such an effect was due to differential binding of BET proteins to the Ins1 and Ins2 promoters, ChIP of BRD2, 3, and 4 was performed in MIN6 cells. We observed strong bindings of BRD2 and 4 to both promoters, which strongly decreased upon JQ1 treatment. Of note, BRD2 and 4 were bound at the proximal region of the Ins1 promoter, whereas they bound at a more distal position of the Ins2 promoter (Fig. 6). This indicates that rather than reducing the ability of amplified NEUROG3⁺ endocrine progenitors to pursue β-cell differentiation, BETis promote β-cell development but in the meantime, downregulate Ins1 expression with little effect on Ins2.

Our results demonstrate that exposure to either I-BET151 or JQ1 enhances pancreatic endocrine cell development. However, Ins1 levels are strongly downregulated upon such chronic treatment. To determine whether BETi exposure during explant development could ultimately lead to an increase of insulin-producing β-cells, we performed pulse-chase experiments. We cultured (pulse period) pancreatic explants with BETis for 5 days. At this stage, Neurog3 expression is at its highest. Pancreatic explants were then cultured for 9 additional days (chase period) without BETis and analyzed (Fig. 7A). Under this setting at d14, Ins1 and Ins2 mRNA levels increased three-to fourfold (Fig. 7B) and insulin content was also strongly increased (Fig. 7C). MafA, but not MafB, mRNA level also was strongly upregulated, with a 13- and 16-fold increase after I-BET151 and JQ1 pulses/chases, respectively (Fig. 7D). It was also the case for Ucn3 mRNA levels (22- and 23-fold increases) (Fig. 7D). Gcg, Sst, and Ghrl expression also were increased (Fig. 7D) as were other β-cell markers, such as Nkx6.1, Pcsk1/3, Pdx1, and Iapp (Supplementary Fig. 10A). Immunohistochemical labeling of MAFA (Fig. 7E and F for quantification) and UCN3 (Fig. 7G) after BETi pulse/chase confirmed this massive increase observed by real-time PCR. There was no coexpression of INS with either SST or GCG, showing that the resulting β-cells are not polyhormonal (Supplementary Fig. 10B). ELISA of insulin secretion did not show a response upon glucose stimulation (Fig. 7H), suggesting that the resulting β-cells yet remain not fully mature. Similar to what was observed in pancreatic buds, transient BETi treatment of hiPSCs during stage 5 reduced C-PEP⁺ cells and INS expression (Fig. 8A and B). GCG⁺ and SST⁺ cells also were reduced (Fig. 8A), whereas GHRL expression was upregulated (Fig. 8B). Cells were then further cultured without BETis during the maturing endocrine cells stage (chase during stage 6). C-PEP expression increased after removal of BETis (Fig. 8A and B compared with 8C and D), but remained lower than in control conditions (Fig. 8C and D). GCG and SST expressions were similar to the control condition (Fig. 8C). On the other end, GHRL expression increased (Fig. 8C and D). Of note, coimmunocytochemistry of C-PEP with either SST and GHRL or GCG and GHRL showed that most endocrine cells at the end of stage 6 were monohormonal (Supplementary Fig. 11).

Taken together, the data indicate that treatment with BETis induces an increased pool of Neurog3⁺ endocrine progenitors that will further differentiate into β-, α-, δ-, or ε-cells in mouse pancreatic buds or ε-cells in hiPSCs.

The molecular mechanisms underlying the transitions from pancreatic progenitors toward mature endocrine cells remain not fully elucidated. Histone modifications are key epigenetic events that play major roles in cell proliferation and differentiation (38), and we previously showed that inhibiting histone deacetylases modulates cell fate during rat pancreas development (30). Here, we extended our comprehension of epigenetic modulation of embryonic pancreas development by studying histone code reader BET proteins. We provide evidence that BET inhibition promotes the pool of NEUROG3⁺ endocrine progenitors, which subsequently gives rise to an increased pool of endocrine cells.

