Aims/hypothesis Most type 2 diabetes-associated genetic variants identified via genome-wide association studies (GWASs) appear to act via the pancreatic islet. Observed defects in insulin secretion could result from an impact of these variants on islet development and/or the function of mature islets. Most functional studies have focused on the latter, given limitations regarding access to human fetal islet tissue. Capitalising upon advances in in vitro differentiation, we characterised the transcriptomes of human induced pluripotent stem cell (iPSC) lines differentiated along the pancreatic endocrine lineage, and explored the contribution of altered islet development to the pathogenesis of type 2 diabetes. Methods We performed whole-transcriptome RNA sequencing of human iPSC lines from three independent donors, at baseline and at seven subsequent stages during in vitro islet differentiation. Differentially expressed genes (q < 0.01, log fold change [FC] > 1) were assigned to the stages at which they were most markedly upregulated. We used these data to characterise upstream transcription factors directing different stages of development, and to explore the relationship between RNA expression profiles and genes mapping to type 2 diabetes GWAS signals. Results We identified 9409 differentially expressed genes across all stages, including many known markers of islet development. Integration of differential expression data with information on transcription factor motifs highlighted the potential contribution of REST to islet development. Over 70% of genes mapping within type 2 diabetes-associated credible intervals showed peak differential expression during islet development, and type 2 diabetes GWAS loci of largest effect (including TCF7L2; −10 log FC = 1.2; q= 8.5 × 10 ) were notably enriched in genes differentially expressed at the posterior foregut stage (q = 0.002), as calculated by gene set enrichment analyses. In a complementary analysis of enrichment, genes differentially expressed in the final, beta-like cell stage of in vitro differentiation were significantly enriched (hypergeometric test, permuted p value <0.05) for genes within the credible intervals of type 2 diabetes GWAS loci. Conclusions/interpretation The present study characterises RNA expression profiles during human islet differentiation, iden- tifies potential transcriptional regulators of the differentiation process, and suggests that the inherited predisposition to type 2 diabetes is partly mediated through modulation of islet development. Marta Perez-Alcantara and Christian Honoré contributed equally to this study. Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00125-018-4612-4) contains peer-reviewed but unedited supplementary material, which is available to authorised users. * Nicola L. Beer Oxford Centre for Diabetes, Endocrinology & Metabolism, email@example.com University of Oxford, Old Road, Oxford OX3 7LE, UK Oxford NIHR Biomedical Research Centre, Churchill Hospital, Oxford, UK Wellcome Centre for Human Genetics, University of Oxford, Oxford, UK Department of Stem Cell Research, Novo Nordisk A/S, Maaloev, Denmark Department of Stem Cell Biology, Novo Nordisk A/S, Maaloev, Denmark Diabetologia (2018) 61:1614–1622 1615 Data availability Sequence data for this study has been deposited at the European Genome-phenome Archive (EGA), under accession number EGAS00001002721. . . . . . Keywords Diabetes Endocrine pancreas In vitro differentiation Islets Stem cells Transcriptome Abbreviations lines, but there is mounting evidence that some of the FC Fold change implicated genetic variants influence islet development GSEA Gene set enrichment analysis . For example, many of the monogenic diabetes GWAS Genome-wide association study genes—most of which impact on islet development — iPSC Induced pluripotent stem cell are also found in or near type 2 diabetes-associated loci NES Normalised enrichment score . Changes in the composition or number of islets as a NKX6-1 NK6 homeobox 1 result of events during development could lead to an al- WGCNA Weighted gene co-expression network analysis tered functional islet mass in later life, increasing risk of type 2 diabetes. Until recently, restricted access to human fetal material Introduction constrained the study of islet development to murine models. However, key differences between human and murine islet Our understanding of the genetic contribution to pathogenesis development , together with the potential of stem cell re- of type 2 diabetes has been greatly facilitated by genome-wide generative approaches to the treatment of diabetes, have mo- association studies (GWASs). These have identified over 100 tivated recent endeavours to differentiate human stem cells genomic regions showing a robust association to disease risk into pancreatic islet-like cells [7–9]. . However, teasing out the biological mechanisms underly- Islet differentiation protocols are rapidly improving [7, 10] ing these disease associations continues to prove difficult, as and are now able to generate functional insulin-producing, most GWAS signals fall outside coding sequences. Broad in- although still somewhat immature, islet-like cells [8, 9]. In this ference across loci has been more successful, demonstrating study, we demonstrate how