To explore the role of BET proteins in embryonic pancreas development, we used an in vitro model of dorsal embryonic pancreatic buds that recapitulates the major steps of endocrine and exocrine development that occur in vivo (13,35). We first observed that Brd2, 3, and 4 expression at E11.5 was similar in the epithelial population that is enriched in pancreatic progenitors and mesenchymal cells. This result was expected because Brd2, 3, and 4 have been described as ubiquitously expressed in the majority of tissues (22), with some exceptions such as the highly enriched expression of BRD2, 3, and 4 in crypts of the small intestine compared with villi (39). Of note, Brd2, 3, and 4 expression at E16.5 is higher in a fraction enriched in hormone-expressing cells than in fractions enriched in acinar or pancreatic progenitor cells, suggesting an important role of BET proteins in establishing or maintaining endocrine cell fate.

In mouse pancreatic explants cultured with I-BET151 or JQ1, Neurog3 expression was amplified at both the mRNA and the protein level. NEUROG3 activity also was increased as demonstrated by the increased expression of two of its targets, NeuroD1 (37) and Fev (36). This increase of NEUROG3⁺ cells was not the consequence of an upstream effect on proliferation of PDX1⁺ pancreatic progenitors or NEUROG3⁺ endocrine progenitors themselves. In mouse pancreatic buds, the expression pattern of Neurog3 was shifted because it peaked at d5 of culture and remained 10 times more expressed than in controls at d7. A possibility is that BETis increase the half-life of NEUROG3. Indeed, NEUROG3 has been described as an unstable protein rapidly degraded by ubiquitin-mediated proteolysis (40), and it can be suspected that Neurog3 mRNA is unstable. Finally, it is notable that the positive effect of BETis on NEUROG3 expression is context and tissue dependent because BETi (CPI203 or I-BET151) treatment induces a loss of progenitor cell markers such as Neurog3 in enteroendocrine cell differentiation in the mouse adult small intestine (39).

In mouse explants, BETi treatment increased the expressions of endocrine markers, with Ghrl being the most strongly upregulated. Ghrelin is expressed by a rare population of pancreatic endocrine cells named ε-cells (41), and GHRL⁺ cells that developed with BETis resemble ε-cells. They do not express PDX1, INS, or SST, and only a few coexpress GCG (Supplementary Fig. 5). GHRL⁺ cell number was shown to be increased in the pancreata of mice deficient for either nkx2.2, pax6, or pax4 (41,42). Our data suggest, however, that ghrelin induction upon BETi treatment does not occur through Nkx2.2 or Pax6, whose expressions did not decrease after explant treatment (Supplementary Fig. 8A). The moderate reduction of Pax4 expression (Supplementary Fig. 8B) seems unlikely to explain the observed effect on ghrelin because supernumerary GHRL⁺ cells of Pax4^−/− mice coexpress GCG and low PDX1 levels (43), which was not the case in our model. The mechanism involved in the effects of BETis on the development of ε-cells hence remains to be clarified. We also observed a strong decrease in amylase expression in pancreatic buds treated with BETis, yet other markers, such as Cela1 or Cpa1, were upregulated by BETis (Fig. 4D and Supplementary Fig. 9), suggesting an incomplete development of the acinar compartment.

Nkx6.1 and MafA are key transcription factors essential for β-cell differentiation and maturation, respectively. Their upregulation should then reflect an increase of β-cell differentiation. We therefore were surprised to observe that Ins1 expression was dramatically reduced after treatment of pancreatic buds with BETis, whereas Ins2 did not vary with JQ1 and only by twofold with I-BET151. Our data indicate that this was also the case in two models of mature β-cells, adult mouse islets, and MIN6 cells, where Ins1, but not Ins2, was downregulated by I-BET151 and JQ1. Brd4 has been the most studied BET member because it was shown to modulate RNA polymerase II activity either by interacting with members of the transcription initiation complex, such as pTEFb and mediator (44–46), or by directly releasing proximally paused RNA polymerase II (47). Of note, Brd4 was shown to associate specifically with active promoters and enhancers of a given cell type and to tether various transcription factors through its extraterminal domain (48,49). The insulin promoter is hyperacetylated in β-cells (50), and Ins1 is among the most-expressed genes in β-cells. ChIP of BRD2, 3, and 4 in BETi-treated MIN6 cells indicated that BRD4, and to a lesser extent, BRD2, were strongly bound at the proximal promoter of Ins1 (−85/+16). We also found them to bind Ins2 promoter, but at a more distal position (−683/−771). This could explain the differences observed in BETi responses because proximally bound BRD4 could be more likely to influence RNA polymerase II activity. Such an approach and high throughput ChIP sequencing analysis on purified β-cells or NEUROG3⁺ progenitors would be of great use for deciphering the underlying mechanisms of BET on pancreas development and insulin regulation. However, to date, there are no available specific markers to efficiently purify such cells from mouse pancreatic buds.