such cellular models of human from both phenotypic and genomic perspectives the impor- pancreatic islet development can provide insights into the role tance of the pancreatic islet in risk of type 2 diabetes [2, 3]. of monogenic diabetes and type 2 diabetes-associated genes in Most functional follow-up of GWAS signals has in- islet development, and highlight the cellular pathways and mechanisms through which they act. volved studies in adult islets and/or a variety of beta cell 1616 Diabetologia (2018) 61:1614–1622 Methods performed at the Oxford Genomics Centre (Wellcome Centre for Human Genetics, Oxford, UK) as previously de- Generation of human induced pluripotent stem cells Human scribed . RNA sequencing libraries were sequenced to a induced pluripotent stem cell (iPSC) lines from three indepen- mean read depth of 148 (±12) million reads per sample. Reads dent individuals without diabetes were obtained from the were mapped to human genome build hg19, with GENCODE StemBANCC consortium (www.stembancc.org) (see ESM v19 (https://www.gencodegenes.org/releases/19.html)asthe Methods). The generation of lines SB Ad2 and SB Ad3 has transcriptome reference, using STAR v.2.5 , followed by previously been described . A third line, SB Neo1, was gene-level quantification with featureCounts from the generated from commercial fibroblasts obtained from a neo- Subread package v.1.5 (http://subread.sourceforge.net/)  natal donor of European descent with no reported diagnosis of (ESM Methods). diabetes (CC-2509, tissue acquisition number 15819; Lonza, Principal component analysis was used to cluster samples Walkersville, MD, USA). Characterisation of all three lines with those from previously published studies [10, 14]. has been reported elsewhere [10, 11]. All lines were free of Correlation of gene expression patterns across all stages was mycoplasma. calculated using the weighted gene co-expression network analysis (WGCNA) package (v.1.51) in R (v.3.3.2) (ESM Ethics All tissue samples for reprogramming were collected Methods)[15, 16]. with full informed consent. Ethical approval for the StemBANCC study (UK) was received from the National Differential expression analysis Analysis was performed on Research Ethics Service South Central Hampshire A research 15,221 autosomal protein-coding and long intergenic non- ethics committee (REC 13/SC/0179). coding RNA (lincRNA) genes present in Ensembl Genes v88 (http://mar2017.archive.ensembl.org/index.html)with In vitro differentiation of iPSCs towards beta-like cells The more than one count per million in all donors of at least one iPSC lines were cultured in mTeSR1 medium (StemCell differentiation stage (ESM Table 4). Genes were normalised Technologies, Vancouver, BC, Canada) at 37°C under 5% using the voom function within the limma package (v.3.32.5) CO , and passaged as single cells every 3–4 days or when in R . The eBayes function in limma was used for differ- confluent. In vitro differentiation involved the timely addi- ential expression analysis, comparing all the differentiation tion of recombinant growth factors and small molecules to stages with iPSC as the baseline, and adjusting for donor ef- sequentially generate cells representing key developmental fects. We adjusted p values for multiple testing (q values) stages of the endocrine pancreas: definitive endoderm, using the Benjamini–Hochberg method . primitive gut tube, posterior foregut, pancreatic endoderm, To define stage-specific marker genes, differentially endocrine progenitors, endocrine-like cells and beta-like expressed genes (q < 0.01) with an absolute log fold change cells. The differentiation protocol was carried out as de- (FC) > 1 were assigned to the stage in which they were most scribed by Rezania and colleagues withsomemodifica- upregulated compared with the baseline iPSC profile. When tions (ESM Tables 1, 2). All three iPSC lines were differ- the log FC was negative for all contrasted stages, the gene was entiated once, in parallel, using the same culture and dif- assigned to iPSCs (ESM Table 5). For comparison with the ferentiation media (ESM Methods). previously reported protocol , published data were reprocessed in an analogous manner for the stages shared Flow cytometry The efficiency of in vitro differentiation was between the protocols (ESM Methods;ESM Tables 6, 7). evaluated by measuring the expression of stage-specific markers indicative of the development of the endocrine pan- Gene ontology and transcription factor binding motif enrich- creas. For each specific stage, these were: definitive endoderm ment Differentially expressed genes in each stage were tested (SRY-box 17 [SOX17] and octamer-binding transcription fac- for enrichment in gene ontology terms for biological processes tor 4 [OCT4, also known as POU5F1]); pancreatic endoderm using the GOstats package (v. 2.40.0) in R . All genes (NK6 homeobox 1 [NKX6-1] and pancreas/duodenum ho- tested for differential expression were used as background. meobox protein 1 [PDX1]); and endocrine-like cells (NKX6- Significant gene ontology terms (q < 0.05) were retained 1, insulin [INS] and glucagon [GCG]) (ESM Fig. 1). Methods (ESM Table 8). for flow cytometry were as previously described , and For transcription factor enrichment, upstream regulators for details of antibodies are listed in ESM Table 3. the differentially expressed genes were predicted using the iRegulon (v. 1.3) Cytoscape plugin (ESM Methods). RNA extraction, sequencing and quantification Cells were Motifs and chromatin immunoprecipitation (ChIP) sequenc- harvested and RNA extracted using TRIzol Reagent ing tracks were ranked based on the normalised enrichment (ThermoFisher Scientific, Paisley, UK) as per the manufac- score (NES), with only those with an NES > 3 (corresponding turer’s guidelines. Library preparation and sequencing was to a false discovery rate (FDR) of 3–9%) being considered. Diabetologia (2018) 61:1614–1622 1617 Experiment Enriched motifs were then matched to transcription factors Current known to bind them (ESM Table 9). PF PE Previous GT Xie Type 2 diabetes and fasting glucose gene enrichment Enrichment analysis was implemented in two ways: as a Stage hypergeometric test in R (using all genes tested for differential iPSC DE EP expression as background) or using the gene-scoring function DE in MAGENTA  followed by a gene set enrichment analy- GT EN PF sis (GSEA) [22, 23](ESM Methods). -50 PE For the hypergeometric test, we analysed the differentially BLC EP expressed genes from each differentiation stage for enrich- Matured in vivo EN ment in genes mapping to type 2 diabetes or fasting glucose -100 BLC GWAS signals, which were defined as protein-coding and iPSC Matured in vivo lincRNA genes located within specified distance bins (0, 50, 100, 200 or 500 kb) surrounding the credible intervals for -100 0 100 trait-associated loci. Credible intervals were defined by the PC1: 39% variance boundaries of the 99% credible sets of variants  from Fig. 1 Principal component analysis of whole-transcriptome data derived from multiple differentiated human islet-like cell models. Data include all DIAGRAM (96 loci)  and ENGAGE (16 loci) con- stages from our current differentiation protocol (Current), the most mature sortium data, respectively (ESM Table 10). A subset of 15 loci stage of a previously published differentiation protocol (Previous) , was considered to influence type 2 diabetes via beta cell dys- and cells derived via in vivo maturation by Xie and colleagues (Xie) . function; these loci included ones causing hyperglycaemia, The first two principal components (PC1, PC2) have been calculated using normalised gene counts for all stages of all three studies and reduced insulin processing and secretion, and reduced fasting corrected for batch effects. DE, definitive endoderm; GT, primitive gut proinsulin levels [27, 28] (ESM Table 11,ESM Methods). tube; PF, posterior foregut; PE, pancreatic endoderm; EP, endocrine pre- For the analysis with MAGENTA and GSEA, we mapped cursor; EN, endocrine-like cells; BLC, beta-like cells. Stages shown from SNPs from the type 2 diabetes GWAS meta-analysis from the current study are iPSC, DE, GT, PF, PE, EP, EN and BLC. The stage shown from the previously reported study  is EN. The stage shown DIAGRAM (96 loci) , and the ranked list of p values for from Xie and colleagues’ in vivo maturation study is ‘Matured each gene was tested in GSEA (ESM Methods). in vivo’ Results and discussion enriched for terms including ‘regulation of insulin secretion’ −4 −5 (q =2.3 ×10 )and ‘hormone transport’ (q =2.0 ×10 ). Characterising an in vitro-derived model of human beta-like Overall, cells generated in this study, compared with those cells To determine whether the differentiated cells followed previously reported , are more aligned to cells that have normal islet development, we profiled gene expression pat- been further matured in vivo  (the current benchmark for terns across iPSC and seven subsequent developmental stages most functionally mature endocrine pancreas-like cells). This in lines from three independent donors (SB Ad2, SB Ad3 and reveals how advances in differentiation protocols are reflected SB Neo1) differentiated in parallel. Each iPSC line success- in the transcriptome, particularly in the later stages of differ- fully generated cells recapitulating key developmental stages entiation where there is a clear increase in the expression of of the endocrine pancreas as confirmed by the expression of genes essential for beta cell function and identity. This is the known marker genes from developing and adult beta cells case for MAFA, which was completely absent in our previous (ESM Fig. 2). differentiation protocol, and INS, whose high expression indi- Principal component analysis of the transcriptome showed cates the correct differentiation towards the last stage of beta that the beta-like cells generated in the current study clustered cell development. more closely with in vivo-matured islet-like cells  than cells from earlier differentiations (Fig. 1,ESM Fig. 3). Identifying transcriptional networks underlying islet develop- Differential expression analysis comparing transcriptomic pro- ment and diabetes To characterise the transcriptomic land- files obtained from differentiations under current and previous scape of each developmental stage in the in vitro-differentiat- protocols (see Methods) showed increasing divergence with ed cells produced in this study, we assigned significantly dif- differentiation stage (from 17 genes showing differential ex- ferentially expressed