Our data indicate that BETis can increase the proportion of NEUROG3⁺ endocrine progenitors both in a model of rodent pancreatic development and in a model of β-cell development from hiPSCs. However, the failure of the generated β-cells to express insulin properly represents a limiting step. Our pulse-chase approach in rodent, however, indicates that NEUROG3⁺ endocrine progenitors amplified with BETis can develop into INS⁺ β-cells after BETi removal. Moreover, newly formed β-cells expressed higher levels of MafA and Ucn3, both considered as markers of β-cell maturity (51,52). Their maturity, however, is not yet complete because their sensitivity to glucose in terms of insulin secretion was not activated, suggesting that MafA or Ucn3 expression cannot be used as sole markers of cell maturity. It is established that in rodent, β-cells appear quite late during development compared with, for example, α-cells (1). Early NEUROG3⁺ cells also have been shown to differentiate into α-cells and those appearing later on into β-cells (53). Thus, it could be postulated that the timing of differentiation correlates with maturity and that β-cells generated upon BETi treatment are more mature because the NEUROG3 expression peak was maintained for a longer period. The current results in hiPSCs show that BET proteins are essential for endocrine progenitor differentiation because BET inhibition during the endocrine progenitor stage enhanced NEUROG3⁺ cells and resulted in a total lack of C-PEP⁺ and GCG⁺ cells. Such an effect was not observed when hiPSCs were cultured with BETis during the earlier pancreatic endoderm stage (Supplementary Fig. 11). Further removing BETis from amplified NEUROG3⁺ progenitors replenished the C-PEP⁺ population that, however, did not quite reach control levels. Further testing of additional conditions may permit the reproduction of the increase in β-cell population observed in mouse pancreatic buds. Indeed, the discrepancies between the two models may come from different susceptibility time windows to BET inhibition, which should be tested. They also might be due to the absence, in hiPSCs, of mesenchymal tissue, which participates in the control of β-cell differentiation (54) and could mediate some BETi effects in explants. Finally, we observed a rise of ε-cells in hiPSCs that were pulsed with BETis during the endocrine progenitor stage (Fig. 8), which is consistent with what was observed in mouse pancreatic buds. Therefore, this should be of particular interest to study ε-cell differentiation, which remains poorly understood. Altogether, our results obtained across two very different models of pancreas development revealed novel roles of BET on pancreatic endocrine progenitors and their differentiation into endocrine cells.

Acknowledgments. The authors acknowledge the transcriptomic platform from the Cochin Institute for performing array hybridizations and thank the flow cytometry platform at the Novo Nordisk Foundation Center for Stem Cell Biology.

Funding. This work was supported by the Agence Nationale de la Recherche (BromoBeta) and the Fondation Francophone pour la Recherche sur le Diabéte (2018). The R.S. laboratory is supported by Fondation Bettencourt Schueller and belongs to the Laboratoire d’Excellence consortium Revive and to the Departement Hospitalo-Universitaire Autoimmune and Hormonal disease. The research leading to these results has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement number 115439, resources of which comprise a financial contribution from the European Union’s Seventh Framework Programme (FP7/2007-2013) and an in-kind contribution from European Federation of Pharmaceutical Industries and Associations companies. The Novo Nordisk Foundation Center for Stem Cell Biology is supported by Novo Nordisk Foundation grant number NNF17CC0027852.

This article reflects only the author’s views, and neither the Innovative Medicines Initiative Joint Undertaking, the European Federation of Pharmaceutical Industries and Associations, nor the European Commission is liable for any use that may be made of the information contained herein.

Duality of Interest. M.H. and C.H. are employees of Novo Nordisk A/S and may hold shares in this company. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. L.H., C.B., V.A., and L.R. performed and analyzed experiments on mouse material. L.H. and R.S. designed the experiments, interpreted the data, and wrote the manuscript. M.B.K.P. and C.H. designed, performed, and analyzed experiments on hiPSCs. M.H. and A.G.-B. provided critical input and analyzed the data. R.S. supervised the study and obtained funding. R.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Data Availability. The data sets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. No applicable resources were generated or analyzed during the current study.