genes to the stage at which they were pression in iPSCs to 2095 at the endocrine-like cell stage) most upregulated: if expression peaked in iPSCs, the gene (ESM Table 7). Gene ontology analysis indicated that genes was assigned to that stage (see Methods). We detected 9409 displaying increased expression at the endocrine-like cell stage significantly differentially expressed genes (q< 0.01, absolute (in comparisons of the current vs previous protocols) were log FC > 1) across all stages, ranging in number from 623 in PC2: 17% variance 1618 Diabetologia (2018) 61:1614–1622 the primitive gut tube stage to 2773 in iPSCs (ESM Table 5). factors with less-well studied roles in human islet develop- Known developmental marker genes, such as NEUROG3 in ment. For example, expression of the transcriptional repressor endocrine progenitors and INS in beta-like cells, were correct- REST peaks in the intermediate steps of in vitro differentiation ly assigned to their canonical stages. Gene ontology analysis and declines at the endocrine-like cell and beta-like cell stages, of the sets of differentially expressed genes (ESM Table 8) with reciprocal expression patterns seen among its predicted showed enrichment in biological terms such as ‘hormone targets. These targets include genes encoding neurexins transport’ in endocrine-like cells (q =0.047) and ‘regulation (NRXN1, NRXN2) and subunits of the glutamate receptor −4 of insulin secretion’ in beta-like cells (q =2.0 ×10 ). channels (GRIA1, GRIA2, GRID1, GRIK2) implicated in in- The expression patterns of monogenic diabetes genes can sulin exocytosis [37, 38]. Correlation of gene expression with point towards stages at which disruption of islet development WGCNA assigns REST to the same cluster as TCF7L2 and has long-term consequences for glucose homeostasis. Of 28 other genes from the Wnt signalling pathway, such as TCF7, genes implicated in monogenic or syndromic diabetes , 24 TCF3 and TCF12 . This pathway is important for islet were differentially expressed in at least one stage of the development and is targeted in many in vitro differentiation in vitro-differentiated model. Nine mapped to the latest beta- protocols [8, 9]. These data therefore indicate that REST is like cell stage, but the other 15 showed significant upregula- likely to be an important transcriptional regulator of human tion earlier in differentiation (ESM Table 12). GATA6, for islet development, both in intermediate (pancreatic endoderm, example, was differentially expressed at the definitive endo- endocrine progenitor) and later (endocrine-like cell, beta-like −11 derm stage (log FC = 9.5, q = 7.6 × 10 ), whereas GATA4 cell)  stages of differentiation, as has also been recently was differentially expressed in posterior foregut cells suggested by studies in mice and humans [41, 42]. −11 (log FC = 8.2, q =1.9 × 10 ); the later expression of GATA TCF7L2 maps to the type 2 diabetes-associated locus with 4 could contribute to the less severe phenotype of individuals the largest common effect on disease risk . Analysis of carrying GATA4 vs GATA6 mutations [29, 30]. TCF7L2 targets (as assessed by ChIP sequencing with The differentiation model used in this study also sheds light iRegulon) shows marked enrichment at the posterior foregut on the developmental role of monogenic diabetes genes with stage (NES = 3.4) that mirrors that of TCF7L2 expression −10 lesser described roles. LMNA, for example, encodes a nuclear (log FC = 1.2;q= 8.5 × 10 ). The expression of several oth- membrane protein involved in chromatin structure and nuclear er Wnt family members also peaks at the posterior foregut stability; it has been implicated in the function and develop- stage; these include the coactivator CREBBP, the binding sites ment of many tissues . The diabetes in carriers of the of which are significantly enriched in type 2 diabetes- LMNA mutation is mostly driven by altered adipose tissue de- associated loci , and HHEX, which maps to a prominent position and insulin resistance . However, the profile of type 2 diabetes-risk locus and is implicated in foregut devel- LMNA expression during in vitro islet differentiation (peaking opment . In the developing embryo, cells of the posterior −3 in pancreatic endoderm; log FC = 1.1, q= 3.1 × 10 )may in- foregut can differentiate into liver as well as endocrine pan- dicate an additional impact on islet development . creas . Alleles associated with risk of type 2 diabetes The developmental competence of differentiating cells is in within the TCF7L2 and HHEX loci may influence early ex- part driven by a subset of transcription factors that initiate and pression of these genes, which could affect development in regulate changes in response to external stimuli, as highlight- multiple metabolic tissues. This view is supported by cellular ed by the many monogenic diabetes genes that are also tran- and murine studies indicating that TCF7L2 regulates beta cell scription factors. To identify potential upstream transcriptional development and function , including via indirect effects regulators active at each stage of islet development, we per- in supporting tissues , as well as affecting hepatic function formed a WGCNA and determined the enrichment of tran- . Similarly, Hhex is essential for the differentiation of the scription factor binding motifs and ChIP sequencing signals posterior foregut into the liver in mice , yet is also thought near differentially expressed genes using iRegulon (see to regulate delta cell identity and function in islets . Methods;ESMTable 9). This analysis confirmed the impact Thus, several key functional candidates mapping within of well-established developmental transcriptional regulators type 2 diabetes GWAS signals, in addition to those which such as the monogenic diabetes gene HNF1B, which showed overlap known monogenic diabetes genes, appear to be active iRegulon enrichment of its targets at the primitive gut tube during this early critical window of pancreatic development. stage (NES 3.0–5.7 [see Methods]). Some of these HNF1B Studying these and other diabetes-relevant genes in stem cell- targets also have known effects on pancreas development derived models can help to decipher the role of multiorgan (SMAD7 , ID2 ), on mature islet function and on the developmental effects on pathogenesis of diabetes. By inte- development of other tissues that also arise from the gut tube grating the differential expression data with genomic annota- (GGCX). tions on transcription factor binding and clustering of longitu- Analysis of the sets of stage-specific differentially dinal expression, we identified novel potential regulators or- expressed genes also highlighted the targets of transcription chestrating gene expression patterns within the different Diabetologia (2018) 61:1614–1622 1619 developmental stages. Such transcriptomic analysis can also Fig. 2a). This enrichment remained significant (q =0.001) if illuminate the mechanisms of action for monogenic diabetes the GWAS genes also implicated in monogenic diabetes genes and inform the search for novel MODY genes that in- (ESM Table 12) were excluded. Using a complementary fluence the same pathways. GSEA approach that ranked the strength of differential ex- pression of each gene (in q value) per stage, we compared Developing and mature cells are enriched in genes within the most differentially expressed genes at each stage for en- type 2 diabetes-associated loci Most of the more than 100 richment among type 2 diabetes GWAS loci; this analysis type 2 diabetes susceptibility loci identified to date map highlighted the beta-like cell stage (q = 0.033, Fig. 2a). This to non-coding regions of the genome and are likely to exert enrichment was no longer significant (q = 0.151) after mono- their effects through altered regulation of nearby genes. We genic diabetes genes had been excluded. examined the transcriptomic data for evidence of develop- As an additional analytical approach, we performed a mental stage-specific enrichment of genes near these loci. hypergeometric test for enrichment in the same set of 117 We first concentrated on genes whose coding sequence was type 2 diabetes credible interval genes (see Methods). As at least partly contained within 99% credible intervals from opposed to the GSEA method above, this analysis does not type 2 diabetes GWAS fine-mapping efforts on the basis that consider the strength of differential expression (or of associ- these represented a set of genes likely to be substantially ation with type 2 diabetes) above the significance threshold. enriched for type 2 diabetes effector transcripts (see This test again demonstrated that genes showing differential Methods). Of the 117 genes so defined, most (86; 73%) expression at the beta-like cell stage were enriched (compared showed differential expression that peaked before the final with background) for location within type 2 diabetes credible beta-like cell stage (ESM Table 13); the stages of maximal intervals (permuted p value =0.049; Fig. 2b). Excluding the differential expression were widely distributed. GSEA, which monogenic diabetes genes, and those that fell in the same considers the strength of association at type 2 diabetes GWAS credible interval, from the differentially expressed genes at signals (see Methods), demonstrated enrichment of the type 2 each stage removed the significance of the beta-like cells diabetes GWAS loci with largest effect for differentially (permuted p value =0.302). We repeated the enrichment test expressed genes at the posterior foregut stage (q = 0.002, using a subset of 15 type 2 diabetes GWAS loci for which the Stage iPSC Ranked T2D GWAS list Ranked differentially DE expressed genes GT PF PE EP 4 EN BLC Fasting glucose T2D (beta cell) T2D (all) Fig. 2 Both developing and mature islet-like cells are enriched for genes tested for all differentially expressed genes per stage in the 96 type 2 within type 2 diabetes-associated loci. (a) Results from the GSEA. SNPs diabetes credible intervals [T2D (all)] from DIAGRAM  and the 16 from the type 2 diabetes GWAS meta-analysis from DIAGRAM (96 loci) fasting glucose credible intervals (Fasting glucose) from ENGAGE   were mapped to genes, and type 2 diabetes association scores were (ESM Table 10), and for all differentially expressed genes in only phys- calculated for each gene using MAGENTA. Two complementary analy- iological type 2 diabetes loci [T2D (beta cell)] (ESM Table 11). We ses were performed: enrichment of all genes ordered by their MAGENTA consider beta cell function loci as 15 loci influencing hyperglycaemia, scores in sets of differentially expressed genes for each stage (Ranked beta cell function and insulin processing [26, 27]. The y-axis represents T2D GWAS list), and enrichment of differentially expressed genes per the results of the hypergeometric test in permuted p values (−log ). The stage (ordered by q value) in significant (p < 0.05 by MAGENTA) gene horizontal grey dashed line marks the 5% significance threshold. T2D, scores (ranked differentially expressed genes). The y-axis represents the type 2 diabetes; DE, definitive endoderm; GT, primitive gut tube; PF, results of the GSEA in FDR-adjusted p values (q values, −log ). The posterior foregut; PE, pancreatic endoderm; EP, endocrine precursor; horizontal grey dashed line marks the 5% significance threshold. (b) EN, endocrine-like cells; BLC, beta-like cells Results for the hypergeometric enrichment analysis. Enrichment was – log permuted (p value) – log (q value) 10 10 1620 Diabetologia (2018) 61:1614–1622 evidence from physiological studies points most emphatically in vitro systems . Stem cell-derived islets may also serve to risk of type 2 diabetes mediated via reduced insulin secre- as a cost-effective platform for drug screening in research into tion (ESM Table 11)[27, 28]. In this analysis, enrichment for treatment of diabetes, and could provide material for trans- genes differentially expressed at the beta-like cell stage be- plant into individuals with diabetes [8, 9]. came more significant (permuted p value =0.007; Fig. 2b). Acknowledgements We thank the High-Throughput Genomics Group at This enrichment was reduced (but not eliminated; permuted p the Wellcome Centre for Human Genetics (University of Oxford, UK) for value =0.03) after excluding the monogenic diabetes genes generation of the sequencing data. and those within the same credible interval. Using the same approach of sampling from the hypergeometric distribution, Data availability Sequence data have been deposited at the European Genome-phenome Archive (EGA), which is hosted by the European we also detected enrichment for genes mapping to credible Bioinformatics Institute (EBI) and the Centre for Genomic Regulation intervals for 16 loci significantly associated with fasting glu- (CRG), under accession number EGAS00001002721, and are also avail- cose (permuted p value =0.0002; Fig. 2b). Earlier stages of able on request from the authors. differentiation did not show significant enrichment for genes Funding The research leading to these results has received funding from within type 2 diabetes or fasting glucose credible intervals. the Innovative Medicines Initiative Joint Undertaking (IMI JU) under Nevertheless, the assignment of differentially expressed Grant Agreement number 115439 (StemBANCC), resources of which genes to a specific stage may lead to a wide distribution of are composed of financial contribution from the European Union’s signal that dilutes the power to detect significant enrichment Seventh Framework Programme (FP7/2007-2013) and EFPIA compa- nies in kind contribution. This publication reflects only the authors’ at stages before the beta-like cell stage. views, and neither the IMI JU, the EFPIA nor the European Type 2 diabetes-associated signals falling in non-coding Commission is liable for any use that may be made of the information regions have a presumed regulatory function: some may contained therein. This work was also supported by the Wellcome Trust map to tissue-specific enhancers acting some distance away (098381, 106130, 090532, 203141), Medical Research Council (MR/ L020149/1, BRR00030) and National Institute for Health Research from their effector transcripts . However, consistent with (NIHR) Oxford Biomedical Research Centre Programme. ALG is a observations that most regulatory GWAS effects operate at Wellcome Trust Senior Fellow in Basic Biomedical Research (95101 relatively short distances , we found attenuation of these and 200837). MIM is a Wellcome Trust Senior Investigator. NLB was a enrichment signals as we extended the analyses to include Naomi Berrie Fellow in Diabetes Research. MvdB was supported by a Novo Nordisk postdoctoral fellowship run in partnership with the genes mapping at increasing distance from the credible inter- University of Oxford. NLB and MvdB are now employees of Novo vals (see Methods), both for genes in all type 2 diabetes cred- Nordisk (although all experimental work was carried out under employ- ible intervals and for the subset implicated in beta cell function ment at the University of Oxford). MPA is supported by a Wellcome Trust (ESM Fig. 4). PhD studentship (H5R00430). The study sponsor was not involved in the design of the study; the collection, analysis and interpretation of data; the The notable overlap between monogenic diabetes genes and writing of the report; or the decision to submit the report for publication. those mapping within type 2 diabetes-associated loci supports the hypothesis that some component of type 2 diabetes suscep- Duality of interest MH and CH are employees of, and shareholders in, tibility arises through impairment of islet development , con- Novo Nordisk. The remaining authors declare that there is no duality of interest associated with this manuscript. cretely in the posterior foregut stage. The final stage in the islet development model (featuring cells expressing genes encoding Contribution statement NLB, CH and MvdB conceived the study. CH and the machinery to support glucose-stimulated insulin secretion) NLB designed and performed the differentiation experiments. MvdB and is also enriched for genes mapping to GWAS signals for both MPA designed and performed the data analyses. ALG and MH gave con- type 2 diabetes and fasting glucose. These data are consistent ceptual advice and edited the manuscript. All the authors interpreted the data. MPA, NLB, MvdB, MIM, AWA and CH wrote the manuscript. All the with the concept that type 2 diabetes-associated loci act both on authors revised the manuscript and approved the final version submitted for the adult islet and during earlier developmental stages. publication. NLB and MvdB are the guarantors of this work. In summary, this study demonstrates how characterisation of gene expression during human islet differentiation can Open Access This article is distributed under the terms of the Creative identify potential novel transcriptional regulators of the differ- Commons Attribution 4.0 International License (http:// entiation process, and provide insights into developmental creativecommons.org/licenses/by/4.0/), which permits unrestricted use, aspects underlying inherited predisposition to type 2 diabetes. distribution, and reproduction in any medium, provided you give appro- priate credit to the original author(s) and the source, provide a link to the Further refinement of in vitro models of endocrine pancreas Creative Commons license, and indicate if changes were made. development will allow more detailed interrogation of the genes and pathways influencing islet development and func- tion in humans. Mechanistic analyses of the contribution of candidate regulators of islet development to long-term islet References function is enhanced by recent advances in clustered regularly interspaced short palindromic repeats- (CRISPR-) based 1. Fuchsberger C, Flannick J, Teslovich TM et al (2016) The genetic approaches that allow their experimental manipulation in architecture of type 2 diabetes. Nature 536:41–47 Diabetologia (2018) 61:1614–1622 1621 2. Benner C, van der Meulen T, Cacéres E, Tigyi K, Donaldson CJ, 24. Maller JB, McVean G, Byrnes J et al (2012) Bayesian refinement of association signals for 14 loci in 3 common diseases. Nat Genet 44: Huising MO (2014) The transcriptional landscape of mouse beta cells compared to human beta cells reveals notable species differ- 1294–1301 ences in long non-coding RNA and protein-coding gene expres- 25. Scott RA, Scott LJ, Mägi R et al (2017) An expanded genome-wide sion. BMC Genomics 15:620 association study of type 2 diabetes in Europeans. Diabetes 66: 3. van De Bunt M, Manning Fox JE, Dai X et al (2015) Transcript 2888–2902 expression data from human islets links regulatory signals from 26. Horikoshi M, M gi R, van de Bunt M et al (2015) Discovery and genome-wide association studies for type 2 diabetes and glycemic fine-mapping of glycaemic and obesity-related trait loci using high- traits to their downstream effectors. PLoS Genet 11:e1005694 density imputation. PLoS Genet 11:e1005230 4. Travers ME, Mackay DJG, Nitert MD et al (2013) Insights into the 27. Dimas AS, Lagou V, Barker A et al (2014) Impact of type 2 molecular mechanism for type 2 diabetes susceptibility at the diabetes susceptibility variants on quantitative glycemic traits KCNQ1 locus from temporal changes in imprinting status in human reveals mechanistic heterogeneity. Diabetes 63:2158–2171 islets. Diabetes 62:987–992 28. Wood AR, Jonsson A, Jackson AU et al (2017) A genome-wide 5. Schwitzgebel VM (2014) Many faces of monogenic diabetes. J association study of IVGTT-based measures of first phase insulin Diabetes Investig 5:121–133 secretion refines the underlying physiology of type 2 diabetes var- 6. Jennings RE, Berry AA, Strutt JP et al (2015) Human pancreas iants. Diabetes 66:2296–2309 development. Development 142:3126–3137 29. Lango Allen H, Flanagan SE, Shaw-Smith C et al (2011) GATA6 7. Bruin JE, Erener S, Vela J et al (2014) Characterization of haploinsufficiency causes pancreatic agenesis in humans. Nat polyhormonal insulin-producing cells derived in vitro from human Genet 44:20–22 embryonic stem cells. Stem Cell Res 12:194–208 30. Shaw-Smith C, De Franco E, Lango Allen H et al (2014) GATA4 8. Pagliuca FW, Millman JR, Gürtler M et al (2014) Generation of mutations are a cause of neonatal and childhood-onset diabetes. functional human pancreatic β cells in vitro. Cell 159:428–439 Diabetes 63:2888–2894 9. Rezania A, Bruin JE, Arora P et al (2014) Reversal of diabetes with 31. Zuela N, Bar DZ, Gruenbaum Y (2012) Lamins in development, insulin-producing cells derived in vitro from human pluripotent tissue maintenance and stress. EMBO Rep 13:1070–1078 stem cells. Nat Biotechnol 32:1121–1133 32. Boschmann J, Engeli JM, Engeli S et al (2010) LMNA mutations, 10. van de Bunt M, Lako M, Barrett A et al (2016) Insights into islet skeletal muscle lipid metabolism, and insulin resistance. J Clin development and biology through characterization of a human Endocrinol Metab 95:1634–1643 iPSC-derived endocrine pancreas model. Islets 8:83–95 33. Kowluru A (2000) Evidence for the carboxyl methylation of nucle- 11. Yang C, Xu Y, Yu M et al (2017) Induced pluripotent stem cell ar lamin-B in the pancreatic beta cell. Biochem Biophys Res modelling of HLHS underlines the contribution of dysfunctional Commun 268:249–254 NOTCH signalling to impaired cardiogenesis. Hum Mol Genet 34. Smart NG, Apelqvist SA, Gu X et al (2006) Conditional expression 26:3031–3045 of Smad7 in pancreatic beta cells disrupts TGF-beta signaling and 12. Dobin A, Gingeras TR (2015) Mapping RNA-seq reads with induces reversible diabetes mellitus. PLoS Biol 4:e39 STAR. Curr Protoc Bioinformatics 51:11.14.1–11.14.19 35. Hua H, Zhang Y-Q, Dabernat S et al (2006) BMP4 regulates pan- 13. Liao Y, Smyth GK, Shi W (2014) featureCounts: an efficient gen- creatic progenitor cell expansion through Id2. J Biol Chem 281: eral purpose program for assigning sequence reads to genomic fea- 13574–13580 tures. Bioinformatics 30:923–930 36. Ferron M, Lacombe J, Germain A, Oury F, Karsenty G (2015) 14. Xie R, Everett LJ, Lim H-W et al (2013) Dynamic chromatin remod- GGCX and VKORC1 inhibit osteocalcin endocrine functions. J eling mediated by Polycomb proteins orchestrates pancreatic differ- Cell Biol 208:761–776 entiation of human embryonic stem cells. Cell Stem Cell 12:224–237 37. Mosedale M, Egodage S, Calma RC, Chi N-W, Chessler SD (2012) 15. Langfelder P, Horvath S (2008) WGCNA: an R package for weight- Neurexin-1α contributes to insulin-containing secretory granule ed correlation network analysis. BMC Bioinformatics 9:559 docking. J Biol Chem 287:6350–6361 16. R Core Team (2017) R: A language and environment for statistical 38. Wu Z-Y, Zhu L-J, Zou N et al (2012) AMPA receptors regulate exo- computing. Vienna, Austria. www.r-project.org cytosis and insulin release in pancreatic β cells. Traffic 13:1124–1139 17. Ritchie ME, Phipson B, Wu D et al (2015) limma powers differen- 39. Cadigan KM, Waterman ML (2012) TCF/LEFs and Wnt signaling tial expression analyses for RNA-sequencing and microarray stud- in the nucleus. Cold Spring Harb Perspect Biol 4:a007906 ies. Nucleic Acids Res 43:e47 40. Atouf F, Czernichow P, Scharfmann R (1997) Expression of neu- 18. Benjamini Y, Hochberg Y (1995) Controlling the false discovery ronal traits in pancreatic beta cells. J Biol Chem 272:1929 rate: a practical and powerful approach to multiple testing. Source J 41. Martin D, Kim Y-H, Sever D, Mao C-A, Haefliger J-A, Grapin- R Stat Soc Ser B 57:289–300 Botton A (2015) REST represses a subset of the pancreatic endo- 19. Falcon S, Gentleman R (2007) Using GOstats to test gene lists for crine differentiation program. Dev Biol 405:316–327 GO term association. Bioinformatics 23:257–258 42. Solimena M, Schulte AM, Marselli L et al (2018) Systems biology 20. Verfaillie A, Imrichová H, Van de Sande B et al (2014) iRegulon: of the IMIDIA biobank from organ donors and pancreatectomised from a gene list to a gene regulatory network using large motif and patients defines a novel transcriptomic signature of islets from in- track collections. PLoS Comput Biol 10:e1003731 dividuals with type 2 diabetes. Diabetologia 61:641–657 21. Segrè AV, Groop L, Mootha VK et al (2010) Common inherited 43. Morris A, Voight B, Teslovich T (2012) Large-scale association variation in mitochondrial genes is not enriched for associations analysis provides insights into the genetic architecture and patho- with type 2 diabetes or related glycemic traits. PLoS Genet 6: physiology of type 2 diabetes. Nat Genet 44:981–990 e1001058 44. Arterbery AS, Bogue CW (2016) Hhex is necessary for the hepatic 22. Subramanian A, Tamayo P, Mootha VK et al (2005) Gene set en- differentiation of mouse ES cells and acts via Vegf signaling. PLoS richment analysis: a knowledge-based approach for interpreting One 11:e0146806 genome-wide expression profiles. Proc Natl Acad Sci U S A 102: 45. Gittes GK (2009) Developmental biology of the pancreas: a com- 15545–15550 prehensive review. Dev Biol 326:4–35 23. Mootha VK, Lindgren CM, Eriksson K-F et al (2003) PGC-1α- 46. Le Bacquer O, Shu L, Marchand M et al (2011) TCF7L2 splice responsive genes involved in oxidative phosphorylation are coordi- variants have distinct effects on β-cell turnover and function. Hum nately downregulated in human diabetes. Nat Genet 34:267–273 Mol Genet 20:1906–1915 1622 Diabetologia (2018) 61:1614–1622 47. Sakhneny L, Rachi E, Epshtein A et al (2018) Pancreatic pericytes 50. Gaulton KJ, Ferreira T, Lee Yet al (2015) Genetic fine mapping and genomic annotation defines causal mechanisms at type 2 diabetes support beta-cell function in a Tcf7l2-dependent manner. Diabetes 67:437–447 susceptibility loci. Nat Genet 47:1415–1425 48. Boj SF, Van Es JH, Huch M et al (2012) Diabetes risk gene and Wnt 51. Beer NL, Gloyn AL (2016) Genome-edited human stem cell- effector Tcf7l2/TCF4 controls hepatic response to perinatal and derived beta cells: a powerful tool for drilling down on type 2 adult metabolic demand. Cell 151:1595–1607 diabetes GWAS biology. F1000Res 5:F1000 Faculty Rev-1711 49. Zhang J, McKenna LB, Bogue CW, Kaestner KH (2014) The dia- betes gene Hhex maintains δ-cell differentiation and islet function. Genes Dev 28:829–834
Diabetologia – Springer Journals
Published: Apr 19, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera