# paper.pdf ## Metadata - PDFFormatVersion=1.4 - Language=en-US - IsLinearized=false - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - Author=Jones K. Ofori - CreationDate=D:20260427181201+05'30' - Creator=Springer - Custom.CrossMarkDomains[1]=springer.com - Custom.CrossMarkDomains[2]=springerlink.com - Custom.doi=10.1038/s42255-026-01498-9 - ModDate=D:20260427181314+05'30' - Subject=Nature Metabolism, doi:10.1038/s42255-026-01498-9 - Title=Cell-specific DNA methylation in human alpha and beta cells regulates gene expression in type 2 diabetes - dc:format=application/pdf - dc:identifier=doi:10.1038/s42255-026-01498-9 - dc:publisher=Springer US - dc:description=Nature Metabolism, doi:10.1038/s42255-026-01498-9 - dc:title=Cell-specific DNA methylation in human alpha and beta cells regulates gene expression in type 2 diabetes - dc:creator=Jones K. Ofori,Sabrina Ruhrmann,Axel Lindström,Alexander Perfilyev,Melina Martin,Alexandros Karagiannopoulos,Lucia Scisciola,Katja Kost,Josefine Jönsson,Åsa Nilsson,Boris Kantor,Monika Dudenhöffer-Pfeifer,Marianne G. Rots,Anna Wendt,Tina Rönn,Lena Eliasson,Karl Bacos,Charlotte Ling - prism:doi=10.1038/s42255-026-01498-9 - prism:url=http://dx.doi.org/10.1038/s42255-026-01498-9 - prism:aggregationtype=journal - prism:publicationname=Nature Metabolism - prism:copyright=© 2026, The Author(s), under exclusive licence to Springer Nature Limited - prism:issn=2522-5812 - pdfx:doi=10.1038/s42255-026-01498-9 - pdfx:crossmarkdomains=springer.comspringerlink.com - crossmark:majorversiondate=2010-04-23 - crossmark:crossmarkdomainexclusive=True - crossmark:doi=10.1038/s42255-026-01498-9 - crossmark:crossmarkdomains=springer.comspringerlink.com - xmp:creatortool=Springer - xmp:modifydate=2026-04-27T18:13:14+05:30 - xmp:createdate=2026-04-27T18:12:01+05:30 - xmp:metadatadate=2026-04-27T18:13:14+05:30 - xmprights:marked=True - jav:journal_article_version=VoR - xmpmm:documentid=uuid:8316de5a-ed00-4563-b4e5-125adf0bc6c3 - xmpmm:instanceid=uuid:8bfe9779-101e-481d-952e-d73371185334 - xmpmm:renditionclass=default - xmpmm:versionid=1 - xmpmm:history=converteduuid:b3619b80-fde7-4538-bdfc-a9964dba31a2converted to PDF/A-2bpdfToolbox2026-04-27T18:13:13+05:30 - pdfaid:part=2 - pdfaid:conformance=B - sn:authorinfo=Jones K. Oforihttp://orcid.org/0000-0001-6484-1544Axel Lindströmhttp://orcid.org/0009-0008-1641-0406Alexandros Karagiannopouloshttp://orcid.org/0000-0001-8458-1065Lucia Scisciolahttp://orcid.org/0000-0003-3636-4298Josefine Jönssonhttp://orcid.org/0000-0003-0709-2828Lena Eliassonhttp://orcid.org/0000-0002-6467-5029Karl Bacoshttp://orcid.org/0000-0002-2461-9073Charlotte Linghttp://orcid.org/0000-0003-0587-7154 - pdfaextension:schemas=http://ns.adobe.com/xap/1.0/mm/xmpMMXMP Media Management SchemainternalUUID based identifier for specific incarnation of a documentInstanceIDURIinternalThe common identifier for all versions and renditions of a document.OriginalDocumentIDURIinternalA reference to the original document from which this one is derived. It is a minimal reference; missing components can be assumed to be unchanged. For example, a new version might only need to specify the instance ID and version number of the previous version, or a rendition might only need to specify the instance ID and rendition class of the original.DerivedFromResourceRefIdentifies a portion of a document. This can be a position at which the document has been changed since the most recent event history (stEvt:changed). For a resource within an xmpMM:Ingredients list, the ResourceRef uses this type to identify both the portion of the containing document that refers to the resource, and the portion of the referenced resource that is referenced.http://ns.adobe.com/xap/1.0/sType/Part#stPartParthttp://ns.adobe.com/pdf/1.3/pdfAdobe PDF SchemainternalA name object indicating whether the document has been modified to include trapping informationTrappedTexthttp://ns.adobe.com/xap/1.0/t/pg/xmpTPgXMP Paged-TextinternalXMP08 Spec: An ordered array of plate names that are needed to print the document (including any in contained documents).PlateNamesSeq TextinternalA structure containing the characteristics of a font used in a documentFontsBag FontA structure containing the characteristics of a font used in a document.http://ns.adobe.com/xap/1.0/sType/Font#stFntFonthttp://ns.adobe.com/pdfx/1.3/pdfxpdfxinternalID of PDF/X standardGTS_PDFXVersionTextinternalConformance level of PDF/X standardGTS_PDFXConformanceTextinternalCompany creating the PDFCompanyTextinternalDate when document was last modifiedSourceModifiedTextexternalMirrors crossmark:CrosMarkDomainsCrossMarkDomainsSeq TextinternalMirrors crossmark:DOIdoiTexthttp://www.aiim.org/pdfa/ns/id/pdfaidPDF/A ID SchemainternalPart of PDF/A standardpartIntegerinternalAmendment of PDF/A standardamdTextinternalConformance level of PDF/A standardconformanceTexthttp://crossref.org/crossmark/1.0/crossmarkcrossmarkinternalCrossMarkDomainsCrossMarkDomainsSeq TextinternalCrossmarkDomainExclusiveCrossmarkDomainExclusiveTextinternalUsual same as prism:doiDOITextexternalThe date when a publication was published.MajorVersionDateTexthttp://prismstandard.org/namespaces/basic/2.0/prismPrismexternalThe aggregation type specifies the unit of aggregation for a content collection. Comment PRISM recommends that the PRISM Aggregation Type Controlled Vocabulary be used to provide values for this element. Note: PRISM recommends against the use of the #other value currently allowed in this controlled vocabulary. In lieu of using #other please reach out to the PRISM group at info@prismstandard.org to request addition of your term to the Aggregation Type Controlled Vocabulary. aggregationTypeTextexternalCopyright copyrightTextexternalThe Digital Object Identifier for the article. The DOI may also be used as the dc:identifier. If used as a dc:identifier, the URI form should be captured, and the bare identifier should also be captured using prism:doi. If an alternate unique identifier is used as the required dc:identifier, then the DOI should be specified as a bare identifier within prism:doi only. If the URL associated with a DOI is to be specified, then prism:url may be used in conjunction with prism:doi in order to provide the service endpoint (i.e. the URL). doiTextexternalISSN for an electronic version of the issue in which the resource occurs. Permits publishers to include a second ISSN, identifying an electronic version of the issue in which the resource occurs (therefore e(lectronic)Issn. If used, prism:eIssn MUST contain the ISSN of the electronic version. See prism:issn. issnTextexternalTitle of the magazine, or other publication, in which a resource was/will be published. Typically this will be used to provide the name of the magazine an article appeared in as metadata for the article, along with information such as the article title, the publisher, volume, number, and cover date. Note: Publication name can be used to differentiate between a print magazine and the online version if the names are different such as “magazine” and “magazine.com.” publicationNameTextexternalThis element provides the url for an article or unit of content. The attribute platform is optionally allowed for situations in which multiple URLs must be specified. PRISM recommends that a subset of the PCV platform values, namely “mobile” and “web”, be used in conjunction with this element. NOTE: PRISM recommends against the use of the #other value allowed in the PRISM Platform controlled vocabulary. In lieu of using #other please reach out to the PRISM group at prism-wg@yahoogroups.com to request addition of your term to the Platform Controlled Vocabulary. urlTexthttp://springernature.com/ns/xmpExtensions/2.0/snSpringer Nature ORCID SchemaexternalAuthor information: contains the name of each author and his/her ORCID (ORCiD: Open Researcher and Contributor ID). An ORCID is a persistent identifier (a non-proprietary alphanumeric code) to uniquely identify scientific and other academic authors.authorInfoBag AuthorInformationSpecifies the types of author information: name and ORCID of an author.http://springernature.com/ns/xmpExtensions/2.0/authorInfo/authorAuthorInformationGives the name of an author.nameTextGives the ORCID of an author.orcidURIhttp://www.niso.org/schemas/jav/1.0/javNISOexternalValues for Journal Article Version are one of the following: AO = Author’s Original SMUR = Submitted Manuscript Under Review AM = Accepted Manuscript P = Proof VoR = Version of Record CVoR = Corrected Version of Record EVoR = Enhanced Version of Recordjournal_article_versionClosed Choice of Text ## Contents ### Page 1 Nature Metabolism | Volume 8 | April 2026 | 957–980 957nature metabolismhttps://doi.org/10.1038/s42255-026-01498-9ResourceCell-specific DNA methylation in human alpha and beta cells regulates gene expression in type 2 diabetesJones K. Ofori 1,9 , Sabrina Ruhrmann1,9 , Axel Lindström 1,2 , Alexander Perfilyev3 , Melina Martin1 , Alexandros Karagiannopoulos 4 , Lucia Scisciola 5 , Katja Kost1 , Josefine Jönsson 1 , Åsa Nilsson4 , Boris Kantor6 , Monika Dudenhöffer-Pfeifer4 , Marianne G. Rots7 , Anna Wendt8 , Tina Rönn1 , Lena Eliasson 2,8 , Karl Bacos 1,10 & Charlotte Ling 1,10Epigenome-wide studies of pancreatic islets provide valuable insights into type 2 diabetes (T2D) but lack methylomes from individual cell types. Here we show changes to alpha and beta cell-specific methylomes and transcriptomes from people with or without T2D, using whole-genome bisulfite sequencing and RNA sequencing. We discover 22,544 differentially methylated regions annotated to 7,975 genes in alpha versus beta cells, such as INS, GCG, PDX1 and PCSK1, with ~50% showing differential expression. CRISPR–dCas9–DNMT3A-based epigenetic editing increases INS and THDNA methylation, while CRISPR–dCas9–TET1-based editing decreases GCGmethylation, each altering INS, TH or GCG expression and content in beta cells. Pre-T2D/T2D-associated differentially methylated regions in alpha and beta cells overlap 12–18% of T2D-associated genome-wide association study candidates. Additionally, ONECUT2 is epigenetically upregulated in beta cells from people with pre-T2D/T2D and elevated in male Goto-Kakizaki rat islets. ONECUT2 overexpression in beta cells/islets downregulates gene sets impacting insulin secretion and glucose homeostasis, and reduces mitochondrial activity, ATP/ADP ratio and insulin secretion. We also provide ‘alpha-beta-methylome’ (https://alpha-beta-methylome.serve.scilifelab. se/app/alpha-beta-methylome/), a resource exploring T2D, age and sex associations on methylation, highlighting cell-specific epigenetic regulation and dysfunctions contributing to T2D. T2D, characterized by hyperglycaemia driven by impaired insulin secretion from pancreatic islets and peripheral insulin resistance1,2 , is rapidly increasing worldwide3 . Pancreatic islets contain several cell types controlling glucose homeostasis. In the fasted state, alpha cells secrete glucagon to raise plasma glucose levels, while in the fed state, beta cells secrete insulin to decrease glucose levels. Impaired insulin secretion is required for T2D development, but excess glucagon also contributes to hyperglycaemia 1,4 . Genome-wide association studies (GWAS) identified single-nucleotide polymorphisms (SNPs) linked to T2D and islet dys- function2 . Additionally, epigenetic mechanisms, linking the environ- ment and genome, are of importance for T2D5,6 . Indeed, earlier studies discovered alterations in DNA methylation in pancreatic islets, liver, adipose tissue and muscle from people with T2D versus controls 7–19 . However, prior studies analysed methylomes in whole tissues, con- taining different cell types, although it is established that epigeneticsReceived: 19 August 2025 Accepted: 3 March 2026Published online: 24 April 2026Check for updatesA full list of affiliations appears at the end of the paper. e-mail: charlotte.ling@med.lu.se ### Page 2 Nature Metabolism | Volume 8 | April 2026 | 957–980 958Resource https://doi.org/10.1038/s42255-026-01498-9factor (TF) binding24 and T2D genetics2 . Finally, we developed the web tool alpha-beta-methylome (https://alpha-beta-methylome.serve. scilifelab.se/app/alpha-beta-methylome/), a comprehensive resource for investigating the impact of age, sex and T2D on DNA methylation and gene expression in human alpha and beta cells.ResultsWGBS and RNA-seq of human alpha and beta cellsUsing MARIS for cell sorting of human islets23 , we obtained on aver- age 322,627 ± 73,055 alpha cells and 311,550 ± 61,250 beta cells from 24 donors with or without T2D (Fig. 1a and Extended Data Fig. 1a). Characteristics of all donors in this study (17 non-diabetic controls and 7 pre-T2D/T2D) and donors for each analysis are presented in Supplementary Table 1. WGBS analysed genome-wide DNA methyla- tion at base-pair resolution on alpha and beta cell samples where we had enough DNA (Fig. 1b). Supplementary Table 2 reports sequenc- ing information for WGBS data passing quality control (n = 10, alpha cells; n = 14, beta cells). After deduplication, ~84% of reads mapped uniquely to GRCh38. After quality control, the transcriptome was ana- lysed by RNA-seq in alpha and beta cells from 22 donors (Fig. 1b and Supplementary Table 1b).Global methylome of human alpha and beta cellsWe next characterized the global methylome in alpha and beta cells, from non-diabetic donors and with WGBS data available from both cell types (Supplementary Tables 1 and 2). DNA methylation of 27.54 million cytosine-phosphate-guanine (CpG) sites was profiled and the distribution of methylation levels through different genomic regions visualized with dmrseq25 . Globally, methylation is bimodal, with most CpGs hypermethylated (right peaks) or unmethylated (left peaks), resulting in average methylation levels of 76.1% and 77.2% in alpha and beta cells, respectively (Fig. 1c). We also analysed the genome-wide methylation levels in relation to different genomic regions and CG density (Extended Data Fig. 1b and Fig. 1d,e). Introns, 3’ untranslated regions (UTRs) and intergenic regions exhibited the highest methyla- tion levels, whereas promoters and 5’ UTRs had the lowest degree of methylation (Fig. 1d). Based on CG density, CpG islands had the lowest methylation level, while Open Sea and Shelves had the highest (Fig. 1e). We then used RNA-seq data from alpha and beta cells (n = 16; Supplementary Table 1b) to explore whether methylation levels in controls cell-type-specific gene expression. Hence, studies focusing on DNA methylation in individual cell types in tissues from people with T2D are needed. To our knowledge, a few studies have analysed cell-specific methylomes in humans20 , but none were done in tissues from people with T2D. A better understanding of epigenetic and molecular defects in individual cell types in islets from people with or at risk of T2D could identify previously unrecognized aetiopathoge- netic mechanisms and drug targets. Additionally, studies exploring whether altering DNA methylation in regulatory regions of alpha and beta cell-specific genes causes expression changes and alters hormone secretion are desirable. Single-cell RNA sequencing (scRNA-seq) and single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq) can identify transcriptome and open chromatin differences in individual cell types of a tissue 21 . However, most single-cell studies are based on smaller numbers of cells from each tissue, and individual, thus rep- resenting only minute parts of a tissue. scRNA-seq studies identified differentially expressed genes (DEGs) in islets from donors with T2D versus controls 21,22 . The DEGs identified in these studies are largely non-overlapping, potentially because many scRNA-seq protocols cap- ture only 5–40% of transcripts in a cell, have low power due to analysis of smaller cohorts, and/or are affected by selection biases for certain cell types 22 . Methods capturing more transcripts from each cell, and more individual cells from each donor, may resolve these issues. To fill these knowledge gaps, we used Method for Analyzing RNA following Intracellular Sorting (MARIS) 23 to sort larger numbers (~300,000 per donor) of alpha and beta cells from human pancre- atic islets of donors with or without T2D, followed by whole-genome bisulfite sequencing (WGBS), the most comprehensive method to study methylation genome wide, and RNA-seq. Firstly, we aimed to character- ize the global methylome and transcriptome of human alpha and beta cells and use epigenetic editing to investigate causative impacts of DNA methylation on transcription of alpha and beta cell-specific genes. Sec- ondly, we aimed to identify previously unrecognized candidate genes contributing to impaired insulin or glucagon secretion by studying the impact of pre-T2D/T2D on methylomes and transcriptomes in human alpha and beta cells, followed by functional validation by manipulation of identified T2D-associated candidates in human islets and cell lines together with analyses of diabetic Goto-Kakizaki (GK) rats. The methy- lomes in alpha and beta cells were also integrated with transcriptionFig. 1 | Analysis of global DNA methylome in human alpha and beta cells.a, Workflow for the current study. Human pancreatic islets from 24 donors with or without T2D were used for sorting of alpha and beta cells after staining for glucagon and insulin. WGBS and RNA-seq were used to analyse the methylome and transcriptome in sorted alpha and beta cells. Bioinformatic analyses were used to identify differential DNA methylation and gene expression. Follow-up experiments were done in human islets, EndoC-βH1 cells, INS-1 beta cells and GK rats, by epigenetic editing, analysing gene expression, insulin secretion and metabolism. b, Filtering islet samples for WGBS and RNA-seq. The Lund University Diabetes Centre (LUDC) islet sorting cohort consists of alpha and beta cells sorted from 24 human islet donors. Based on samples with enough DNA, WGBS was performed on sorted alpha and beta cells from 11 and 14 islet preparations, respectively. Among WGBS samples passing quality control (one alpha cell sample did not), 3 donors had pre-T2D/T2D (both alpha and beta cell), and 7 (alpha cell) and 11 (beta cell) were controls. RNA-seq was performed on sorted alpha and beta cells from 22 and 24 islet preparations, respectively. Among RNA-seq samples passing quality control (two beta cell samples did not), 7 donors had pre-T2D/T2D and 15 were controls (both alpha cells and beta cells).c, Density plot showing the global methylome of human alpha (mean 76.1%) and beta (mean 77.2%; n = 7 donors) cells. Peaks near 0% and 100% methylation.d,e, Density plots showing the degree of DNA methylation in alpha and beta cells in different genomic (d) and CpG island (e) regions. d, The region 1 to 5 kilobase pairs (kb) upstream of the transcription start site (TSS) had mean methylation levels of 66.7% (alpha cells) and 67.0% (beta cells). Promoters (1,000 bp upstream of TSS) had mean methylation levels of 25.5% (alpha cells) and 25.4% (beta cells). 5’ UTRs had mean methylation levels of 13.5% (alpha cells) and 13.1% (beta cells). Exons had mean methylation levels of 62.4% (alpha cells) and 62.2% (beta cells). Introns had mean methylation levels of 79.8% (alpha cells) and 79.9% (beta cells). 3’ UTRs had mean methylation levels of 80.6% (alpha cells) and 80.0% (beta cells). Intergenic regions had mean methylation levels of 78.1% (alpha cells) and 80.2% (beta cells). e, The mean degree of DNA methylation of Shelves (81.9% in alpha and 82.6% in beta cells), Shores (62.2% in alpha and 62.4% in beta cells), CpG islands (15.8% in alpha and in beta cells) and Open sea (82.4% in alpha and 83.7% in beta cells). A CpG island is defined as a ≥200-bp-long stretch of DNA with a CG content of ≥50% and an observed CpG/expected CpG in excess of 0.6. The Shores are the 2-kb flanking regions of CpG islands, the Shelves are the 2-kb regions outside island shores, and the Open Sea is everything else. f,g, Average DNA methylation levels, using the WGBS data, in different gene regions of genes divided according to expression level (no, low, medium and high expression) in alpha (f; n = 7 donors) and beta cells (g; n = 11 donors), respectively. The WGBS data were analysed by dmrseq 25 , and the Friedman rank-sum test was used to test differences in the average DNA methylation levels between the groups. Data are presented as the mean ± s.e.m.; for f, ***P1-5kb = 0.000105, Ppromoters = 0.000105,P5’ UTR = 0.000105, Pexons = 0.000105, Pintrons = 0.000172, P3’ UTR = 0.000105; forg, ***P1–5 kb = 0.000000835, Ppromoters = 0.000000322, P5’ UTR = 0.000000322,Pexons = 0.000000835, Pintrons = 0.000000322, P3’ UTR = 0.000000322, as analysed by two-sided Friedman rank-sum tests, not corrected for multiple testing. The vertical lines in c–e represent the mean DNA methylation. Schematic in a created in BioRender; Ofori, J. https://biorender.com/h6zumch (2026). ### Page 3 Nature Metabolism | Volume 8 | April 2026 | 957–980 959Resource https://doi.org/10.1038/s42255-026-01498-9different genomic regions associate with the expression level of genes in alpha and beta cells. Genes with <2 mean normalized counts were cat- egorized as not expressed, and the remaining 24,730 and 24,726 genes in alpha and beta cells, respectively, were divided into three groups of similar size, categorized into low (0–33% quantile), medium (33–66% quantile) and high (66–100% quantile) expression. We found associa- tions between the mean DNA methylation levels in different genomic regions and expression levels (Fig. 1f,g; Friedman rank-sum test). In promoters and 5’ UTR regions, lower methylation was associated with higher gene expression; while in intragenic regions, for example, exons, introns and 3’ UTR, lower methylation was associated with lower expres- sion (Fig. 1f,g). Additionally, we found small significant differences between alpha and beta cells for the mean degree of methylation for different genomic regions linked to the four categories of expressed genes (Extended Data Fig. 1c–h).DMRs in human alpha versus beta cellsTo further dissect the epigenetic differences between alpha and beta cells, we used dmrseq 25 and WGBS data from seven donors not diag- nosed with T2D and with the methylome available in both cell types (Supplementary Tables 1 and 2 and Extended Data Fig. 1i). On autosomal chromosomes, WGBS covered 25,907,683 CpGs. We found 22,544 differ- entially methylated regions (DMRs) annotated to 7,975 autosomal genes in alpha versus beta cells, when requiring three consecutive differen- tially methylated CpGs and false discovery rate (FDR) ≤ 5% (q < 0.05; Fig. 2a,b, Extended Data Fig. 1j and Supplementary Table 3a–c). These genes are enriched in Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways including insulin secretion, cAMP signalling, gluca- gon signalling, calcium signalling, metabolic pathways and T2D mellitus (Fig. 2c and Supplementary Table 3d). The DMRs covered 3–379 CpGs (average 10 CpGs) and 9–10,752 base pairs (bp; mean length 637 bp).0.3 0.2 0.1 0 0 25 50 75 100 DensityαDNA methylation (%) Global methylome 0.6 0.4 0.2 0 Density 1–5 kb DNA methylation (%) High expression Medium expression Low expression Not expressed 1–5 kb Promoters 5′ UTR Exons Introns 3′ UTR0 20 40 60 80 100 α-cells β-cellsf gIslets of langerhans Donors with or without T2D (n = 24) Pancreas Single-cell dispersion Stain for glucagon/insulin FACS Laser Detector DNA/RNA extraction Library preparation Whole-genome bisulfite sequencing RNA-seq Differential DNA methylation Differential expression Human islets EndoC and INS1 cells Diabetic GK rat Gene expression, epigenetic editing, insulin secretion, metabolism, etc. LUDC islet sortingn = 24a bRNA-seq (n = 22 α + 24 β) WGBS (n = 11 α + 14 β) Quality control RNA-seq data (n = 22 α + 22 β) WGBS data (n = 10 α + 14 β) Controls (n = 7 α + 11 β) pre-T2D/T2D (n = 3 α + 3 β)cα β β α = 76.1%β = 77.2%α = 66.7%β = 67.0%α = 25.5%β = 25.4%d Promotersα = 13.5%β = 13.1% 5′ UTRα = 62.4%β = 62.2% 0.6 0.4 0.2 0 Intronsα = 79.8%β = 79.9%α = 80.6%β = 80.0% 3′ UTR Intergenicα = 78.1%β = 80.2% 25 50 75 0 100 25 50 75 0 100 25 50 75 0e100 Shelves Shoresα = 81.9%β = 82.6%α = 62.2%β = 62.4% Density 1.0 0.5 0 CpG islands Open seaα = 15.8%β = 15.8%α = 82.4%β = 83.7% Density 1.0 0.5 0 25 50 75 0 100 25 50 75 0 100 DNA methylation (%) DNA methylation (%) Controls (n = 15 α + 15 β) pre-T2D/T2D (n = 7 α + 7 β) Exons 25 50 75 0 100 Density 0 20 40 60 80 100 DNA methylation (%) 1–5 kb Promoters 5′ UTR Exons Introns 3′ UTRα β α β*** *** *** *** *** *** *** *** *** *** *** *** ### Page 4 Nature Metabolism | Volume 8 | April 2026 | 957–980 960Resource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 1k shows the annotation of the DMRs in relation to different genomic regions and CG density (Supplementary Table 3b,c). Interestingly, DMRs annotated to key alpha and beta cell genes 21,24,26,27 , for example, PDX1, PLUT, INS, INS-IGF2, PCSK1, GCG, SLC2A2, MAFA,GATA6, SIX2 and GPR119, were among the most significant (Fig. 2d–i, Extended Data Fig. 1l–n and Supplementary Table 3b). Several genes have multiple DMRs, for example POU6F2 (n = 16), PLUT (n = 11), PCSK1(n = 10), INS/INS-IGF2 (n = 7) and PDX1 (n = 3; Fig. 2d,e, Extended Data Figs. 2 and 3 and Supplementary Table 3b). The longest DMRs (≥6,104 bp) were annotated to, for example, GCG, PLUT, PDX, HADH, RXRG, PAX3,MAFA, ZC3H3 and KCNQ5, while DMRs annotated to PLUT, PDX1 andMAFA/ZC3H3 included most CpGs (≥212; Supplementary Table 3b, Fig. 2d,g and Extended Data Fig. 4a). We then compared the list of 7,975 genes with alpha versus beta cell DMRs identified in the current study with a list of genes presented in Supplementary Tables 9 and 10 in a similar study by Manduchi et al.28 . This showed that 468 of 706 (66%) genes, including GCG, INS, MAFA and PDX1, presented in their study exhibit DMRs also identified in our study (hypergeometric enrichment score (HES) = 2.07, P = 2.63 × 10 −58 ). Epigenetics may regulate expression by controlling TF binding to the genome29 . Therefore, we investigated TF motifs and binding in iden- tified DMRs. Using HOMER, we found enrichment of TF motifs, includ- ing NEUROD1, FOXA2, PDX1, FOXA3, FOXP1 and MAFB, in these DMRs (Fig. 2j and Supplementary Table 3e). Additionally, we present DMRs that overlap with sequences bound by the islet-specific TFs, FOXA2, MAFB, NKX2.2, NKX6.1 and PDX1 in Fig. 2k, and Supplementary Table 3f (enriched overlap, P ≤ 0.001)24 . Moreover, 270 DMRs, annotated to for example, CREB5, GLP1R, HDAC9, OPRD1 and SYT13, were bound by all five TFs. Next, since genetics affects T2D, especially via islet dysfunction2 , and because we previously demonstrated interactions between genetics and epigenetics 30,31 , we studied the overlap between genes linked to T2D-asociated SNPs 2 and the 7,975 genes with DMRs pre- sented in Supplementary Table 3. Interestingly, 381 of the 849 (45%) T2D candidate genes also have DMRs in alpha versus beta cells (Fig. 2l and Supplementary Table 3g, HES = 1.74, P = 2.63 × 10 −34 ). Addition- ally, 25 T2D-associated SNPs are located within DMRs annotated to for example, INS-IGF2, CDKN2B, PCSK1, GLP1R, MEG3 and KCNQ1(Fig. 2m and Supplementary Table 3g). Notably, we found that 446 of 1,289 T2D SNPs 2 are in CpGs (Supplementary Table 3g). Together, this supports that T2D-risk alleles can affect DNA methylation, and potentially alter T2D risk, via effects on methylation30 . Additionally, 831 of the DMRs, for example, annotated to GCG, SLC2A2 and AKT1, overlap with 1,086 CpGs with differential islet methylation in T2D (Supplementary Table 3a)12 . Finally, we intersected our alpha versus beta cell DMRs with three-dimensional chromatin map data 32 . Here, 30 of 89 chromatin interaction points that differ between alpha and beta cells overlap with an alpha versus beta cell DMR, with 15 over- lapping with more than one DMR (Supplementary Table 3h, overlapz-score = 8.3, P ≤ 0.001). Overall, these data clearly show cell-specific differences in the islet epigenome, overlapping key TF binding sites24 , T2D-associated SNPs2 and T2D-associated islet DNA methylation12 . Epigenetic differences on sex chromosomes in alpha versus beta cells were analysed separately in males and females. In males, we analysed 1,298,421 CpGs on X and Y chromosomes. We identi- fied 453 DMRs, annotated to 178 genes, on the X chromosome, but none on the Y chromosome (Supplementary Table 4). Some of the most significant DMRs on the male X chromosome are annotated toARX, USP27X, PRKX, GPR119, POLA1, MPP1, DLG3, HDAC8 and KDM6A(Fig. 2n, Extended Data Fig. 4b,c and Supplementary Table 4b). The longest X-chromosome DMRs (≥2,685 bp) are annotated to NEXMIF,GPR119, HS6ST2 and DACH2, while DMRs annotated to BCOR, HCFC1and GPR119 have most CpGs (Supplementary Table 4b). Additionally, we present X-chromosome DMRs that overlap with sequences bound by FOXA2, MAFB, NKX2.2, NKX6.1 and/or PDX1 in Extended Data Fig. 4d and Supplementary Table 4d (enriched overlap, P ≤ 0.001). As there were only two non-diabetic females with WGBS data available in alpha and beta cells, we could not identify X-chromosome DMRs in females.DEGs in human alpha versus beta cellsNext, we studied gene expression in alpha and beta cells from 16 non-diabetic donors (Supplementary Table 1b). Using RNA-seq, we found 13,001 and 12,876 genes expressed in alpha and beta cells, respectively (≥20 mean normalized counts in ≥80% of samples, unique Ensembl gene ID; Supplementary Table 5a,b). Notably, 93% and 94% of the genes found to be expressed in alpha and beta cells in our data- set overlap with previously published work 27 , after reanalysing that data using the same conditions as ours (Supplementary Table 5a,b). While 12,384 genes were expressed in both cell types, 617 and 492 tran- scripts were uniquely expressed in either alpha or beta cells (Fig. 3a and Supplementary Table 5a,b). Based on gene ontology (http://www. webgestalt.org/), transcripts expressed in alpha cells, but not beta cells, showed enrichment in, for example, cAMP-mediated signal- ling, regulation of ion transport, regulation of secretion, exocytosis, regulation of apoptotic process and cell proliferation (Fig. 3b and Supplementary Table 5c). The 492 transcripts expressed only in beta cells, showed enrichment for 12 biological processes, for example, gland development (including NKX3.1), regulation of hormone levels (includ- ing MAFA, PDX1, GLP1R and SLC2A2), synaptic signalling (includingFig. 2 | Genes important for islet cell function and T2D exhibit widespread DNA methylation differences in sorted human alpha versus beta cells. a, Heat map displaying significant DMRs between human alpha and beta cells (n = 7 donors) sorted from the same islet preparations (FDR < 5%, q < 0.05). b, Volcano plot showing hypomethylated and hypermethylated DMRs identified by dmrseq 25 in human beta compared to alpha cells (n = 7 donors). q < 0.05 indicates the cut-off (dashed line) and the ‘methylation difference’ on the x axis represents the beta coefficient generated by dmrseq and included in Supplementary Table 3a. c, KEGG pathways with FDR below 5% (q < 0.05) based on a WebGestalt analysis of 7,975 genes with DMRs between sorted alpha and beta cells. d–i, Left, Top DMRs analysed and visualized by dmrseq 25 in alpha versus beta cells are annotated to PDX1 (d), INS/INS-IGF2 (e), PCSK1 (f), GCG (g), SLC2A2 (h) and MAFA(i; n = 7 donors). Right, Expression of respective gene in sorted alpha versus beta cells (n = 16 donors, ***q < 0.001 based on DESeq2 (ref. 82). Data are presented as the mean ± s.e.m. Women and men are represented by filled and open circles, respectively. j, The significant DMRs between alpha and beta cells (q < 0.05, based on dmrseq 25 ) are enriched for putative binding motifs of TFs such as NEUROD1, FOXA2, FOXA3, FOXP1 and MAFB (q < 0.05, as analysed by HOMER, http://homer. ucsd.edu/homer/motif/). k, Venn diagram and a table showing the significant DMRs between alpha and beta cells (q < 0.05) overlapping with binding sites for five islet-specific TFs: FOXA2, MAFB, NKX2.2, NKX6.1 and PDX1 (ref. 24). Overlaps are enriched with P ≤ 0.001 based on region-based permutation tests. For individual P values, see Supplementary Table 3f. l, 381 of 849 genes (45%) with T2D-associated SNPs 2 overlapped with genes having DMRs in alpha versus beta cells (HES = 1.74, P = 2.63 × 10−34 , based on one-tailed hypergeometric test). m, A significant DMR in alpha versus beta cells annotated to INS-1GF2directly overlaps the T2D-associated SNP (rs1124699, dashed line). n, The top X-chromosome DMR (only in males, five donors) annotated to ARX (q < 0.0082, based on dmrseq25 ) and with expression in sorted alpha versus beta cells (n = 10 male donors, ***q < 0.001, based on DESeq2 (ref. 82). Data are presented as the mean ± s.e.m. Women and men are represented by filled and open circles, respectively. Supplementary Table 6 presents P values for DEGs in d–i. In all DMR plots, locations of the methylation loci are shown with black lines at the bottom, and average, smoothed lines based on the dmrseq analysis are shown in pink (beta cells) or blue (alpha cells), including the standard deviation. The shaded area shows the DMR. Annotations to CpG island regions and genes are shown below each graph. Women and men are represented by filled and open circles, respectively. P and q values for a–i, m and n were corrected for multiple testing. CPM, counts per million. ### Page 5 Nature Metabolism | Volume 8 | April 2026 | 957–980 961Resource https://doi.org/10.1038/s42255-026-01498-90.6 0.8 0.2 0.4d f h j k l bID 3α-cells β-cells 75 100 25 0 50 Methylation (%)PDX1/PLUT Chr13: 27921596−27930664 (width = 9,069 bp), q = 0.0017 0 1 × 10 3 2 × 10 3 3 × 10 3 4 × 10 3 5 × 103α βPDX1 expression (CPM)β αIsland Shore Shelf Open seaPDX1Exons CpG island 75 100 25 0 50 Methylation (%)PCSK1 Chr5: 96425887−96431524 (width = 5,638 bp), q = 0.0017 0 0.5 × 104 1.0 × 104 1.5 × 10 4 2.0 × 10 4 2.5 × 104PCSK1 expression (CPM)PCSK1IslandShoreShelfOpen sea Exons CpG islandβ αID 6 ID 9 ID 11 ID 14 ID 19 ID 24 ID 3 ID 6 ID 9 ID 11 ID 14 ID 19 ID 24PCSK1 GLS GCG MAFA/ZC3H3 PDX1/PLUT SIX2 IGF2/INS−IGF2 GATA6Cutoff: q value 0.05 13,895 DMRs 8,649 DMRs 5 4 3 2 1 0 −2 0 Methylation difference 2 −log10 (P value) Lower methylation in β-cells Higher methylation in β-cellseName FOXP1 NEUROD1 FOXA2 PDX1 FOXA3 MAFB Motif P value 1.0 × 10 –121 1.0 × 10–101 1.0 × 10–93 1.0 × 10–90 1.0 × 10–70 1.0 × 10–15 0 0.5 1.0 1.5 2.0 Metabolic pathways Calcium signalling pathway FoxO signalling pathway AMPK signalling pathway Glucagon signalling pathway Insulin signalling pathway cAMP signalling pathway Type II diabetes mellitus Insulin resistance Enrichment ratio 75 100 25 0 50 Methylation (%)INS/INS-IGF2 Chr11: 2157235−2161277 (width = 4,043 bp), q = 0.0017INS expression (CPM) 0 2 × 105 4 × 105 6 × 105 8 × 105 10 × 105 IslandShore Shelf Open seaINS/INS-IGF2Exons CpG islanda c45% GWAS (n = 849) DMR (n = 381) e.g., BCL6, CDKAL1, CDKN2B, CRY2, FTO, GLIS3, GLP1R, GRB10, HDAC9, HNF1A, HNF4A, IGF1R, IGF2, IGF2BP2, IGF2BP3, INS, IRS1, IRS2, KCNQ1, KDM4B, NCOR2, NEUROG3, NOTCH2, ONECUT1, PTEN, SIX2, TCF7L2, THADA, VEGFA75 100 25 0 50 Methylation (%)SLC2A2 Chr3: 171023531−171026852 (width = 3,322 bp), q = 0.0017 0 100 200 300 400 500SLC2A2 expression (CPM) Exons CpG islandSLC2A2IslandShoreShelfOpen sea 75 100 25 0 50 Methylation (%)GCG Chr2: 162141841−162152592 (width = 10,752 bp), q = 0.0017 0 0.5 × 10 6 1.0 × 106 1.5 × 10 6 2.0 × 106 2.5 × 106GCG expression (CPM) Open seaGCGExons CpG island 75 100 25 0 50 Methylation (%)MAFA Chr8: 143423362−143427602 (width = 4,241 bp), q = 0.0017 0 2 × 103 4 × 103 6 × 103 8 × 103 10 × 103MAFA expression (CPM)MAFAIsland ShoreShelf Exons CpG island FOXA2 MAFB PDX1 NKX2.2 NKX6.1 1,424 491 450 1,604 136 270 75 142 96 1 18 18 15 25 51 104 150 14 31 133 283 1 16 96 220 139 36 150 406470 275 19 649 e.g., ACVR1C, CREB5, GLP1R, HDAC9, JAZF1, OPRD1, SYT1375 100 25 0 50 Methylation (%)ARX ChrX: 25014494−25015241 (width = 748 bp), q = 0.0082 0 2,000 4,000 6,000ARX expression (CPM) ShoreARXExons CpG islandg nINS-IGF2 Chr11: 2143347−2146620 (width = 3,274), q = 0.0017IGF2/INS-IGF2CpG island 75 100 25 50 Methylation (%) Shore Island Shelfmβ α β αiFOXA2 MAFB NKX2.2 NKX6.1 PDX1 TF DMRs overlapping TF binding sites TF binding sites overlapping DMRs 3,969 (17.6%) 1,752 (7.8%) 4,897 (21.8%) 1,914 (8.2%) 2,952 (13.1%) 4,054 (7.4%) 1,791 (8.3%) 5,072 (8.9%) 1,943 (12.5%) 3,011 (14.1%) 0 T2D SNP: rs1124699P = 2.63 × 10–34 Insulin secretion ***α β α βα β α β α β α β*** *** ***β α***β α***β α β α*** ### Page 6 Nature Metabolism | Volume 8 | April 2026 | 957–980 962Resource https://doi.org/10.1038/s42255-026-01498-9NPY and SYT1), cell–cell signalling (including GAD1 and GLP1R) and ion transport (including GLRA1; Fig. 3c and Supplementary Table 5d). In both alpha and beta cells, ~40% of the 20 most abundantly expressed genes are mitochondrially encoded (Fig. 3d,e). Additionally, 8,434 protein-coding genes were DEGs in alpha versus beta cells (q < 0.05; Fig. 3f and Supplementary Table 5e). Notably, ~40% (3,502) of these DEGs also have DMRs (Supplementary Table 5f), sup- porting epigenetic regulation of cell-specific expression (see for exam- ple, Fig. 2d–i, Extended Data Fig. 1l–n and Fig. 3g–o). Additionally, the DMRs correlated negatively with the DEGs (Supplementary Table 5g). We proceeded to focus on the biological relevance of the most significant DEGs in alpha versus beta cells (q < 10 × 10−8 ; Supplementary Table 6a). Of these 3,618 DEGs, ~50% had higher expression in each cell type, and they were enriched in KEGG pathways including insulin secretion, T2D mellitus and circadian entrainment, as well as glucagon, cAMP, AMPK, FoxO and calcium signalling (Fig. 3p and Supplementary Table 6b–d). DEGs showing higher expression in alpha cells (q < 10 × 10 −8 ) were enriched for ten pathways, for example, cAMP signalling, while DEGs showing higher expression in beta cells (q < 10 × 10−8 ) were enriched for pathways related to, for example, insulin secretion and maturity-onset diabetes of the young (Supplementary Table 6e,f). Among DEGs withq < 10 × 10−8 in alpha versus beta cells, 1,835 (51%) also have DMRs anno- tated to them (Supplementary Tables 3 and 6). Next, we intersected these DEGs with the 849 GWAS T2D candidate genes2 and 186 (22%) genes, including ADCY5, GLIS3, GLP1R, HNF4A andINS, were overlapping (Fig. 3q and Supplementary Table 6g, HES = 1.67,P = 1.07 × 1031 ).Epigenetic editing reveals a causal role for DNA methylation in regulation of INS, TH and GCG expressionThe cell-specific DNA methylation differences, which we found linked to large differences in expression of key alpha and beta cell genes, suggest that the methylome directly regulates expression and potentially hormone levels in individual islet cell types (Fig. 2d–i and Supplementary Table 3b). To test this, in human EndoC-βH1 beta cells, we used epigenetic editing with CRISPR–dCas9–DNMT3A or CRISPR– dCas9–TET1 (ref. 33) together with guide RNAs (gRNAs) targeting three alpha versus beta cell DMRs annotated to INS, TH and GCG(Figs. 2e,g and 4a–f). CRISPR–dCas9–DNMT3A-mediated editing, with theINS-DMR-gRNA targeting the INS DMR (Figs. 2e and 4a,b) and theTH-DMR-gRNA targeting a region annotated to TH (Fig. 4c,e), resulted in increased methylation of 22 and 13 CpG sites, respectively, as ana- lysed by pyrosequencing and EPIC v2.0 methylation arrays (Fig. 4g–j). Notably, after editing, 5 CpGs near the INS-DMR-gRNA target loca- tion showed absolute increased methylation of 20–43%, representing 2.46- to 30-fold increases, and these sites may have a stronger impact on gene expression than the other sites (Fig. 4h). DNMT3A-mediated editing using INS-DMR-gRNA and TH-DMR-gRNA caused reduced INSand TH expression, respectively (Fig. 4k,l). The targeted region in TH is located very near a potential INS enhancer (Fig. 4a,c) but editing using the TH-DMR-gRNA did not alter INS expression (Fig. 4k). Importantly, DNMT3A-mediated editing using the INS-DMR-gRNA led to reduced insulin content in human EndoC-βH1 beta cells (Fig. 4m). Next, we used CRISPR–dCas9–TET1 and a gRNA (GCG-DMR-gRNA) targeting a DMR located in the promoter of GCG, encoding glucagon (Figs. 2g and 4d). This led to decreased DNA methylation at two CpG sites analysed by pyrosequencing and EPIC v2.0 (Fig. 4n,o). Addi- tionally, the ten-eleven translocation methylcytosine dioxygenase 1 (TET1)-based editing using the GCG-DMR-gRNA resulted in increasedGCG expression and glucagon content in the EndoC-βH1 beta cells (Fig. 4p,q). Importantly, to our knowledge, this is the first study where epi- genetic editing reveals a causal role for DNA methylation on INS, THand GCG expression in beta cells, supporting future development of epigenetic-based therapies for regulation of, for example, insulin.Pre-T2D/T2D-associated DMRs in human alpha and beta cellsTo explore the epigenetic component of T2D, we compared the methy- lome in alpha and beta cells from donors with pre-T2D/T2D, based on having a haemoglobin A1c (HbA1c) ≥ 6% (≥42 mmol mol−1 ) and/ or a T2D diagnosis, with cells from controls, including non-diabetic donors with HbA1c < 6% (Supplementary Table 1a,b). In alpha cells, we found 3,207 autosomal DMRs associated with pre-T2D/T2D (hereafter called T2D-associated; permutation P < 0.05, Fig. 5a and Supplementary Table 7a–c). These DMRs were annotated to 1,871 genes, including GCG, INS-IGF2 and PDX1, enriched in KEGG pathways, including circadian entrainment (Fig. 5b and Supplementary Table 7d). The DMRs covered 3–257 CpGs and 10–18,509 bp. For protein-coding genes, T2D-associated alpha cell DMRs with lowest P values were anno- tated to ZNF566, UNC93A, CROCCP2, TBCA, USP18 or SNRPN, and the longest DMRs (>5,025 bp) were annotated to TUBBP5 and FLG, with the one annotated to TUBBP5 also covering most CpGs (Fig. 5a and Supplementary Table 7b). Figure 5c–h presents some T2D-associated alpha cell DMRs annotated to genes previously linked to T2D and/or islet function, for example, SNRPN34 , SLC52A2/TMEM249 (ref. 35), CHURC1(ref. 36) and TULP4 (ref. 37) or encoding epigenetic enzymes, for exam- ple KDM1B and KDM3B. Using HOMER, we found enrichment of TF motifs, including HNF1B and OCT4, in the T2D-associated alpha cell DMRs (Fig. 5i and Supplementary Table 7e), and we identified DMRs that overlap with sequences bound by the islet-specific TFs FOXA2, MAFB, NKX2.2, NKX6.1 and PDX1 (Fig. 5j and Supplementary Table 7f; enriched overlap, P ≤ 0.004). Additionally, 103 of 849 (12%) T2D candidate genes discovered by GWAS2 also have T2D-associated alpha cell DMRs (Fig. 5k and Supplementary Table 7g; HES = 7.34, P = 1.37 × 10 −238 ). This repre- sents 5.5% of the 1,871 genes annotated to T2D-associated alpha cell DMRs. The overlapping genes include GIP, GLIS3, GRB10, HDAC9, IGF2,IGF2BP3, KCNQ1, KDM4B, MEG3, TH and TCF7L2. The T2D-associated SNP rs883541 (ref. 2) lies within a DMR annotated to WIPI1 (Fig. 5l). Addi- tionally, 28 T2D-associated alpha cell DMRs overlap with 32 CpGs previ - ously found to be differentially methylated in islets from donors with T2D versus controls 12 (Supplementary Table 7a). T2D-associated dif- ferences in methylation of individual CpGs of GCG, INS, PDX1/PLUT andMAFB in alpha cells are shown in Fig. 5m and Extended Data Fig. 4e–g.Fig. 3 | The transcriptome of sorted human alpha and beta cells. a, Venn diagram showing the number of overlapping (12,384) and uniquely expressed protein-coding genes in human alpha and beta cells, with 617 and 492 genes, respectively, requiring ≥20 mean normalized counts in ≥80% of the samples. b,c, Gene ontology of the protein-coding genes uniquely expressed in human alpha (b) or beta (c) cells, showing significantly enriched pathways (q < 0.05). d,e, The 20 most abundantly expressed genes in alpha (d) and beta (e) cells of 16 donors. Data are presented as the mean ± s.e.m. Women and men are represented by filled and open circles, respectively. f, Volcano plot showing DEGs at q < 10 × 10−8 andq < 0.05, based on DESeq2 (ref. 82), in human alpha versus beta cells sorted from the same islet preparations (n = 16 donors). g–o, Alpha versus beta cell DEGs based on DESeq2 (ref. 82; ***q < 0.001, n = 16 donors) that also have significant DMRs (q < 0.05, based on dmrseq 25 ): TM4SF4 (g), F10 (h), IRX2 (i), FEV (j), LDHA(k), ADCY2 (l), NR3C1 (m), GLIS3 (n) and SYT13 (o). Data are presented as the mean ± s.e.m. Women and men are represented by filled and open circles, respectively. p, Selected KEGG pathways with significant enrichment (q < 0.05 based on WebGestalt analysis with correction for multiple testing) of DEGs in alpha versus beta cells with q < 10 × 10−8 . q, 186 of 849 genes (22%) with T2D- associated SNPs identified by GWAS 2 overlapped with the top DEGs (q < 10 × 10−8 , based on DESeq2) 82 in human alpha versus beta cells (HES = 1.67, P = 1.07 × 10−31 , based on one-tailed hypergeometric test). Exact P values for the expression data and DEGs are found in Supplementary Table 6. P/q values for b, c and f–p were corrected for multiple testing. ### Page 7 Nature Metabolism | Volume 8 | April 2026 | 957–980 963Resource https://doi.org/10.1038/s42255-026-01498-9In the beta cells, we discovered 5,106 T2D-associated autoso- mal DMRs annotated to 3,020 genes, for example, INS and INS-IGF2(permutation P < 0.05; Fig. 6a and Supplementary Table 8a–c). The 3,020 genes were enriched in KEGG pathways including cAMP, Wnt, calcium, PI3K-Akt, FoxO and AMPK signalling (Fig. 6b and Supplementary Table 8d). T2D was associated with lower meth- ylation of 75.5% (3,857) of these DMRs. The DMRs covered 3–315 CpGs (average 10 CpGs) and 11–7,594 bp (mean length 662 bp). The longest DMRs (>3,587 bp) were annotated to, for example TUBBP5,LINC01597, MYO16, SHISAL2B, FAM242A, LOC101927078, OR5D16and MIR4321, while a DMR annotated to MIR4321, AMH and JSRP1has the most CpGs (315 loci; Fig. 6a and Supplementary Table 8b). Figure 6c–j presents some T2D-associated beta cell DMRs annotated to genes previously linked to T2D and/or islet function, includingEPS8L1 (ref. 38), VEGFA39 , NKX6.3 (ref. 40), CTNND2 (ref. 41), RGS6(ref. 42), FEM1B43 , SEC16A44 and SORBS1 (ref. 45). Using HOMER, we617 492 12,384p edα-cellsqβ-cells (q < 0.05) Ion transport Cell–cell signalling Synaptic signalling Regulation of hormone levels Gland developmentα-cells (q < 0.05) Regulation of apoptotic process Cell proliferation Exocytosis Regulation of secretion Regulation of ion transport cAMP-mediated signalling 0 0.5 2.52.01.0 1.5 3.0 Enrichment ratioCalcium signalling pathway Insulin signalling pathway FoxO signalling pathway AMPK signalling pathway cAMP signalling pathway Glucagon signalling pathway Insulin resistance Circadian entrainment Type II diabetes mellitus Insulin secretion0 0.5 2.52.01.0 1.5 Enrichment ratio GWAS (n = 849) DEGs (n = 186)e.g., ACSL1, ADCY5, AUTS2, CASR, CDKN2B, DLK1, GINS2, GLIS3, GLP1R, GRB10, HDAC9, HNF1A, HNF4A, KCNJ8, KDM4B, KLF13, IGF1R, IGF2BP2, INS, INSR, NCOR2, NKX6-1, NRXN3, PCSK1, PDE3A, SCTR, SOCS2, TLE1, TTN, VEGFA22%α-cellsn = 13,001β-cellsn = 12,876h2,000 4,000 6,000 8,000α β0 1,000 2,000 3,000 4,000 5,000α β500 1,000 1,500i l0.5 × 10 5 1.0 × 10 5 1.5 × 10 5 2.0 × 10 5 2.5 × 10 5α β0TM4SF4 expression (CPM) 500 1,000 1,500 2,000 2,500α β0F10 expression (CPM)IRX2 expression (CPM) 0LDHA expression (CPM)ADCY2 expression (CPM) 0 0 5,000 10,000 15,000 20,000α β α βSYT13 expression (CPM)a500 1,000 1,500 2,000 2,500jFEV expression (CPM) 0α β0 2,000 4,000 6,000βm n0 1,000 2,000 3,000 4,000α βGLIS3 expression (CPM)αNR3C1 expression (CPM)q < 0.05b c f FXYD5 TM4SF4F10 FAM83B ADGRG2 IRX2 CYP2U1 HS3ST3B1 LOXL4TTR TMEM236CRYBA2 VSTM2LMRC1 FEV KIAA1549L ITPR1 PPP2R2B KCNH6 TMEM45B MAFA NKX6-1 ARX GATA6 PDX1 INS INS-IGF2 ISL1 NEUROD1 GCGHigher eenes300 200 100ONECUT2 ADCY9 NKX2-2 PAX6GNAL FBLL1 q-value < 0.05 −5 0 5 log2 fold change −log10(P value)g k o12 9 6 3 04,335 genes 4,111 genes Higher expression in β-cells, protein-coding genes: 1,566 Higher expression in α-cells, protein-coding genes: 1,590RFX6 MAFB q-value < 10–8GCG TTR MT-CO1 CHGB VGF FTL MT-ND4 GNAS EEF1A1 SLC7A2 CPE MT-CO3 CLU MT-ND5 PDK4 ALDH1A1 MT-ND2 CHGA MT-CYB MT-ATP6Relative expression (CPM) 2.5 × 106 2.0 × 106 10 × 105 8 × 105 6 × 105 4 × 105 2 × 105 0 Relative expression (CPM) 0 10 × 10 5 8 × 105 6 × 105 4 × 105 2 × 105MT-CO1 MT-ND4 GNAS MT-CO3 EEF1A1 ABCC8 MT-CYB MT-ND2 MT-ND1 CPE MT-ATP6 EEF2 PTPRN MT-CO2 MT-ND5 SCD IAPP NPTX2 INS FTL0 0.5 2.52.01.0 1.5 3.0 Enrichment ratio 3.5β-cells *** *** *** *** *** *** *** *** ***P = 1.07 × 10–31 ### Page 8 Nature Metabolism | Volume 8 | April 2026 | 957–980 964Resource https://doi.org/10.1038/s42255-026-01498-9Transduction dCas9 TET1gRNA Puro Transduced EndoC-βH1 cells Selection 10 days DNA methylation Gene expression Hormone contentaDNMT3A/ TET1INS CpG Chr11: 2161277 DNA methylation (%) 15 12 9 6 3 0INS CpG Chr11: 2161227 DNA methylation (%) 8 6 4 2 0 DNMT3AINS-DMR-gRNA + ++ + – +– +bTH Chr11: 2166536−2166805 (width = 270), q = 0.0017 75 100 25TH0 50 Methylation (%) Exons CpG islandβ αdCas9 DNMT3AgRNA Puroc d e30 20 10 DNMT3AINS-DMR-gRNA + + – + Insulin content (mU/mg protein) 0 EndoC-βH1 cells 0INS expression 50 40 30 20 10 DNMT3AINS-DMR-gRNA + + – + + –TH-DMR-gRNA– – +TH expression 0.008 0.006 0.004 0.002 0 DNMT3ATH-DMR-gRNA + + – + DNMT3AINS-DMR-gRNA + DNMT3A 0 20 40 60 80 100INSDNA methylation (%) 0 10 20 30 55 60 65 70GCG CpG Chr2: 162152562 DNA methylation (%) TET1GCG-DMR-gRNA + + – + 30 35 20 25 10 15 0 5 Glucagon (pg/mg protein) TET1GCG-DMR-gRNA + + – + 0 0.001 0.002 0.003 0.004GCG expression TET1GCG-DMR-gRNA + + – + 0 10 20 30 40 50 60 DNMT3ATH-DMR-gRNA + DNMT3ATHDNA methylation (%) 0 20 40 97 98 99 100GCG cg05108984 DNA methylation (%) TET1GCG-DMR-gRNA + + – + IslandTH CpG Chr11: 2165314 DNA methylation (%) 100TH CpG Chr11: 2165318 DNA methylation (%) 95 90 85 40 20 0 100 95 90 85 40 20 0 DNMT3AINS-DMR-gRNA + + + + + – + –f g h i j k l m n o p qgRNA dCas9 cg13670288 cg26337010 cg03128167 cg12219390 cg16641322 cg09606495 cg01181416 cg00613255 cg25336198 cg11208229 cg22353154 cg14128708cg13670288cg26337010cg03128167cg24522478cg15923395cg16641322cg26452040cg19448770cg10781463cg04484994cg09606495cg13993218cg10991175cg01181416cg00613255cg25336198cg03366382cg24338752cg11208229 Chr11INS INS-DMR-gRNA PDX1 NKX2.2 MAFB PROMOTER Islet regulome TFs 2156000 2158000 2160000 2162000 2164000 2166000INSα vs β INS DMR 2161215 2161225 2161235 2161245 2161255 2161265 2161275INS-DMR-gRNA 2156000 2156500 2157000 2157500 2158000 2158500 2159000 2159500 2160000 2160500 2161000 2161500 11p14.1 11.11 q11 11q22.1Chr11INScg13670288 cg26337010 cg03128167 cg24522478 cg15923395 cg16641322 cg26452040 cg19448770 cg10781463 cg04484994 cg09606495 cg13993218 cg10991175 11.11 q11 11p15.4 11.11 q11 11q22.1 23.2Chr11 2154000 2156000 2158000 2160000 2162000 2164000 2166000 2168000 2170000 2172000 2174000 2176000 2178000 2180000INS TH TH-DMR-gRNA cg26337010 cg03128167 cg12219390 cg16641322 cg00613255 cg22353154 CpG Chr11: 2165314 CpG Chr11: 2165318 cg01181416 cg00613255 cg25336198 cg03366382 cg24338752 cg11208229 CpG Chr11: 2161227 CpG Chr11: 2161277 Chr2 162137000 162141000 162145000 162149000 162153000 162157000 162161000GCG GCG -DMR-gRNA cg05108984 CpG Chr2: 162152562 ENHANCER ENHANCERTHcg14128708 cg13670288 cg09606495 cg01181416 cg25336198 cg11208229 *** ******* * * * * ** * * ****** * ** * * * ** ** * * * * ** * * * * ** * * * * * * * * * * * * * * * * * * * * * * * * ** * * * * * * *T G A G G G C T G C T G G G G A G C C C A T C T C C C C T A C C T C T C A A C C C C T G C C G CC C C C C G C T G G C T T T A T A G T C T C A ### Page 9 Nature Metabolism | Volume 8 | April 2026 | 957–980 965Resource https://doi.org/10.1038/s42255-026-01498-9found enrichment of TF motifs including FOXA3, FOXA2, HNF1B, OCT4, FOXA1, PAX7, FOXP1, BACH2 and PDX1 in T2D-associated beta cell DMRs (Fig. 6k and Supplementary Table 8e). Addition- ally, we present T2D-associated beta cell DMRs that overlap with sequences bound by the islet-specific TFs FOXA2, MAFB, NKX2.2, NKX6.1 and PDX1 (Fig. 6l and Supplementary Table 8f; enriched overlap, P ≤ 0.001). Further, 153 of 849 (18%) T2D candidate genes identified by GWAS 2 also have T2D-associated beta cell DMRs, including CDKN2A, GLIS3, INS, KCNQ1, NKX6-3 and LPL (Fig. 6m and Supplementary Table 8f; HES = 2.69, P = 3.13 × 10 −85 ). This repre- sents 5.1% of the 3,020 genes annotated to T2D-associated beta cell DMRs. Three T2D SNPs 2 (rs362307/HTT, rs150015181/NKX6-3 and rs34245505/BMF) are located within T2D-associated beta cell DMRs (Fig. 6n and Supplementary Table 8g). Notably, 65 T2D-associated beta cell DMRs overlap 91 CpGs exhibiting differential islet meth- ylation in T2D 12 (Supplementary Table 8a). These include DMRs annotated to FER, ASCL5, HDAC4, CIT, PLX2 and FGF8. Additionally, Fig. 6o–q presents T2D-associated methylation differences of indi- vidual CpGs of GCG, INS and PDX1/PLUT in beta cells.Pre-T2D/T2D-associated DEGs in human alpha and beta cellsWe next investigated the impact of pre-T2D/T2D on gene expression in alpha and beta cells. By comparing RNA-seq data of alpha cells from 7 pre-T2D/T2D donors with 15 controls (Supplementary Table 1a,b), we identified 345 protein-coding DEGs with P < 0.05 (Fig. 7a and Supplementary Table 9a). When integrating these data with our published RNA-seq data from whole islets of a large T2D case–con- trol cohort 26 , and a sorted alpha cell T2D case–control cohort 46 , we found 11 and 3 overlapping DEGs, respectively, including GNAL(Extended Data Fig. 5a and Supplementary Table 9b,c). Additionally, 26 T2D-associated DEGs also have DMRs in alpha cells, for example,BACH1 (Supplementary Table 9a). To identify regulators of insulin secretion that may contribute to T2D, we compared RNA-seq data of beta cells from 7 pre-T2D/T2D donors with 15 controls (Supplementary Table 1a,b). This identi- fied 573 protein-coding DEGs with P < 0.05, including ADCY9, FBLL1,ZNF431, ZSWIM9, GNAL, ONECUT2, SNX1 and TRAPPC13 with FDR < 10% (Fig. 7b,c and Supplementary Table 10a). Notably, 43 of the DEGs found in beta cells overlap with T2D-associated DEGs found in wholeFig. 4 | Epigenetic editing shows a causative role for DNA methylation of INS,TH and GCG on gene expression in beta cells. a, The location of a human alpha versus beta cell INS DMR (purple; see also Fig. 2e), gRNA binding site (pink), TF binding in islets (https://pasqualilab.upf.edu/app/isletregulome/; blue), promoter (green) and enhancers (red). The zoomed-in part (bottom) shows the sequence targeted by the gRNA (black) and the two CpG sites covered by the pyrosequencing assay (orange). The figure was made with the help of the UCSC Genome Browser. INS runs from right to left in this image. b–d, The gRNA binding sites (pink), enhancer (c; red) and CpG sites where DNA methylation changed after epigenetic editing (P < 0.05) within the targeted loci for INS (b),TH (c) and GCG (d) analysed by pyrosequencing (displayed in blue), or EPIC v2.0 array (displayed in black). e, The TH DMR with differential DNA methylation in human alpha versus beta cells, in the region targeted by epigenetic editing. qvalue was corrected for multiple testing. Locations of the methylation loci are shown with black lines at the bottom, and average, smoothed lines based on the dmrseq 25 analysis are shown in pink (beta cells) or blue (alpha cells), including the standard deviation. The shaded area shows the DMR. Annotations to CpG island regions and genes are shown below the graph. f, Design of the epigenetic editing experiments performed to investigate the causal relationship between altered DNA methylation and changes in gene expression/cell function.g, Pyrosequencing shows that epigenetic editing with CRISPR–dCas9– DNMT3A and the gRNA targeting the INS DMR (INS-DMR-gRNA) increased DNA methylation of two analysed sites (CpG Chr11: 2161227 and CpG Chr11: 2161277, blue; b), compared with the negative control including DNMT3A but not theINS-DMR-gRNA in EndoC-βH1 beta cells. n = 4 biological replicates, **P < 0.01, based on one-tailed paired t-tests (left, P = 0.008; right, P = 0.006). h, The EPIC v2.0 array identified edited CpG sites between 10 kb upstream and downstream of INS, discovering absolute methylation changes of up to 40% between beta cells that had received both CRISPR–dCas9–DNMT3A and the INS-DMR-gRNA and the negative control (only DNMT3A). n = 4 biological replicates, *P < 0.05, **P < 0.01 and ***P < 0.001 based on one-tailed paired t-tests. i, Pyrosequencing data of two CpG sites with altered methylation after epigenetic editing with CRISPR–dCas9– DNMT3A together with TH-DMR-gRNA versus control (only DNMT3A). n = 4 biological replicates, *P ≤ 0.05 based on one-tailed paired t-tests (left, P = 0.05; right, P = 0.04). j, presents CpG sites identified by EPIC v2.0 array, where DNA methylation was altered after epigenetic editing with CRISPR–dCas9–DNMT3A and the TH-DMR-gRNA versus control (only DNMT3A). n = 4 biological replicates, *P < 0.05, **P < 0.01 and ***P < 0.001 based on one-tailed paired t-tests. k–m, CRISPR–dCas9–DNMT3A-based editing of INS or TH reduced INS (P = 0.001) andTH (P = 0.023) expression, respectively (quantitative PCR (qPCR), n = 4 biological replicates, k and l), and insulin content (P = 0.021, only after editing with INS- DMR-gRNA, n = 3 biological replicates, m) in EndoC-βH1 beta cells. *P < 0.05, **P < 0.01, based on one-tailed paired t-tests. n,o, Pyrosequencing and EPIC v2.0 array data show that epigenetic editing with CRISPR–dCas9–TET1 and the gRNA,GCG-DMR-gRNA, targeting the GCG DMR in Fig. 2g decreased DNA methylation of two CpG sites (CpG Chr2: 162152662 for pyrosequencing and cg05108984 for EPIC v2.0) versus the control (only TET1) in EndoC-βH1 beta cells. n = 6 biological replicates for pyrosequencing, *P = 0.05 based on one-tailed paired t-test. n = 4 biological replicates for EPIC v2.0 array, **P = 0.00741, based on one-tailed pairedt-test. p,q, CRISPR–dCas9–TET1-based editing with GCG-DMR-gRNA caused increased GCG expression (*P = 0.047, qPCR, n = 6 biological replicates, p), and glucagon content (**P = 0.008, ELISA, n = 5 biological replicates, q) versus the control (only TET1) in EndoC-βH1 beta cells, based on one-tailed paired t-tests. The colours of the dots in g–m and n–q represent samples from the same experiments. Data are presented as the mean ± s.e.m. Bars shown in light grey include CRISPR–dCas9–DNMT3A or CRISPR–dCas9–TET1 plus gRNA, while white bars are controls only including CRISPR–dCas9–DNMT3A or CRISPR–dCas9– TET1, and no gRNA.Fig. 5 | Key islet and T2D candidate genes exhibit pre-T2D/T2D-associated DMRs in human alpha cells. a, Volcano plot showing hypomethylated and hypermethylated DMRs in pre-T2D/T2D cases (n = 3) versus non-diabetic controls (n = 7) in alpha cells (P < 0.05 based on dmrseq 25 and 10 permutations). The ‘methylation difference’ on the x axis represents the beta coefficient values generated by dmrseq and included in Supplementary Table 7a. b, Selected enriched KEGG pathways among the 1,871 genes with pre-T2D/T2D-associated DMRs in alpha cells. For exact q values, corrected for multiple testing, see Supplementary Table 7d. c–h, T2D-associated DMRs in alpha cells of genes previously linked to T2D and/or islet function, or in genes encoding epigenetic enzymes: SNRPN (c), SLC52A2/(TMEM249) d), CHURC1 (e), TULP4 (f), KDM1B (g) and KDM3B (h). The DMRs were analysed and visualized by dmrseq 25 including ten permutations. i, The pre-T2D/T2D-associated DMRs in alpha cells are enriched for putative binding motifs of TFs such as HNF1B and OCT4, as analysed by HOMER (http://homer.ucsd.edu/homer/motif/; q < 0.05). j, Venn diagram and table showing pre-T2D/T2D-associated DMRs in alpha cells overlapping with binding sites for five islet-specific TFs: FOXA2, MAFB, NKX2.2, NKX6.1 and PDX1 (ref. 24). Overlaps are enriched with P ≤ 0.004 based on region-based permutation tests. For individual P values, see Supplementary Table 7f. k, 103 of 849 (12%) T2D candidate genes identified by GWAS 2 overlap with genes that had T2D-associated DMRs in alpha cells annotated to them (HES = 7.34,P = 1.37 × 10−238 based on a one-tailed hypergeometric test). l, A T2D-associated SNP2 (rs883541, dashed line) is located in a pre-T2D/T2D-associated alpha cell DMR, annotated to WIPI1 (based on dmrseq 25 ). m, pre-T2D/T2D-associated differences in the absolute DNA methylation level (%) of individual CpG sites annotated to GCG in alpha cells from seven non-diabetic donors and three donors with pre-T2D/T2D, *P < 0.05, **P < 0.01 based on two-tailed Wilcoxon rank-sum test. Data are presented as the mean ± s.e.m. Women and men are represented by filled and open circles, respectively. In c–h and l, locations of the methylation loci are shown with black lines at the bottom, and average smoothed lines based on the dmrseq analysis are shown in pink (pre-T2D/T2D) or blue (control), including the standard deviation. The shaded area shows the DMR. Annotations to CpG island regions and genes are shown below each graph. ### Page 10 Nature Metabolism | Volume 8 | April 2026 | 957–980 966Resource https://doi.org/10.1038/s42255-026-01498-9a bZNF566/ZNF566−AS1 UNC93A ZFPM1−AS1/ZFPM1 SNRPN TBCA PCSK5GLT1D1 SBNO2 USP18 CCDC102A PRPH2 EBF2 THEG BAIAP2 PITPNC1 SLC52A2/TMEM249 KDM1B CHURC1−FNTB/FNTB/MAX EPS8L1 ADCY9 ONECUT2Cut-off: P value 0.05 6 4 2 0 −2 0 Methylation difference 2 −log10(P value) Lower methylation in T2D1,695 DMRsHigher methylation in T2D 1,512 DMRsc d e f g h i75 100 25 0 50 Open sea CpG island75 100 25 0 50SNRPN Chr15: 24919379–24920345 (width = 967 bp), P = 0.00017 Island Shore Exons CpG islandTMEM24975 100 25 0 50 Exons CpG islandCHURC1/FNTB Chr14: 65031963–65032639 (width = 677 bp), P = 0.00044 Open seaCHURC1-FNTB/FNTB0 60 80 100162138895 162141148 162144640 162150909 162157034T2D ControlCpG locationName HNF1B OCT4 Motif q-value 0.0007 0.040TULP4 Chr6: 158509642–158509905 (width = 264 bp), P = 0.0020 Open sea 75 100 25 0 50 CpG island75 100 25 0 50 CpG island Open seaKDM3B Chr5: 138411108–138411224 (width = 117), P = 0.003975 100 25 0 50 Exons CpG islandKDM1B Chr6: 18207237–18208068 (width = 832 bp), P = 0.00038KDM1B HH1PL1FOXA2 MAFB NKX2.2 NKX6.1 PDX1 TF DMRs overlapping TF binding sites TF binding sites overlapping DMRs 133 (4.2%) 64 (2.0%) 191 (6.0%) 43 (1.3%) 63 (2.0%) 140 (0.26%) 65 (0.30%) 201 (0.35%) 43 (0.28%) 63 (0.29%)FOXA2 MAFB PDX1 NKX2.2 NKX6.1 65 26 6 80 3 3 0 3 0 3 0 0 0 2 0 0 0 2 7 4 10 9 1 3 1513 7 3 30 5 1SHBmGCG IGF2/INS-IGF2PDX1Methylation (%)Methylation (%)Methylation (%) Methylation (%)WIPI1 Chr17: 68452759−68453363 (width = 605), P = 0.009875 100 25 0 50 Exons CpG island Open sea ShelfWIPI1Methylation (%)j k lGWAS (n = 849) DMRs (n = 103) 12%e.g., CAMK1D, GCKR, GIP, GLIS3, GRB10, HDAC9, IGF2, IGF2BP3, JAZF1, KCNQ1, KDM4B, MEG3, NFATC1, SIX2, TCF7L2, VEGFACircadian entrainment Insulin secretion Glutamatergic synapse 0 0.5 2.52.01.0 1.5 3.53.0 Enrichment ratio T2D α-cells (q < 0.05) Methylation (%)SLC52A2 Chr8: 144353828–144354442 (width = 615 bp), P = 0.00023 Open sea Methylation (%) T2D SNP: rs883541 40 20GCGDNA methylation (%)P = 1.37 × 10–238 T2D Control T2D Control T2D Control T2D Control T2D Control T2D Control T2D Control ** ** * * ### Page 11 Nature Metabolism | Volume 8 | April 2026 | 957–980 967Resource https://doi.org/10.1038/s42255-026-01498-9islets, 26 including GABRA2, SOCS1, SYT1, PAX5, CABLES1 and TBC1D4; and 90% of these went in the same direction (Extended Data Fig. 5a–c and Supplementary Table 10b). Moreover, 20 and 19 DEGs found in our beta cells overlap with T2D-associated beta cell DEGs found in studies by Walker et al. and Martínez-López et al., respectively (Supplementary Table 10c,d)46,47 . Additionally, 88 genes exhibited both T2D-associated DEGs and DMRs in beta cells (Supplementary Table 10e). For the top DEGs (Fig. 7c), we identified individual CpGs annotateda cAMPK signalling pathway FoxO signalling pathway GABAergic synapse PI3K–Akt signalling pathway Calcium signalling pathway Wnt signalling pathway cAMP signalling pathway Glutamatergic synapse 0 0.5 2.52.01.0 1.5 3.0 T2D β-cells (q < 0.05)e f g h i l bSTEAP1B MIR4321/AMH/JSRP1 GPR146/C7orf50 EPS8L1FER LHPP CCDC102AVEGFAZFPM1−AS1/ZFPM1 GRIN3B/WDR18 SEMA5B GRID2 HMG20B/MFSD12 INSC ERRFI1 WDR97/SHARPIN CAPN15 LAMB1Lower methylation in T2D 3,857 DMRs Higher methylation in T2D 1,249 DMRs 6 4 2 0 −2 −1 0 1 Methylation difference 2 −log10(P value) Cut-off: P value 0.05FIGNL2dEPS8L1 Chr19: 55086536–55087518 (width = 983 bp), P = 5.6 × 10 −5 Shore CpG island Exons IslandEPS8L1Open seaVEGFA VEGFA Chr6: 43784557–43785555 (width = 999 bp), P = 0.00016NKX6-3 Chr8: 41647150–41648285 (width = 1,136), P = 0.00037 IslandNKX6-3ShoreCTNND2 Chr5: 11115400–11116334 (width = 935 bp), P = 0.0015RGS6 Chr14: 72433580–72433965 (width = 386 bp), P = 0.0016 ShelfFEM1B Chr15: 68281287–68281962 (width = 676 bp), P = 0.0025 Open seaSORBS1 Chr10: 95508224–95508714 (width = 491 bp), P = 0.0034 Enrichment ratio FOXA3 FOXA2 HNF1B OCT4 FOXA1 PAX7 FOXP1 BACH2 PDX1 Name Motif P value 1.0 × 10 –11 1.0 × 10–7 1.0 × 10–7 1.0 × 10–6 1.0 × 10–4 1.0 × 10–4 1.0 × 10–4 1.0 × 10 –2 1.0 × 10 –2SEC16AGWAS (n = 849) DMRs (n = 153) e.g., CDKN2A, GLIS3, IGF1R, IGF2BP2, INS, KCNQ1, GPC6, GRM8, NKX6-3, LPL, NCOR2, PDE3A, SIX2, SKI, WNT7B, SYT14, THADA, VEGFA18%FOXA2 MAFB PDX1 NKX2.2 NKX6.1 158 57 20 239 15 7 6 7 4 5 1 0 0 3 8 1 0 4 5 4 31 20 4 2 2318 12 1 78 14 6 e.g., CDC42BPA, FOXK1, GALNT8, KC6, KLHDC4, NEDD4Lm n HTT Chr4: 3239818−3240197 (width = 380), P = 0.045Open seaHTTMethylation (%) 75 100 25 0 50 CpG island Exons T2D SNP: rs362307 0 60 80 100 162145878 162152122 162152263 CpG location 2165739 CpG location T2D Controlo p qOpen sea 75 100 25 0 50 CpG island 75 100 25 0 50 CpG island Exons 75 100 25 0 50SEC16A Chr9: 136458946–136459334 (width = 389 bp), P = 0.0030 Methylation (%)Methylation (%)Methylation (%) Shelf Open sea CpG island Exons 75 100 25 0 50 Methylation (%) CpG island 75 100 25 0 50 Methylation (%) Open sea CpG island 75 100 25 0 50 Methylation (%)jCpG island Exons 75 100 25 0 50 Methylation (%) CpG island 75 100 25 0 50 Methylation (%)kFOXA2 MAFB NKX2.2 NKX6.1 PDX1 TF DMRs overlapping TF binding sites TF binding sites overlapping DMRs 336 (6.6%) 144 (2.8%) 472 (9.3%) 86 (1.7%) 126 (2.5%) 344 (0.63%) 146 (0.67%) 482 (0.84%) 87 (0.56%) 127 (0.59%) 40 20 0 60 80 100 40 20 0 60 80 100GCGDNA methylation (%)INSDNA methylation (%)PDX1/PLUTDNA methylation (%) 40 20 27816277 27819753 27822808 27827155 27853610 27860112 27860991 27893473 27895562 27918236 27920524 * * * * * * * * * * * * * ** *P = 3.13 × 10–85 T2D Control T2D Control T2D Control T2D Control T2D Control T2D Control T2D Control T2D Control T2D Control ### Page 12 Nature Metabolism | Volume 8 | April 2026 | 957–980 968Resource https://doi.org/10.1038/s42255-026-01498-9to ADCY9, ZNF431, ZSWIM9, GNAL, ONECUT2, SNX1 and TRAPPC13with altered methylation in beta cells from donors with pre-T2D/T2D versus controls (Extended Data Fig. 5d–j). In line with this, CpGs anno- tated to ADCY9 (n = 5), ZNF431 (n = 1), ONECUT2 (n = 2) and SNX1 (n = 1) show differential islet methylation in T2D 12 , supporting epigenetic dysregulation of these genes (Supplementary Table 10a,f). Nota- bly, three T2D-associated islet DEGs 26 were differentially expressed in both alpha and beta cells in this study—GNAL, NTN1 and PRKCH(Extended Data Fig. 5a,k and Supplementary Tables 9b and 10b).Exploring and visualizing DNA methylation by a web-based tool, alpha-beta-methylomeWe proceeded to develop a web-based tool, alpha-beta-methylome (https://alpha-beta-methylome.serve.scilifelab.se/app/alpha-beta- methylome/), a comprehensive open resource based on the human WGBS and RNA-seq data included in this study (Fig. 1a and Supplementary Table 1c), for investigators to explore and visualize the impact of age, sex and pre-T2D/T2D on autosomal DNA methyla- tion and/or expression of any gene in alpha and beta cells, as well as differences between alpha and beta cells, with the option to selectP-value cut-offs and DNA methylation sequencing coverage. More detailed information is available at https://alpha-beta-methylome. serve.scilifelab.se/app/alpha-beta-methylome/.Functional follow-up identifies T2D-associated genes that impair beta cell functionWe then explored whether any of the T2D-associated DEGs pre- sented in Supplementary Tables 9 and 10 have a functional role in alpha or beta cells. Protein-coding DEGs with FDR < 10% were selected for functional investigations, resulting in eight DEGs in beta cells (Extended Data Fig. 6a). Among the selected T2D-associated DEGs, ADCY9, FBLL1, ZNF431 and ZSWIM9 were downregulated, while GNAL, ONECUT2, SNX1 and TRAPPC13 were upregulated in beta cells from donors with pre-T2D/T2D (Fig. 7b,c and Extended Data Fig. 6a). A systematic PubMed search showed that little, or nothing, was known about these genes in relation to beta cell function, andGNAL, TRAPPC13, ZNF431, ZSWIM9 and FBLL1 had previously not been linked to insulin secretion or beta cell function. SNX1 defi- ciency increased cAMP and insulin secretion from MIN6B1 cells in response to a glucagon-like peptide-1 analogue 48 . Reduced Onecut2expression in rat beta cells led to increased granuphilin expression, linked to decreased insulin secretion 49 . Adcy9-deficient insulinoma cells showed reduced glucose-stimulated cAMP response and insulin secretion50 . To mimic the situation in pre-T2D/T2D, selected genes showing lower expression in beta cells from donors with pre-T2D/T2D, that is,ADCY9, FBLL1, ZNF431 and ZSWIM9, were silenced using siRNA trans- fection in both human islets, and rat INS-1 832/13 beta cells (hereafter called INS-1 beta cells; Fig. 7c–e). We chose this cell line rather than EndoC-βH1 as, in our experiments, INS-1 beta cells behaved more like primary human beta cells than the fetal human EndoC-βH1 in gene manipulation experiments (Extended Data Fig. 6b) 12 . siRNA transfection resulted in a 40–80% reduction in mRNA expression compared to islets/beta cells transfected with a negative con- trol siRNA (siNC; Fig. 7d,e). Unfortunately, the ZNF431-targeting siRNA in islets failed in silencing the expression, and Znf431 is not expressed in INS-1 beta cells. ZNF431 was hence not investigated further. Interestingly, silencing of ADCY9, FBLL1 or ZSWIM9 reduced glucose-stimulated insulin secretion (GSIS) in both human isletsFig. 7 | Functional follow-up of T2D-associated DEGs in human alpha and beta cells. a,b, Volcano plots showing T2D-associated DEGs in human alpha (a) and beta (b) cells, P < 0.05 based on DESeq2 (ref. 82). c, mRNA expression of T2D-associated beta cell DEGs selected for functional follow-up based on q < 0.1 and DESeq2 (ref. 82). Data from 15 non-diabetic donors and 7 donors with T2D. Women and men are represented by filled and open circles, respectively. Data are presented as the mean ± s.e.m. d, siRNA-mediated knockdown of ADCY9,FBLL1 and ZSWIM9 in human islets from five donors. e, Silencing of Adcy9, Fbll1and Zswim9 using siRNA in rat INS-1 832/13 beta cells. n = 4 (Fbll1) or 6 (Adcy9 andZswim9) biological replicates. f, Insulin secretion from human islets stimulated with 1 mM and 20 mM glucose. Measurements were performed on islets from the same five donors as in d. Data are presented as the mean ± s.e.m. g, Insulin secretion from INS-1 832/13 beta cells during 1 h of stimulation with low (2.8 mM) or high (16.7 mM) glucose. n = 6 biological replicates. h, Depolarization-induced insulin secretion (35 mM K + ) from INS-1 832/13 beta cells. n = 6 biological replicates from the same experiments as e. i, Overexpression of GNAL (OE-GNAL), ONECUT2 (OE-ONECUT2), SNX1 (OE-SNX1) or TRAPPC13 (OE-TRAPPC13) in human islets from five (SNX1) or six (GNAL, ONECUT2 and TRAPPC13) donors.j, OE-GNAL, OE-ONECUT2, OE-SNX1 and OE-TRAPPC13 in INS-1 832/13 beta cells. n = 5 (ONECUT2) or 6 (GNAL, SNX1 and TRAPPC13) biological replicates.k, Insulin secretion from human islets stimulated with 1 mM or 20 mM glucose for 1 h. Measurements were performed on islets from the same donors as in i.l,m, Insulin secretion from INS-1 832/13 beta cells during 1 h of stimulation with 2.8 mM or 16.7 mM glucose in absolute values (l) and as fold change, that is, secretion at 16.7 mM divided by secretion at 2.8 mM glucose (m). n = 6 biological replicates. n, Depolarization-induced insulin secretion by stimulation with 35 mM K+ in INS-1 832/13 beta cells. n = 3 biological replicates. In d–n, data are presented as the mean ± s.e.m. and *P < 0.05, **P < 0.01, ***P < 0.001 versus siNC or GFP, based on two-tailed (d–n) paired Student’s t-tests. The same colour represents paired samples included in the same experiment, for human islets from the same donor and for the INS-1 832/13 beta cells from the same cell passage. The q values for b,c were corrected for multiple testing, while the otherP values throughout were not.Fig. 6 | Key islet and T2D genes exhibit pre-T2D/T2D-associated DMRs in human beta cells. a, Volcano plot showing hypomethylated and hypermethylated DMRs in pre-T2D/T2D cases (n = 3) versus non-diabetic controls (n = 11) in beta cells (P < 0.05 based on dmrseq 25 and ten permutations). The ‘methylation difference’ on the x axis represents the beta coefficient values generated by dmrseq and included in Supplementary Table 8a. b, Selected enriched KEGG pathways among the 3,020 genes with T2D-associated DMRs in beta cells. For exact q values, corrected for multiple testing, see Supplementary Table 8d. c–j, T2D-associated DMRs in beta cells of genes previously linked to T2D and/or islet function, including EPS8L1 (c), VEGFA (d), NKX6.3 (e), CTNND2(f), RGS6 (g), FEM1B (h), SEC16A (i) and SORBS1 (j). The DMRs were analysed and visualized by dmrseq 25 including ten permutations. k, The T2D-associated DMRs in beta cells are enriched for putative binding motifs of TF such as FOXA3, FOXA2, HNF1B, OCT4, FOXA1, PAX7, FOXP1, BACH2 and PDX1, as analysed by HOMER (http://homer.ucsd.edu/homer/motif/, P < 0.05). l, Venn diagram and table showing T2D-associated DMRs in beta cells overlapping with binding sites for five islet-specific TFs: FOXA2, MAFB, NKX2.2, NKX6.1 and PDX1 (ref. 24). Overlaps are enriched with P ≤ 0.001 based on region-based permutation tests. For individual P values, see Supplementary Table 8f. m, 153 of 849 (18%) T2D candidate genes identified by GWAS 2 overlap with genes that had T2D-associated DMRs in beta cells annotated to them (HES = 2.69, P = 3.13 × 10−85 based on a one- tailed hypergeometric test). n, A T2D-associated SNP 2 (rs362307; dashed line) is located in a T2D-associated beta cell DMR annotated to HTT. o–q, T2D-associated differences in the absolute DNA methylation level (%) of individual CpG sites annotated to GCG (o), INS (p) and PDX1 (q) in beta cells from 11 (o and q) or 8 (p) non-diabetic donors and 3 donors with pre-T2D/T2D. *P < 0.05, **P < 0.01 based on two-tailed Wilcoxon rank-sum test. Data are presented as the mean ± s.e.m. Women and men are represented by filled and open circles, respectively. Inc–j and n, locations of the methylation loci are shown with black lines at the bottom, and average, smoothed lines based on the dmrseq analysis are shown in pink (pre-T2D/T2D) or blue (controls), including the standard deviation. The shaded area shows the DMR. Annotations to CpG island regions and genes are shown below each graph. ### Page 13 Nature Metabolism | Volume 8 | April 2026 | 957–980 969Resource https://doi.org/10.1038/s42255-026-01498-9and INS-1 beta cells (Fig. 7f,g) compared to siNC, with ADCY9 defi- ciency showing the strongest effect. Both Adcy9-deficient andZswim9-deficient INS-1 beta cells showed reduced insulin secretion in response to K+ stimulation relative to siNC (Fig. 7h), indicative of issues downstream of closing ATP-sensitive potassium channels in the insulin secretion pathway. Next, the selected genes that exhibited higher beta cell expres- sion from donors with pre-T2D/T2D, GNAL, ONECUT2, SNX1 ande0 0.5 1.0 1.5 2.0 0 0.5 1.0 1.5 2.0 0 0.5 1.0 1.5 2.0 siNC siADCY9 siFBLL1 siZSWIM9 1 mM 20 mM 0 0.5 1.0 1.5 0 0.5 1.0 1.5 0 0.5 1.0 1.5f g h i dGFP OE-ONECUT2OE-SNX1OE-TRAPPC13OE-GNALj k lHuman islets INS1 β-cells Human islets Human islets 2.8 mM 16.7 mM 0 5 10 15 20 0 10 20 30 40 0 500 1,000 1,500 2,000 2,500 mRNA expression (normalized counts)SNX1 ONECUT2 ZNF431 ZSWIM9 ADCY9 GNAL TRAPPC13 FBLL1 ADCY9 expressionFBLL1 expressionZSWIM9 expressionAdcy9 expressionFbll1 expressionZswim9 expression siNC siADCY9 siNC siFBLL1 siNC siZSWIM9 siNC siAdcy9 siNC siFbll1 siNC siZswim9 Insulin secretion (% of total content) Insulin secretion (ng/mg protein/h) Insulin secretion (ng/mg protein/h)ONECUT2 expression 0 40 80 120SNX1 expression 20 40TRAPPC13 expression 00 200 400 600 800 1,000GNAL expression 0 500 1,000 1,500 2,000 siAdcy9 siNC siZswim9 siFbll1 GFP OE-ONECUT2GFP OE-SNX1GFP OE-TRAPPC13GFP OE-GNAL GNAL expression 2.0 × 105 1.5 × 105 1.0 × 105 0ONECUT2 expression 1.5 × 104 1.0 × 104 0.5 × 10 4TRAPPC13 expression 4 × 105 2 × 105 1 × 105 0SNX1 expression 6 × 103 4 × 103 2 × 103 0 GFP GFP GFP GFP OE-GNALOE-ONECUT2OE-TRAPPC13OE-SNX1Insulin secretion (% of total content) Insulin secretion (% of total content) Insulin secretion (% of total content) 0 2 4 6 8 2.8 mM 16.7 mM 0 100 200 300 400 500 0 100 200 300 400 500 0 3 × 10 5 0.5 × 105 ND T2D 1.0 mM 20.0 mM 0 5 10 15 P = 0.115P = 0.177n0 4 8 12 16 Insulin secretion (fold change) GFP OE-ONECUT2OE-SNX1 OE-TRAPPC13OE-GNALGFP OE-ONECUT2OE-SNX1 OE-TRAPPC13OE-GNAL P = 0.051 Human islets INS1 β-cellsAPC2 STAB1 ADGRD1 MYH14 SEC22A GOLGA8A AHR ATAD2 NOXA1 ASAP2 SLC12A5 SLC9C1 TMEM167A PRPF31 SULT1C2 ZBTB43AADAT TRAPPC13 ESRRG DUOX2Downregulated in T2D: 157 genes Upregulated in T2D: 188 genes Cut-off: P value 0.05 6 4 2 0 −3 −2 −1 0 1 log 2 fold change log 2 fold change 2 3 4 −log10(P value)SNX1 ZNF431 ONECUT2 ADCY9 ZSWIM9 GNAL TRAPPC13FBLL1Cut-off: q-value 0.1 271 genes 294 genes Cut-off: P value 0.05 −3 −2 −1 0 1 2 3 −log10(P value)a b 6 4 2 0 Downregulated in T2D: 4 genes Upregulated in T2D: 4 genesm60csiNCsiAdcy9 siFbll1 siZswim9GFP OE-ONECUT2OE-SNX1OE-TRAPPC13OE-GNALINS1 β-cells INS1 β-cells * * * * * * * * ** ** * *** * *** * **** *** *** ** ** *** * *** *** *** ** *** * ** *** * ** * ** * ** ### Page 14 Nature Metabolism | Volume 8 | April 2026 | 957–980 970Resource https://doi.org/10.1038/s42255-026-01498-9TRAPPC13, were overexpressed using lentiviral transduction in human islets and INS-1 beta cells (Fig. 7c,i,j). Interestingly, only overexpres- sion of ONECUT2 (OE-ONECUT2) significantly reduced GSIS in human islets (Fig. 7k), while OE-ONECUT2 or overexpression of TRAPPC13(OE-TRAPPC13) significantly reduced GSIS in INS-1 beta cells (Fig. 7l). It is worth mentioning that in INS-1 beta cells, OE-ONECUT2 caused a severe reduction in GSIS and elevated basal secretion, similar to what is observed in individuals with T2D 51 . These changes resulted in reduced fold change of insulin secretion, that is, secretion at high glucose divided by secretion at low glucose, for OE-ONECUT2 and OE-TRAPPC13 (Fig. 7m). Overexpression of neither candidate affected K+ -stimulated insulin secretion in INS-1 beta cells (Fig. 7n), suggesting that OE-ONECUT2 and OE-TRAPPC13 impair insulin secretion through mechanisms acting proximal to closure of ATP-sensitive potassium channels, for example in mitochondrial function.OE-ONECUT2 causes reduced mitochondrial activity and Adcy9 deficiency impairs stimulus–secretion couplingBecause manipulation of ADCY9 and ONECUT2 had the strongest effects on insulin secretion (Fig. 7d–m), we decided to dissect the molecular mechanisms of these perturbances in INS-1 beta cells. We confirmed the Adcy9 deficiency and OE-ONECUT2 at protein levels by western blot analysis (Extended Data Fig. 6c,d). To determine whether Adcy9 defi- ciency and/or OE-ONECUT2 affect mitochondrial metabolism, we meas- ured oxygen consumption rates (OCRs) using the Seahorse analyser. OE-ONECUT2 reduced mitochondrial respiration with lower basal and glucose-stimulated respiration, ATP production-coupled respiration, proton leak, maximal mitochondrial respiration and spare respiratory capacity in INS-1 beta cells, versus GFP (Fig. 8a). Adcy9 deficiency did not affect the OCR (Extended Data Fig. 6e). In line, OE-ONECUT2 caused a small significant reduction in ATP5A in complex 5 of the electron transport chain (Extended Data Fig. 6f). Using PercevalHR to assess ATP:ADP in live single cells52 , we found impaired cytosolic ATP lev- els in OE-ONECUT2 INS-1 beta cells (Fig. 8b), demonstrating reduced mitochondrial activity in these beta cells. These data support that OE-ONECUT2 dampens mitochondrial activity, while Adcy9 deficiency impairs stimulus–secretion coupling in beta cells. The latter fits well with the extensive literature on the role of cAMP in enhancing insulin granule exocytosis in beta cells 53 . To further characterize the mechanisms underlying the dysfunc- tion caused by ADCY9 deficiency and OE-ONECUT2, we performed global transcriptomic analyses in human islets. We found 57 and 114 DEGs in ADCY9-deficient and OE-ONECUT2 islets, respectively (q < 0.1; Supplementary Tables 11a and 12a). Interestingly, NR4A3and PCSK2 were DEGs in both ADCY9-deficient islets and in T2D beta cells (Supplementary Tables 10a and 11a). In addition to ONECUT2, also SYNPO2, CFB, VEGFB and PER3 were DEGs in both OE-ONECUT2islets and T2D beta cells (Supplementary Tables 10a and 12a).HDAC9 and AZIN1 are T2D-associated genes in GWAS 2 , and DEGs inADCY9-deficient (Supplementary Table 11a) and OE-ONECUT2 islets (Supplementary Table 12a), respectively. In line with the impaired ATP production in OE-ONECUT2 INS-1 beta cells, OE-ONECUT2 human islets exhibited reduced expression of genes involved in oxidative phosphorylation (NDUFA1, NDUFA7 and COX10; Extended Data Fig. 6g) and glycolysis (GPI, ALDOA, ENO1 and ENO2; Supplementary Table 12a). Using gene-set enrichment analysis (GSEA), we tested whether sets of related genes had altered expression in ADCY9-deficient or OE-ONECUT2 islets. GSEA yielded 3 and 50 gene sets with downregu- lated expression, and two and one gene sets with upregulated expres- sion in ADCY9-deficient and OE-ONECUT2 islets, respectively (q < 0.05; Supplementary Tables 11b and 12b). Interestingly, in OE-ONECUT2islets, downregulated gene sets include cellular response to carbohy- drates and glucose, and regulation of peptide secretion and insulin secretion (Fig. 8c,d). Individual genes contributing to enrichment for ‘regulation of insulin secretion’ in OE-ONECUT2 islets are presented in Fig. 8d and Extended Data Fig. 6h. In ADCY9-deficient islets, upregu- lated gene sets were linked to apoptosis (Extended Data Fig. 6i and Supplementary Table 11b), and cAMP signalling and insulin secretion were among enriched KEGG pathways (Supplementary Table 11c). Next, because ONECUT2 is a TF, we used weighted gene correlation network analysis (WGCNA)54 on 114 DEGs from OE-ONECUT2 in human islets (q < 0.1; Supplementary Table 12a) to test whether ONECUT2 is part of a dysregulated network. WGCNA revealed a network with ONE- CUT2 and 107 other genes, where ONECUT2 had stronger connection to 30 DEGs (Fig. 8e and Supplementary Table 12c). Four genes (ONE- CUT2, SYNPO2, PER3 and VEGFB) were both in the WGCNA ONECUT2network and among T2D-associated beta cell DEGs (Supplementary Table 10a). We proceeded exploring Onecut2 in a rodent diabetes model, the GK rat. We recently demonstrated mitochondrial dysfunction in GK islets12 . In line with beta cells from donors with pre-T2D/T2D show- ing elevated ONECUT2 expression (Fig. 7c), islets from GK rats had increased Onecut2 expression versus non-diabetic Wistar rats (Fig. 8f). In line with previous studies, islet Ins1 expression was decreased (Pnominal = 0.12) and blood glucose increased in GK versus Wistar rats (Fig. 8g,h). Finally, we showed that silencing ONECUT2 in EndoC-βH1 beta cells exposed to high glucose plus palmitate (HGP; 19 mM glucose plus 1 mM palmitate), mimicking a diabetogenic environment, could rescue the perturbed insulin secretion seen in beta cells exposed to HGP (Extended Data Fig. 6j,k). Our data support ONECUT2 as a key transcription factor in beta cells, controlling expression of genes involved in mitochondrial func- tion, glucose homeostasis and insulin secretion.DiscussionHere, we provide a unique resource, alpha-beta-methylome (https:// alpha-beta-methylome.serve.scilif elab.se/app/alpha-beta-methyl- ome/), where the community can explore associations between T2D, age and sex, and cell-type specific methylation and gene expressionFig. 8 | Dysregulation of ONECUT2 leads to perturbed mitochondrial function, reduced insulin secretion and dysregulation of many genes in beta cells.a, Reduced OCR in INS-1 832/13 beta cells overexpressing ONECUT2 (OE-ONECUT2). The left graph shows the oxygen consumption trace, and the right graph the quantification of derived respiratory parameters. n = 4 independent experiments. *P < 0.05 based on two-tailed paired Student’s t-tests. b, Reduced ATP production in OE-ONECUT2 INS-1 832/13 beta cells, n = 73 (GFP) and 116 (OE-ONECUT2) cells from six different experiments. ***P = 4.61 × 10−14 based on a two-tailed Student’s t-test (right). c, emapplot visualizing results of the GSEA of the transcriptome data generated in OE-ONECUT2 versus control (GFP) human islets of six non-diabetic donors presented in Supplementary Table 12. The R packages clusterProfiler and enrichplot were used to generate the emapplot, and 23 gene sets are preselected. GSEA was corrected for multiple testing. d, The genes contributing to the enrichment for the downregulated gene set ‘regulation of insulin secretion’ in the GSEA of OE-ONECUT2 human islets, as well as selected individual downregulated genes in the gene set: FOXO1 (*P = 0.026), G6PC2(*P = 0.0152), GCK (**P = 0.0087), GIPR (**P = 0.0022), GPR119 (*P = 0.0441), PDX1(*P = 0.0441) and SLC2A2 (*P = 0.0441). P values are based on two-tailed Mann– Whitney U-tests and presenting unlogged data. n = 6 biological replicates of cells expressing GFP or ONECUT2. e, WGCNA based on weighted correlations among OE-ONECUT2 DEGs with q < 0.1 showed that ONECUT2 is part of an expression cluster containing 114 DEGs (in network) and 108 are shown in the figure (due to Cytoscape cut-off 0.55), with direct connection to 30 DEGs. f–h, Male GK rats, a T2D animal model, showed higher Onecut2 (*P = 0.0458; f), and nominally lower Ins (P = 0.12; g) expression in islets, as well as higher blood glucose levels (***P = 6.11 × 10−7 ; h) compared with male control Wistar rats (n = 5) based on two- tailed Student’s t-tests. Data in a, b, d and f–h are presented as the mean ± s.e.m.P values presented in a, b, d and f–h were not corrected for multiple testing. ### Page 15 Nature Metabolism | Volume 8 | April 2026 | 957–980 971Resource https://doi.org/10.1038/s42255-026-01498-9aTime (min) 0 40 80 120 0 1 × 104 2 × 104 3 × 104 4 × 104 GFP OE-ONECUT2Glucose Oligomycin FCCP Rotenone Antimycin A OCR (pmol min−1 mg protein−1 ) Basal mitochondrial respiration Glucose-stimulated respiration ATP-production-coupled respiration Proton leak Maximal mitochondrial respiration Spare respiratory capacity 0 1 × 104 2 × 104 4 × 104 OCR (pmol min−1 mg protein−1 ) 0 1 2 3 4Onecut2 expression 0 0.5 1.0 1.5Ins1 expressionP = 0.12ABCA12, ABCC8, ACVR1C, ADCY5, ANO1, BAIAP3, BMP8A, BRSK2, CASR, CFTR, CHGA, CHRM3, CPLX1, CPT1A, DRD2, ENY2, F2RL1, FOXA2, FOXO1, G6PC2, GCK, GIPR, GPR119, HNF1B, HNF4A, IL1B, IL1RN, IRS2, KCNJ11, KLF7, LRRC8A, NEUROD1, NKX61, NR0B2, NR1D1, PCK2, PDX1, PER2, PFKL, PICK1, PLA2G6, PLCB1, PRKACA, PRKCE, PTPRN, RAF1, RBP4, REST, RIMS2, RPH3AL, SIDT2, SLC2A2, SLC8B1, SNX19, STX1A, STX4, SYT7, TFAP2B, TNF, UCN3Regulation of insulin secretion 0 20 40 60GIPR expression 0 50 100 150 200 250PDX1 expression 0 500 1,000 1,500 2,000 2,500GPR119 expression 200 400 600 800 1,000FOXO1 expression 0 0 0.5 × 105 1.0 × 105 1.5 × 105G6PC2 expression 20 40 60 80 100GCK expression 0 Cellular response carbohydrate hexose Insulin protein transport extracellular Intracellular peptide glucose homeostasis NES–1.5 –1.6 –1.7 –1.8 –1.9bTime (s) 0 200 400 600 0 0.4 0.8 1.2 1.6 [ATP]cyto PercevalHR emission (520 nm) 2.8G 16.7G Oligomycin GFP OE-ONECUT20 10 20 30 Blood glucose (mM) Wistar GK Wistar GK Wistar GKe c f g h0.7 1.4 OE-ONECUT2 vs GFP 0.4 1.1 1.7ONECUT2 S100A5 PRKY POTEF SEMA3E DNAJB1 1 PANX3 EQTN DUX4L8 OR2A14; OR2A2 AZIN1 CLRN2 ZNF354B LRRCC1 YTHDF3 SPNS3 PPM1N SNF8 HORMAD2 MAGEL2 CMA1 BIK RRP12 JMJD7; PLA2G4B; JMJD7-PLA2G4B WDR89 CF AP46 PDE3B ORM2 ZFC3H1 LY6G6E OR52E8 NPIPB3 PFDN2 PPOX AR T1 BTNL3 GFI1B ATM TCEAL2 VEGFB C10orf25 NPNT MYOG TMEM127 LMNB1 ZNF852 IRF6 GRIK3 SAP130 CREB5 GBP7 EGR2 SEPP1 KCNQ3 FLNA SYNPO2 EIF4EBP1 ACADL ZER1 FAM219A C9orf91 INVS RHBDD2 ASB8 MUC3A SH3BP4 ATP2A3 GOLM1 GJD2CFI PCDH17 ATXN2 ATP6V1B2 CASR LOC389834; AL354822.1 PPIL2 TTC8 ABCG1 DGKB RHBDL3 PLAGL2 NR1D1 GALNT1 DOK5 CACNA1A PER3TLDC2 PTBP1; MIR4745 SUL T2B1 CLIC6 FOXF1 ERICH4 ASGR2 LDLRAD3 VPS9D1 CES1 OR8G1 AKAP5 ELMSAN1 RRN3 KA TNAL1 NAA25 KIF4B KIRREL3 MPZL3 RBM4B USP6NL PHF245.3 × 10–4 Node size: P value 1.8 × 10 –10 0 50 100 150 200 250 300SLC2A2 expression GFP OE-ONECUT23 × 104 GFP OE-ONECUT2% increase at 16.7 mM 50 0 –50 –100 100 150 GFP OE-ONECUT2d*** * * * * ** **** *** * * * * * * *** *** *** *** *** *** ***Amide transport Peptide transportPeptide hormone secretion Establishment of protein localization to extracellular region Intracellular glucose homeostasis Glucose homeostasis Regulation of peptide hormone secretion Protein localization to extracellular region Response to carbohydrate Regulation of peptide transport Regulation of protein transport Insulin secretion Protein secretion Carbohydrate homeostasis Regulation of insulin secretion Cellular response to hexose stimulus Response to glucose Number of genes 80 100 120 140 160 Response to monosaccharide Response to hexose Regulation of protein secretion Regulation of peptide secretion Peptide secretion ### Page 16 Nature Metabolism | Volume 8 | April 2026 | 957–980 972Resource https://doi.org/10.1038/s42255-026-01498-9in human alpha and beta cells. By comprehensive analyses, we dis- covered regulatory regions which were unmethylated in one cell type, while they were highly methylated in the other, near genes with central roles in alpha and/or beta cells, including INS/INS-IGF2,PDX1/PLUT1, MAFA, PCSK1, GCG and ARX. The cell-specific epigenetic pattern was clearly linked to gene expression, and half of genes with DMRs were DEGs in alpha versus beta cells. Intriguingly, by epigenetic editing using CRISPR–dCas–DNMT3A 33 , we increased INS promoter methylation, causing decreased INS expression and insulin levels in beta cells, providing evidence that methylation directly regulates transcription of this important gene. By integrating beta cell methy- lomes and transcriptomes from people with or without T2D, with comprehensive functional analyses in human islets and beta cells, we discovered that ONECUT2 has a key role in beta cell function and T2D pathogenesis. TFs establish and maintain alpha and beta cell identity. Dysregu- lation of TFs contributes to islet dysfunction and diabetes55,56 . By dis- secting methylomes, we discovered DMRs with large methylation differences in alpha versus beta cells, annotated to such TFs. Meth- ylation in these DMRs likely controls TF expression and thereby cell fate. For example, ARX, essential for alpha cells, is hypomethylated and highly expressed in alpha cells, while it is hypermethylated and non-expressed in beta cells. MAFA and PDX1, essential for beta cells, were hypomethylated and highly expressed in the beta cells. We found cell-specific DMRs for NEUROG3, KLF4 and PAX6, important for alpha and beta cells56 . Long non-coding RNAs control TFs in alpha and beta cells. We identified DMRs annotated to PLUT, regulating PDX1 andPAUPAR, regulating PAX6 (refs. 57,58). We found cell-specific DMRs inHNF1A, HNF1B and HNF4A, TFs contributing to maturity-onset diabe- tes of the young, and in PAX5, an upregulated TF in islets from people with T2D 26 . By using our web tool, alpha-beta-methylome, we found T2D-associated methylation changes annotated to these TFs and long non-coding RNAs. In line with this, we previously discovered hyper- methylation of PDX1 in islets of donors with T2D8,11 . Additionally, DMRs identified in this study overlap with sequences bound by FOXA2, MAFB, NKX2.2, NKX6.1 and PDX1. T2D-associated SNPs impact islet function 2 . Previously, we dis- covered associations between genetics and methylation as well as the chromatin structure in human islets 30,31,59 . Here, we found that numerous genes linked to T2D SNPs also have T2D-associated DMRs in alpha or beta cells, and/or show differential methylation between alpha and beta cells, for example, INS, TCF7L2 (top T2D gene) 2 andGLP1R (treatment target) 60 , as well as TFs mentioned above. Addition- ally, T2D SNPs are located within DMRs, further supporting interac- tions between genetics and epigenetics in T2D. We also found that 45% and 22% of T2D-associated genes/loci identified by GWAS overlap the alpha versus beta cell DMRs and DEGs, respectively. The reason for the higher overlap with genes annotated to DMRs than to DEGs could be because DNA methylation is more stable than gene expres- sion, where expression of some genes represents more of a ‘snap- shot’ at the time samples were collected. However, this needs to be further investigated. We add ONECUT2 ‘as a piece of the puzzle’ of genes causing ham- pered mitochondrial activity and reduced GSIS in islets/beta cells, contributing to T2D (Extended Data Fig. 6l) 12,26,61,62 . When analysing ~300,000 beta cells per donor, ONECUT2 expression was increased ~2-fold in people with pre-T2D/T2D, and ONECUT2 expression was higher in beta versus alpha cells. OE-ONECUT2 in human islets and INS-1 beta cells, mimicking T2D, caused reduced GSIS, OCR, ATP produc- tion and expression of gene sets regulating glucose metabolism and insulin secretion. The latter includes reduced expression of GCK, which encodes glucokinase, the rate-limiting enzyme in glycolysis. This could be an important change, leading to reduced production of pyruvate to fuel the mitochondrial tricarboxylic acid cycle, oxidative phospho- rylation and ATP production63 . Rodent models showed downregulatedOnecut2 when overexpressing miR-29a/miR-29b/miR-29c64 , and links between insulin resistance and downregulated ductal Onecut2/ONE- CUT2 (ref. 65), supporting differences between mice and men66 . In 10x Genomics’ scRNA-seq data from non-diabetics (5,000–10,000 islet cells), ONECUT2 is predominantly expressed in ductal cells and to a lesser degree in beta cells, with limited expression in other cell types 67 . While our data strongly implicate ONECUT2 in T2D pathogenesis, one should keep in mind that the overexpression of ONECUT2 is stronger in our in vitro model than in T2D beta cells in vivo. Interestingly, ONECUT2 and ONECUT1 have nearly identical DNA binding sites, but data from our study and published studies suggest different roles for these two TFs68 . Mutations and common variants inONECUT1 contribute to neonatal diabetes and T2D, respectively, and evidence supports that ONECUT1 impacts pancreas development 69–71 . However, ONECUT1 expression was not detected in beta cells in our study, nor in fetal or adult beta cells sorted by Blodgett et al. 27 . Fur- thermore, there was a 10-fold higher expression of ONECUT2 in adult compared to fetal beta cells27 , and ONECUT1 showed very low expres- sion compared to ONECUT2 in beta cells in our human islet single-cell RNA-seq data67 . This study focused on dissecting cell-specific and pre-T2D/ T2D-associated DNA methylation in alpha and beta cells. However, since we previously showed that age and sex impact DNA methyla- tion, gene expression and insulin secretion in human islets 72,73 , we included analyses of age and sex associations on methylation and gene expression in our web tool, alpha-beta-methylome. Interestingly, sex-associated methylation and gene expression differences previously found in islets of, for example, BCL11A73 , were also identified in beta cells by exploring alpha-beta-methylome. Bcl11a-silencing reduced GSIS in beta cells, mimicking lower BCL11A expression and insulin secretion found in islets from males 73 . Additionally, age-associated beta cell methylation of, for example, KLF14, FHL2 and ZNF518B, was identified using our web tool. These genes also showed age-associated methylation in islets 72 . We also explored whether results generated by Manduchi et al. 28 could be validated by using our web tool and, for example, we found age-associated hypomethylation of specific CpG sites in beta cells of genes involved in insulin secretion, such as FFAR1,RAPGEF4, RAB11FIP5, SMAD2, MYO5A, PDX1 and PTPRN. It should be noted that while our web tool performs a linear model between age and DNA methylation of individual CpG sites in beta cells from donors with an age range between 36 and 81 years, Manduchi et al. measured the average methylation differences between their older (39–58 years) and younger (18–35 years) donors. This study provides a valuable resource for islet research by iden- tifying genomic regions where DNA methylation seems to regulate cell-specific gene expression in human alpha and beta cells. For exam- ple, we discovered a region annotated to the promoter of SLC2A2, encoding GLUT2, with hypomethylation in beta cells compared to alpha cells. In line with this, we found SLC2A2 expression in beta but not alpha cells. GLUT2 is an important glucose transporter in rodent islets74 , but its importance in human islets has been questioned 75 . Nevertheless, Hattersley’s group found SLC2A2 mutations causing neonatal diabetes, and they proposed that GLUT2 affects insulin secretion in humans 76 . The epigenetic and expression data of SLC2A2 in human beta cells in the present study together with data from our previous study, where we found reduced SLC2A2 expression in islets from donors with T2D and that SLC2A2 knockdown reduced insulin secretion in human islets26 , support a role for GLUT2 in mature beta cells. We used different cut-offs for unexpressed genes in our study. We first used <2 mean normalized counts to categorize genes as not expressed, and we categorized the remaining genes into low, medium and high expression. In the second analysis, we used ≥20 mean normal- ized counts in ≥80% of samples for a gene to be considered expressed in alpha and beta cells. The reason for using different cut-offs was that in the second analysis we wanted to include expressed genes that could ### Page 17 Nature Metabolism | Volume 8 | April 2026 | 957–980 973Resource https://doi.org/10.1038/s42255-026-01498-9have a biological role in the cells, while in the first analysis our goal was to investigate the degree of DNA methylation in relation to four different levels of expression. However, one should keep in mind that the cut-offs are based on our ‘best guess’. This study has strengths and limitations. By using WGBS, we cov- ered most of the methylome, compared with DNA methylation arrays, covering ~5%. We provide alpha and beta cell methylomes and tran- scriptomes from the same donors. As the number of samples is limited, partly because MARIS requires many islets and WGBS is challenging, future studies should include additional donors. While this study provides new data, some cell-specific findings could be ‘validated’ in larger cohorts of human islets, for example sex and age associations described above72,73 , and T2D associations described in the results, for example, 65 T2D-associated beta cell DMRs overlap 91 CpGs exhibiting differential islet methylation in T2D 12 , supporting their robustness. Functional validations and results from pathway analyses further sup- port the validity of our data. Interestingly, 75.5% of the T2D-associated beta cell DMRs identified in this study were hypomethylated, which is in line with our previous study where 77% of T2D-associated methylation sites had lower methylation in whole islets from donors with T2D12 , and the recent study by Manduchi et al. supporting hypomethylation in beta cells from donors with T2D28 . Additionally, genes discovered to have alpha versus beta cell DMRs in our study significantly overlapped with the ones found by Manduchi et al., and both our study and theirs found enriched gene pathways affecting T2D and secretion. A limitation is that only a few T2D-associated alpha cell DEGs could be validated in two other studies26,46 . However, one needs to keep in mind that Walker et al. used a gentler cut-off for inclusion of genes compared to our study 46 . Additionally, since T2D is a heterogeneous and polygenic disease, diverse mechanisms and different genes contribute to the disease in different groups of people2,77 . This may affect the reproducibility between some studies, despite technical soundness. Together, we provide comprehensive cell-specific methylomes and transcriptomes in alpha and beta cells from people with or without T2D and a web tool allowing the exploration of epigenetic dysregula- tions contributing to impaired insulin secretion and potentially T2D.MethodsEthics statementWritten informed consent to donate organs for biomedical research was obtained from the pancreatic donors or their relatives, and all procedures were approved by the ethics committees at Uppsala Uni- versity and Lund University (permit nos. 2007-05 and 2011-263) follow- ing the Helsinki declaration. Donors were not compensated for their participation. Animal experiments were performed with permission from the Animal Ethics Committee of Lund University (permit no. 5.8.18-04115/2021) in accordance with the legal requirements of the European Community (86/609/EEC).Human pancreatic isletsPancreases were obtained from multi-organ donors, and pancreatic islets were isolated by collagenase digestion and density gradient purification at the Nordic Network for Clinical Islet Transplantation in Uppsala, Sweden. The findings are based on inclusion of both sexes, determined by self-reporting and concordant genetic testing. Sex is considered as a covariate in genome-wide analyses as described below. Supplementary Table 1a presents the clinical characteristics for all 24 donors included in this study, while Supplementary Table 1b presents characteristics for the donors included in each individual analysis of this study. These analyses include (i) pairwise analyses, that is, comparing data in alpha versus beta cells, in donors not diagnosed with T2D and with data available for both alpha and beta cells, and (ii) comparing data in alpha or beta cells from donors with pre-T2D/ T2D, based on HbA1c ≥ 6% (≥42 mmol mol−1 ) and/or T2D diagnosis, with data in alpha or beta cells from controls, including donors with HbA1c < 6% (<42 mmol mol−1 ). Additionally, Supplementary Table 1c presents the characteristics of the donors included in each possible analysis using the web-based tool alpha-beta-methylome (https:// alpha-beta-methylome.serve.scilifelab.se/app/alpha-beta-methy- lome/), which also includes (iii) associations between age and DNA methylation or gene expression using linear regression models in alpha or beta cells from non-diabetic donors, and (iv) sex associations by comparing data in alpha or beta cells from male versus female non- diabetic donors.MARIS staining and FACS sorting of human alpha and beta cellsMARIS23 was used to sort alpha and beta cells from islets from 24 donors. Briefly, islets were dispersed and fixed and then stained with antibodies for 30 min with guinea pig anti-insulin (A0564, 10119708; 1:600 dilu- tion, Dako) and mouse anti-glucagon (MAB1249, HSQ0416111; 1:200 dilution, R&D Systems). Cells were then washed twice in DEPC-treated PBS containing 10% BSA, 1% saponin and RNase inhibitor before sec- ondary staining with donkey anti-guinea pig Alexa Fluor 647 (706-605- 148, 129003; 1:400 dilution, Jackson ImmunoResearch) and donkey anti-mouse Alexa Fluor 488 (A-21202, 1644644; 1:400 dilution, Invit- rogen). The sorting of alpha and beta cells was performed on an Aria Fusion (BD Biosciences), achieving purity of 92.0% ± 7.6% for alpha cells and 96.5% ± 4.0% for beta cells. We obtained 322,627 ± 73,055 alpha cells and 311,550 ± 61,250 beta cells from each donor. When the number of sorted cells from a donor was deemed sufficient, they were split into two portions for extraction of RNA and DNA, respec- tively. Sample purity was also determined by deconvolution according Loyfer et al. 20 using GCG and INS as marker genes for alpha and beta cells, respectively, and either the gene expression (purity of alpha cells = 99.3% ± 0.3% and beta cells = 99.5% ± 0.1%) or DNA methylation (purity of alpha cells = 94.2% ± 0.8% and beta cells = 99.1% ± 0.4%) data generated by RNA-seq or WGBS (see below). Purity for individual sam- ples was included in Supplementary Table 1a.RNA and DNA isolation from human alpha and beta cellsAfter sorting, the alpha and beta cells were pelleted by centrifugation at 3,000g for 6 min at 4 °C. Total RNA and DNA were then extracted from the cells using the RecoverAll Total Nucleic Acid Isolation kit (Ambion, Life Technologies) according to the recommended protocol from the manufacturer, except that we incubated the RNA samples in digestive buffer for 3 h at 50 °C and then for 15 min at 80 °C instead of 15 min at 50 °C and 15 min at 80 °C, according to the MARIS protocol. The quality and integrity of the RNA and DNA samples were analysed using the Agilent ScreenTape and Fragment Analyzer systems (Agilent Technologies). Next, RNA and DNA concentrations were measured by Qubit/Quant-iT assay (Thermo Fisher Scientific) to verify sufficient quantity and quality for library preparations.RNA-seqSequencing libraries were generated from 1 μg RNA from alpha and beta cells using the TruSeq RNA Library Preparation Kit followed by RNA-seq on a NextSeq 500 (Illumina). For preparation and analysis of data, quality and adaptor trimming was done using Trim Galore (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/), and then Salmon was used for quantification of transcript expres- sion 78 . Quality control was performed using fastQC (https://www. bioinformatics.babraham.ac.uk/projects/fastqc/) and multiQC tools 79 . Downstream analyses were done using R80 . The R package tximport81 was used to import transcript-level abundance, estimated counts and transcript lengths into R using Gencode annotation v32. Differential gene expression analysis was done using DESeq2 (ref. 82) which applies Wald’s test. For the analysis of DEGs and linear models, we considered a transcript to be expressed in a particular cell type if it has at least 20 reads in 40% of the samples. Correction for age, sex, body mass index (BMI) and days in culture was included in the analyses. Gene symbols ### Page 18 Nature Metabolism | Volume 8 | April 2026 | 957–980 974Resource https://doi.org/10.1038/s42255-026-01498-9referring to human mRNA expression are capitalized and italicized, and mRNA expression is referred to as ‘gene expression’.WGBS and data analysisGenomic DNA (2.2–50 ng) was treated with sodium bisulfite using the EZ DNA Methylation-Gold D5005 kit (Zymo Research) according to the manufacturer’s protocol. Sequencing libraries were prepared using the SPlinted Ligation Adapter Tagging (SPLAT) protocol83 . Cluster genera- tion and paired-end 150-bp read length sequencing were performed using S4 flow cells with a NovaSeq 6000 and v1.5 sequencing chemis- try (150 cycles, Illumina). The number of reads, alignment statistics and sequencing depth are reported in Supplementary Table 2. The sequencing data were processed using R80 and MultiQC79 . Paired-end reads were trimmed using Trim Galore 84 . Genome indexing was done using Genome Reference Consortium Human Build 38 (GRCh38/hg38). Gene annotations were retrieved using the R package TxDb.Hsapiens. UCSC.hg38.knownGene (version 3.4.6. 2019). Sequencing reads were aligned, and count data were generated using Bismark 85 . Extended Data Figs. 7 and 8 present heat maps and dendrograms of the WGBS samples, respectively. To identify DMRs, we used dmrseq25 , which uses transformed meth- ylation proportions, and fits a linear regression model using generalized least squares with a nested autoregressive correlated error structure. The dmrseq package ‘weighs’ samples based on coverage and methylation state, meaning a CpG with 10 of 10 reads methylated is not treated the same as a CpG with 50 of 50 reads methylated, that is, higher coverage allows for more certainty of the methylation state. For the DMR calling, we used the following settings: >2 coverage in 70% of samples, a minimum DMR length of three CpGs and ten permutations. The t-statistics from Wald’s test were used to evaluate the strength of the covariate’s effect within each region, and statistical significance was determined via per- mutations using a pooled null distribution. The permutations involved calculating an initial test statistic from the observed data, then repeat- edly shuffling the data and recalculating the statistic to build a reference distribution. The final P value is the proportion of shuffled results that are as extreme or more extreme than the original statistic. Additionally, dmrseq’s cut-off for the scalar value that represents the absolute value for the cut-off of the single CpG coefficient that is used to discover candidate regions was set to 0.20 for the analysis of alpha versus beta cells and to 0.05 for the analysis of pre-T2D/T2D versus controls. DMR annotations to gene and CpG island regions were done using hg38 and annotatr 86 . The analysis of DMRs on autosomal chromosomes included data from both males and females, while X-chromosome and Y-chromosome DMRs were analysed in males and females separately.HOMER TF motif discoveryHOMER (v.4.11, http://homer.ucsd.edu/homer/motif/) was run with findMotifsGenome.pl, to search for enrichments of TF binding motifs in identified DMRs.Culturing cell linesThe human EndoC-βH1 beta cell line (female origin, EndoCells) was seeded in Matrigel and fibronectin-coated (100 μg ml −1 and 2 μg ml −1 , respectively; Sigma-Aldrich) culture vessels in DMEM contain- ing 5.6 mM glucose, 2% BSA fraction V, 10 mM nicotinamide, 50 μM 2-mercaptoethanol, 5.5 μg ml −1 transferrin, 6.7 ng ml −1 sodium selenite, 100 units per ml penicillin and 100 μg ml −1 streptomycin. The rat INS-1 832/13 beta cell line (male origin, shared by C. New- gard, Duke Molecular Physiology Institute) was cultured in RPMI 1640 medium with 11.1 mM D-glucose supplemented with 10% FBS, 2 mM L-glutamine, 1 mM sodium pyruvate, 50 μM β-mercaptoethanol, 5 ml penicillin–streptomycin (10,000 U/10 mg ml−1 ) and 10 mM Hepes (HyClone). All cells were incubated in a humidified atmosphere with 5% CO 2 at 37 °C.Epigenetic editing, pyrosequencing, MethylationEPIC v2.0 array analyses, real-time qPCR and insulin and glucagon ELISAThree different sets of epigenetic editing experiments were done in human EndoC-βH1 beta cells, which were seeded (180,000 cells per well) in 48-well plates 48 h before transduction. Next, the cells were transduced with all-in-one lentiviral constructs 33 at a multiplicity of infection (MOI) of 10 in the growth medium without antibiotics (peni- cillin–streptomycin) for 6 h. The constructs contain inactivated Cas9 (dCas9) fused to the DNA methyltransferase DNMT3A, and eitherINS-DMR-gRNA, a gRNA (sequence TGAGACTATAAAGCCAGCGGGGG) targeting an INS DMR (Figs. 2e and 4a,b), or TH-DMR-gRNA, a gRNA (sequence GCAGCTGTACTGGTTCACGG) targeting TH (Fig. 4c) as well as a potential INS enhancer (Fig. 4a,c). In the third set of editing experiments, EndoC-βH1 beta cells were transduced with a construct containing GCG-DMR-gRNA, a gRNA (sequence CATGCGTGATT- GAAAGTAGAGCG) targeting the DMR presented in Figs. 2g and 4d, dCas9 fused to TET1, an enzyme involved in demethylating DNA. The virus-containing medium was then removed and replaced with com- plete growth medium. Forty-eight hours after transduction, cells were transferred into 24-well plates, and after an additional 48 h, cells were put into selection medium, including 4 μg ml −1 puromycin, for 7 days. The selection medium was renewed after 3 days. After puromycin selec- tion, cells from the same well were collected for RNA and DNA extrac- tion, and cells from different wells were collected for protein extraction. DNA and RNA were then prepared from the EndoC-βH1 beta cells using the Allprep RNA/DNA Mini Kit (80204, QIAGEN, Hilden, Ger- many) according to the manufacturer’s instructions. The DNA and RNA concentrations were measured using a NanoDrop (ND-1000; Thermo Fisher Scientific). Next, we performed bisulfite conversion using 200 ng of DNA using the EpiTect Bisulfite Kit (59104, QIAGEN) according to the manufacturer´s instructions. Using the PyroMark Assay design software 2.0 (QIAGEN), the following pyrosequencing assays were designed: INS-forward primer GATTTTTTTATTTTAGGTTTTAATGGG,INS-reverse primer ACAAACCTACTTAATAACCTCTTCTAATA,INS-sequencing primer TTTATTTTAGGTTTTAATGGGT, TH-forward primer GGGGTGATTTTATTTAAGGAATTTATTTAG, TH-reverse primer CCTAACCCACCTAAACTTATCCTT, TH-sequencing primer ATTGTTT- GTTTGAGGAG, GCG-forward primer GGATATGTATAAAATAGGGATGGT- TATGG, GCG-reverse primer AAAAAAACAACTTAAAAATCCACCTTCTA and GCG-sequencing primer AGGGATGGTTATGGG. Pyrosequenc- ing was then run on a PyroMark Q48 autoprep instrument (QIA- GEN). Notably, several pyrosequencing assays covering the genomic regions targeted by and surrounding the three investigated gRNAs failed, potentially due to CG-rich regions. Subsequently, we were unable to analyse methylation of all CpG sites in these regions by using pyrosequencing. We proceeded to use the Infinium MethylationEPIC v2.0 BeadChip array (Illumina) to study DNA methylation of CpG sites covered by EPIC v2.0 probes in the INS, TH and GCG loci in samples from the epigenetic editing experiments. These arrays were analysed at Eurofins, Denmark. DNA was treated with bisulfite using the EZ-96 DNA Methylation kit (Zymo Research). Preprocessing and quality control of the methylation data were performed using R Statistical Software version 4.3.3 (https:// www.R-project.org/). Probes on the array that target SNPs, non-CpG sites, cross-reactive probes87 , Y-chromosome-specific probes, probes targeting single-nucleotide variants of common cancer driver muta- tions and probes with a detection P value > 0.01 in more than 10% of the samples were removed. The intensities from the remaining probes were then corrected for background and dye bias using the NOOB function from the R package minfi (version 1.48.0)88 , and intra-sample normali- zation performed with the BMIQ function included in the R package wateRmelon (version 2.8.0) 89 . To minimize batch effects, all paired editing samples (control and gRNA) were placed on the same array. Paired t-tests were used to analyse the methylation data covered by EPIC ### Page 19 Nature Metabolism | Volume 8 | April 2026 | 957–980 975Resource https://doi.org/10.1038/s42255-026-01498-9v2.0 probes from 10 kb upstream of the TSS to 10 kb downstream of the gene end of INS, TH and GCG. Here, EPIC v2.0 array covered 56 sites forINS, 62 for TH and 14 for GCG. M values were used for all bioinformatic and statistical analyses, and they were then converted to beta values (1–100% methylation) for presenting the results. cDNA was generated with the RevertAid First Strand synthesis kit (K1622, Thermo Fisher Scientific). Gene expression was analysed by real-time qPCR in triplicates on a 384-well plate with TaqMan assays (Thermo Fisher Scientific) for INS (HS00355773_m1), TH (Hs00165941_ m1), GCG (Hs01031536_m1) and PPIA (Hs04194521_s1, endogenous control), in a QuantStudio 7 Flex Real Time PCR system (Thermo Fisher Scientific) under default cycling parameters. Data collection and most analysis for epigenetic editing experiments were not per - formed blinded to the conditions of the experiment because samples needed to be loaded on plates and discs in a certain order. However, EPIC array experiments performed by Eurofins were blinded. No data points were excluded from the analyses. Total protein from wells with edited EndoC-βH1 beta cells was extracted by using 200 μl RIPA buffer: 0.1% sodium dodecyl sulfate (SDS), 150 nM NaCl, 1% Triton X-100, 50 mM Tris-Cl, pH 8 and ethylen- ediaminetetraacetic acid (EDTA)-free protease inhibitor (Roche). The protein amount/concentration was analysed using the BCA Protein Assay kit (Thermo Fisher Scientific). Insulin and glucagon levels in beta cells were then analysed by ELISA (insulin: 10-1113-01; glucagon: 10-1271-01, Mercodia). BCA and ELISA plates were read on a CLARIOstar microplate reader (BMG Labtech).Gene silencing, overexpression and insulin secretion in human pancreatic isletsHandpicked human islets (100–200) were placed in culture dishes containing RPMI 1640 media (11879020, Thermo Fisher Scientific) with 5 mM glucose, 10% FBS and 200 mM L-glutamine. Human islets were then transfected in a final volume of 2.5 ml per dish contain- ing serum free Opti-MEM, 6.25 μl of Lipofectamine RNAiMAX, 50 nM of Silencer Select Pre-Designed siRNA against ADCY9 (s1037), FBLL1(s196099), ZNF431 (s46878) ZSWIM9 (s51624) or negative control no. 2 siRNA (siNC, 4390846; all from Thermo Fisher Scientific). In line with previous studies12,26 , a second transfection was performed 24 h after the first, and all functional experiments were performed 72 h after the first transfection. For overexpression experiments, lentiviral plasmids containing the human phosphoglycerate kinase 1 promoter in front of cDNA of human GNAL, ONECUT2, SNX1 or TRAPPC13, with and without His-tag, followed by an internal ribosomal entry site and the cDNA forGFP, were cloned by GenScript. Lentiviral vectors were produced by Lund University Cell and Gene Therapy core facility. Lentivirus confer- ring gene expression of only GFP was used as control. Around 100–200 islets were transduced with these lentiviruses for 72 h with a MOI of 5. Gene symbols referring to human mRNA expression are capitalized and italicized, whereas when referring to human protein levels, protein symbols are capitalized but not italicized. For insulin secretion, six batches of 6–12 human islets per donor were pre-incubated for 30 min in KREBS buffer (2.5 mM CaCl 2, 4.7 mM KCl, 120 mM NaCl, 25 mM NaHCO3, 1.2 mM KH2 PO 4, 1.2 mM MgSO4 and 10 mM HEPES) with 1 mM glucose before stimulation for 1 h in KREBS with either 1 mM or 20 mM glucose (n = 3 for each condition). After glucose stimulation, islets were dissolved in 100 μl RIPA buffer (50 mM TRIS-HCl, 150 mM NaCl, 0.5 mM sodium deoxycholate, 2 mM EDTA, 50 mM NaF, 1% Triton-X, 0.1% SDS). Insulin was measured by ELISA (10-1113-01, Mercodia), and secreted insulin was normalized to the total insulin content in the islets.Gene silencing and overexpression in INS-1 beta cells and EndoC-βH1 beta cellsThe INS-1 832/13 beta cells were seeded (250,000 cells per well) in 24-well plates containing 1 ml medium without antibiotics a day before transfection. For gene silencing, cells were transfected the following day, in a total volume of 600 μl of serum free Opti-MEM and 1.5 μl of Lipofectamine RNAiMAX with 25 nM Silencer Select Pre-Designed siRNA against Adcy9 (s235414), Fbll1 (s170925) or Zswim9 (s1587760), or silencer select negative control no. 2 siRNA (4390846) as negative control. All reagents were from Thermo Fisher Scientific. Cells were assayed 72 h after transfection, when reaching 90–100% confluence. For overexpression experiments in INS-1 beta cells, the same lentiviral vectors as described for human islets were used. Cells (79,000 cells per well in 48-well plates) were transduced for 24 h with a MOI of 1.25 and all experiments were performed 72 h after initiation of transduc- tion. EndoC-βH1 beta cells were transfected as described above withRHOT1-targeting siRNA (s30650) or the negative control.Insulin secretion in INS-1 beta cells and EndoC-βH1 beta cellsConfluent plates of transfected or transduced INS-1 832/13 beta cells or EndoC-βH1 beta cells, the latter cultured overnight in low-glucose media before experiments, were washed twice carefully with pre-warmed Secretion Assay Buffer (SAB), pH 7.2 (1.16 mM MgSO 4, 4.7 mM KCl, 1.2 mM KH2 PO 4, 114 mM NaCl, 2.5 mM CaCl2 , 25.5 mM NaHCO 3, 20 mM HEPES and 0.2% bovine serum albumin) containing 2.8 mM glucose. Cells were then pre-incubated in fresh SAB with 2.8 mM glucose for 2 h. Next, the cells were stimulated for 1 h in SAB with 2.8 mM glucose, 16.7 mM glucose or 2.8 mM glucose with 35 mM KCl at 37 °C. Insulin levels were measured by ELISA (10-1145-01 (rat) or 10-1113-01 (human), Mercodia) and normalized to total protein or total insulin content in each well. Total protein from each well was extracted by using 100–200 μl RIPA buffer: 0.1% SDS, 150 nM NaCl, 1% Triton X-100, 50 mM Tris-Cl, pH 8 and EDTA-free protease inhibitor (Roche). The protein content was analysed by BCA assay (Thermo Fisher Scientific). Plates were read on a CLARIOstar microplate reader.Real-time qPCR in human islets, INS-1 beta cells, rat islets and EndoC-βH1 beta cellsTotal RNA was extracted with the miRNeasy isolation kit (QIAGEN) according to the manufacturer’s instructions. The equipment and rea- gents mentioned below were purchased from Thermo Fisher Scientific. RNA concentrations were measured using NanoDrop (ND-1000). cDNA was generated with the RevertAid First Strand cDNA synthesis kit. qPCR was performed in triplicates on 384-well plates in a QuantStudio 7 Flex Real Time PCR system under default cycling parameters. TaqMan assays were used to measure expression levels of human ADCY9 (Hs00181599_ m1), human FBLL1 (Hs01584505_s1), ZNF431 (Hs00819968_m1), humanZSWIM9 (Hs00736775_m1), human ONECUT2 (Hs00191477_m1), humanRHOT1 (Hs00430256_m1), human SNX1 (Hs00162052_m1), humanTRAPPC13 (Hs01556088_m1), human GNAL (Hs00181836_m1), rat Adcy9(Rn01424753_m1), rat Fbll1 (Rn06322963_s1), rat Zswim9 (Rn01753793_ m1), rat Onecut2 (Rn01265320_m1), rat Snx1 (Rn01418446_m1), ratTrappc13 (Rn01493919_m1), rat Gnal (Rn01491898_m1) and rat Ins1(Rn02121433_g1). Human PPIA (Hs04194521_s1) and human HPRT1(Hs02800695_m1), as well as rat Ppia (Rn00690933_m1) and Hprt1(Rn01527840_m1), were used as endogenous controls for normaliz- ing mRNA expression. Data are presented as relative quantification, describing the difference in expression of the gene compared with a control group. Threshold levels of all Ct values were automatically set, and the gene expression levels were normalized using the geomet- ric means of the two endogenous controls, PPIA and HPRT1. Relative expressions were calculated with the ΔΔCt method.Western blot analysisTo extract protein, cells were first lysed in ice-cold RIPA buffer (50 mM Tris-HCl pH 7.4–7.6, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.5% sodium deoxycholate and 0.1% SDS) containing protease inhibitor cocktail (Sigma-Aldrich). Lysates were spun at 11,000g for 3 min to pel - let debris and protein concentration of the supernatant was measured ### Page 20 Nature Metabolism | Volume 8 | April 2026 | 957–980 976Resource https://doi.org/10.1038/s42255-026-01498-9by BCA assay (Thermo Fisher Scientific). Next, 10 μg of protein was mixed with 5x sample buffer containing 10% β-mercaptoethanol and boiled before electrophoretic separation on 4–15% TGX Stain-Free gels (Bio-Rad). The gels were then activated with ultraviolet light for 1 min to visualize total protein on the blotted membrane. Protein was transferred to PVDF membrane with a TransBlot Turbo Transfer System (Bio-Rad). Membranes were blocked with 5% milk and 1% BSA in a buffer consisting of 150 mM NaCl, 20 mM Tris-HCl, pH 7.5, and 0.1% Tween for 1 h. The blots were probed with antibodies against Anti-6X His-tag anti- body [EPR20547] - ChIP grade (ab213204, 1000344-30, clone EPR20547; 1:500 dilution, Abcam), ADCY9/AC9 (ab191423, GR269903-5, clone EPR16188; 1:500 dilution, Abcam) or total OXPHOS rodent antibody cocktail (ab110413-MS604, lot 2101021674; 1:250 dilution, Abcam) and incubated overnight at 4 °C. Secondary antibodies were horseradish peroxidase-conjugated goat anti-rabbit (7074, lot 32; 1:10,000 dilution, Cell Signaling Technology) or goat anti-mouse (1706516, 1:2,000 dilu- tion, Bio-Rad). Clarity Western ECL Substrate was used for visualiza - tion of proteins with a ChemiDoc XRS+ System (Bio-Rad). The signal intensity of each protein band was measured with Image Lab software (version 5.2.1; Bio-Rad) and normalized to the total amount of protein loaded in the lane.OCR measurementsMitochondrial OCRs were evaluated with an XFe24 extracellular flux analyser (Agilent). Details of this analysis are described in Supple- mentary File 1.ATP:ADP ratio measurement in live single cells using PercevalHRThe ATP/ADP ratio was measured with the fluorescent biosensor Perce- valHR. A total of 70,000 INS-1 beta cells were seeded on poly-D-lysine (1 mg ml−1 )-coated Lab-Tek chambered cover glass (Thermo Fisher Scientific), followed by transfection with Lipofectamine 3000 (Thermo Fisher Scientific) with 1 μg of PercevalHR plasmid DNA (Addgene ID: 21737) per well together with lentivirus transduction containing either the cDNA of GFP or ONECUT2. Seventy-two hours later, the cells were pre-incubated in experimental buffer (pH 7.4): 3.6 mM KCl, 1.3 mM CaCl2 , 0.5 mM MgSO 4, 0.5 mM Na 2HPO4, 10 mM HEPES, 5 mM NaHCO 3 and 135 mM NaCl, supplemented with 2.8 mM glucose, for 90 min at 37 °C. Cover glass with adhered cells was mounted on the stage of a Nikon ECLIPSE Ti2 microscope (Bergman Labora) equipped with a confocal unit. PercevalHR was excited with laser light at 488 nm and emission was detected at 520 nm at 2.8mM glucose and after addition of 16.7 mM glucose and 5 μM oligomycin sequentially.Gene expression microarray of ADCY9-deficient orONECUT2-overexpressing isletsTotal RNA was extracted from human islets after silencing ADCY9using siADCY9 (n = 5 donors) or ONECUT2 lentivirus overexpression (OE-ONECUT2, n = 6 donors) using the miRNeasy isolation kit (QIAGEN) according to the manufacturer’s recommendations. Details of the analyses of the microarray data are described in Supplementary File 2.Rat modelDiabetic GK rats12 , developed by selective breeding of Wistar rats, and control Wistar rats from Janvier labs in France were kept in standard controlled housing conditions; 21–22 °C, 55–65% humidity and 12-h light–dark cycle, and were given standard chow (SAFE A40) and water ad libitum. Animals were used at 11–13 weeks of age. Only male rats were included to reduce the number of animals used. Blood glucose levels were measured from the tail vein using a glucometer (ACCU-CHEK Aviva, Roche). Pancreatic islets were isolated by collagenase diges- tion and handpicked in cold Hank’s buffer with 1 mg ml −1 bovine serum albumin. Islets (75–200) were washed in ice-cold PBS, lysed in Qiazol (QIAGEN), transferred to microcentrifuge tubes, and homogenized by vortexing tubes for 1 min before storage at −20 °C until RNA extraction using the MiRNeasy Mini Kit protocol (217004, QIAGEN). RNA concen- tration was measured using NanoDrop (ND-1000 Spectrophotometer).Rescue experiment in EndoC-βH1 beta cells exposed to HGPDetails are described in Supplementary File 3.Statistical methods and data collectionClinical characteristics between individuals with pre-T2D/T2D and non-diabetic controls were analysed using two-sample, two-tailedt-tests. DMR calling in alpha versus beta cells was done using dmrseq25 (details are found above) including donors with data available in both cell types. DMR calling in donors with pre-T2D/T2D versus controls in alpha and beta cells was done in a similar way; however, here the analy- ses were adjusted for age, sex, BMI and days in culture of islets before MARIS. DEGs in alpha versus beta cells were analysed in paired samples using DESeq2 (ref. 82). DEGs in donors with pre-T2D/T2D versus con- trols in alpha and beta cells were identified in a similar way, and here the analyses were adjusted for age, sex, BMI and days in culture of islets. FDR analysis was used to account for multiple testing, and FDR below 5% and 10% (q < 0.05 and q < 0.1) were applied. No statistical methods were used to pre-determine sample sizes, but they were based on results from our previous study where we analysed DNA methylation and gene expression in whole human islets 11 . T2D-associated differences in DNA methylation of individual CpGs were analysed using a non-parametric Wilcoxon rank-sum test. Student t-tests (paired and two-tailed) were used in silencing experiments comparing siRNAs of target genes with control (siNC cells), as well as the overexpression experiments in human islets and INS-1 beta cells. Student’s t-tests (unpaired, two-tailed) were used for analyses of GK versus Wistar rats. Paired analysis of variance was used for analyses of the siONECUT2/HGP rescue experiment in EndoC-βH1 beta cells, with post hoc tests based on paired t-tests. Tests were two-tailed unless stated otherwise. To show if overlapped sets display statistically significant enrich- ment, we used gene-set enrichment hypergeometric tests for gene overlaps and region-based permutation tests for genomic region over- laps. Details of these analyses are described in Supplementary File 4. While the non-parametric tests used do not require normal distri- bution, parametric tests were used for normally distributed data. For EPIC v2.0 array data, M values were used for bioinformatic and statis- tical analyses. The collection and analysis of genome-wide data were performed blinded to the conditions of the experiment. Data collection and analysis for the remaining experiments were not performed blind to the conditions of the experiments. No data points were excluded from analysis unless they were technical outliers.Figure preparationGraphs in Figs. 1–8 and Extended Data Figs. 1–8 were created using GraphPad Prism and arranged in CorelDraw version 24.3.05.571. Fig. 1a and Extended Data Fig. 6l were created in BioRender; Ofori, J. https:// biorender.com/h6zumch and https://biorender.com/1w54ey4 (2026). Euler diagrams for Figs. 2k, 5j and 6l were created using eulerr 90 . Fig. 8e was generated using Cytoscape (v.3.10.3).Reporting summaryFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article.Data availabilityThe WGBS and RNA-seq data generated from human alpha and beta cells are deposited in Zenodo (https://doi.org/10.5281/ zenodo.18656853) 91 , and in the LUDC repository (https://www.ludc. lu.se/resources/repository; accession nos. LUDC2025.08.1 for the alpha cell WGBS data, LUDC2025.08.3 for the beta cell WGBS data, LUDC2025.08.4 for the alpha cell RNA-seq data and LUDC2025.08.5 ### Page 21 Nature Metabolism | Volume 8 | April 2026 | 957–980 977Resource https://doi.org/10.1038/s42255-026-01498-9for the beta cell RNA-seq data). Data can be requested through the repository portal and individual-level data from the human pancreatic islets are not publicly available due to ethical and legal restrictions related to the Swedish Biobanks in Medical Care Act, the Personal Data Act and European Union’s General Data Protection Regulation and Data Protection Act. The human alpha and beta cell RNA-seq data used for validation are available in the Gene Expression Omnibus under accession number GSE67543. The gene expression microarray data from ADCY9-deficient or ONECUT2-overexpressing human islets are available in the Gene Expression Omnibus under accession numbers GSE319105 and GSE319109, respectively). alpha-beta-methylome, an open resource based on the WGBS and RNA-seq data included in this study, is available at https://alpha-beta-methylome.serve.scilifelab.se/ app/alpha-beta-methylome/, and allows users to explore and visualize the impact of age, sex and pre-T2D/T2D on autosomal DNA methyla- tion and/or expression of any gene in alpha and beta cells, as well as differences between alpha and beta cells. Source data are provided with this paper.Code availabilityThe computer code used to generate the results described in ‘WGBS and data analysis’ and ‘Statistical methods and data collection’ is deposited in Zenodo (https://doi.org/10.5281/zenodo.18681937)92 . Other details are available from the corresponding author on on request.References1. Wewer Albrechtsen, N. J. et al. The liver–α-cell axis and type 2 diabetes. Endocrine Rev. 40, 1353–1366 (2019). 2. Suzuki, K. et al. Genetic drivers of heterogeneity in type 2 diabetes pathophysiology. Nature https://doi.org/10.1038/s41586-024- 07019-6 (2024). 3. Federation, I. D. IDF Diabetes Atlas, 11th edition https:// diabetesatlas.org/ (2025). 4. Wendt, A. & Eliasson, L. Pancreatic alpha-cells - The unsung heroes in islet function. Semin. Cell Dev. Biol. 103, 41–50 (2020). 5. Ling, C. & Rönn, T. Epigenetics in human obesity and type 2 diabetes. Cell Metab. 29, 1028–1044 (2019). 6. Ling, C., Bacos, K. & Rönn, T. Epigenetics of type 2 diabetes mellitus and weight change — a tool for precision medicine?. Nat. Rev. Endocrinol. 18, 433–448 (2022). 7. Yang, B. T. et al. Insulin promoter DNA methylation correlates negatively with insulin gene expression and positively with HbA1c levels in human pancreatic islets. Diabetologia 54, 360–367 (2011). 8. Yang, B. T. et al. Increased DNA methylation and decreased expression of PDX-1 in pancreatic islets from patients with type 2 diabetes. Mol. Endocrinol. 26, 1203–1212 (2012). 9. Dayeh, T. et al. Genome-wide DNA methylation analysis of human pancreatic islets from type 2 diabetic and non-diabetic donors identifies candidate genes that influence insulin secretion. PLoS Genet. 10, e1004160 (2014). 10. Volkmar, M. et al. DNA methylation profiling identifies epigenetic dysregulation in pancreatic islets from type 2 diabetic patients.EMBO J. 31, 1405–1426 (2012). 11. Volkov, P. et al. Whole-genome bisulfite sequencing of human pancreatic islets reveals novel differentially methylated regions in type 2 diabetes pathogenesis. Diabetes 66, 1074–1085 (2017). 12. Ronn, T. et al. Genes with epigenetic alterations in human pancreatic islets impact mitochondrial function, insulin secretion, and type 2 diabetes. Nat. Commun. 14, 8040 (2023). 13. Rönn, T., Perfilyev, A., Oskolkov, N. & Ling, C. Predicting type 2 diabetes via machine learning integration of multiple omics from human pancreatic islets. Sci. Rep. https://doi.org/10.1038/ s41598-024-64846-3 (2024). 14. Nilsson, E. et al. Altered DNA methylation and differential expression of genes influencing metabolism and inflammation in adipose tissue from subjects with type 2 diabetes. Diabetes 63, 2962–2976 (2014). 15. Nilsson, E. et al. Epigenetic alterations in human liver from subjects with type 2 diabetes in parallel with reduced folate levels. J. Clin. Endocrinol. Metab. 100, E1491–E1501 (2015). 16. Nitert, M. D. et al. Impact of an exercise intervention on DNA methylation in skeletal muscle from first-degree relatives of patients with type 2 diabetes. Diabetes 61, 3322–3332 (2012). 17. Ling, C. et al. Multiomics profiling of DNA methylation, microRNA, and mRNA in skeletal muscle from monozygotic twin pairs discordant for type 2 diabetes identifies dysregulated genes controlling metabolism. BMC Med. https://doi.org/10.1186/ s12916-024-03789-y (2024). 18. Davegårdh, C. et al. VPS39-deficiency observed in type 2 diabetes impairs muscle stem cell differentiation via altered autophagy and epigenetics. Nat. Commun. https://doi.org/10.1038/s41467- 021-22068-5 (2021). 19. Kirchner, H. et al. Altered DNA methylation of glycolytic and lipogenic genes in liver from obese and type 2 diabetic patients.Mol. Metab. 5, 171–183 (2016). 20. Loyfer, N. et al. A DNA methylation atlas of normal human cell types. Nature 613, 355–364 (2023). 21. Wang, G. et al. Integrating genetics with single-cell multiomic measurements across disease states identifies mechanisms of beta cell dysfunction in type 2 diabetes. Nat. Genet. 55, 984–994 (2023). 22. Ngara, M. & Wierup, N. Lessons from single-cell RNA sequencing of human islets. Diabetologia 65, 1241–1250 (2022). 23. Hrvatin, S., Deng, F., O’Donnell, C. W., Gifford, D. K. & Melton, D. A. MARIS: method for analyzing RNA following intracellular sorting.PLoS ONE 9, e89459 (2014). 24. Pasquali, L. et al. Pancreatic islet enhancer clusters enriched in type 2 diabetes risk-associated variants. Nat. Genet. 46, 136–143 (2014). 25. Korthauer, K., Chakraborty, S., Benjamini, Y. & Irizarry, R. A. Detection and accurate false discovery rate control of differentially methylated regions from whole genome bisulfite sequencing. Biostatistics 20, 367–383 (2019). 26. Bacos, K. et al. Type 2 diabetes candidate genes, including PAX5, cause impaired insulin secretion in human pancreatic islets.J. Clin. Invest. https://doi.org/10.1172/jci163612 (2023). 27. Blodgett, D. M. et al. Novel observations from next-generation RNA sequencing of highly purified human adult and fetal islet cell subsets. Diabetes 64, 3172–3181 (2015). 28. Manduchi, E. et al. Epigenetic adaptation of beta cells across lifespan and disease. Nat. Metab. https://doi.org/10.1038/s42255- 026-01495-y (2026). 29. Kaluscha, S. et al. Evidence that direct inhibition of transcription factor binding is the prevailing mode of gene and repeat repression by DNA methylation. Nat. Genet. 54, 1895–1906 (2022). 30. Dayeh, T. A. et al. Identification of CpG-SNPs associated with type 2 diabetes and differential DNA methylation in human pancreatic islets. Diabetologia 56, 1036–1046 (2013). 31. Olsson, A. H. et al. Genome-wide associations between genetic and epigenetic variation influence mRNA expression and insulin secretion in human pancreatic islets. PLoS Genet. 10, e1004735 (2014). 32. Su, C. et al. 3D chromatin maps of the human pancreas reveal lineage-specific regulatory architecture of T2D risk. Cell Metab.34, 1394–1409 (2022). 33. Tagliafierro, L. et al. Lentiviral vector platform for the efficient delivery of epigenome-editing tools into human induced pluripotent stem cell-derived disease models. J. Vis. Exp.https://doi.org/10.3791/59241 (2019). ### Page 22 Nature Metabolism | Volume 8 | April 2026 | 957–980 978Resource https://doi.org/10.1038/s42255-026-01498-934. Jazayeri, M., Eftekhari-Yazdi, P., Sadighi Gilani, M. A., Sharafi, M. & Shahverdi, A. Epigenetic modifications at DMRs of imprinting genes in sperm of type 2 diabetic men. Zygote 30, 638–647 (2022). 35. Ghosal, A. & Said, H. M. Mechanism and regulation of vitamin B2 (riboflavin) uptake by mouse and human pancreatic beta-cells/ islets: physiological and molecular aspects. Am. J. Physiol. Gastrointest. Liver Physiol. 303, G1052–G1058 (2012). 36. Khamis, A. et al. Laser capture microdissection of human pancreatic islets reveals novel eQTLs associated with type 2 diabetes. Mol. Metab. 24, 98–107 (2019). 37. Wu, L. et al. Circ-Tulp4 promotes β-cell adaptation to lipotoxicity by regulating soat1 expression. J. Mol. Endocrinol. 65, 149–161 (2020). 38. Cai, X., Yang, R., Shi, W., Cai, Y. & Ma, Z. Exploration of the common pathogenic link between COVID-19 and diabetic foot ulcers: an in silico approach. Health Sci. Rep. https://doi. org/10.1002/hsr2.1686 (2023). 39. Jabs, N. et al. Reduced insulin secretion and content in VEGF-a deficient mouse pancreatic islets. Exp. Clin. Endocrinol. Diabetes116, S46–S49 (2008). 40. Broadaway, K. A. et al. Loci for insulin processing and secretion provide insight into type 2 diabetes risk. Am. J. Hum. Genet. 110, 284–299 (2023). 41. Arpón, A. et al. Insulin sensitivity is associated with lipoprotein lipase (LPL) and catenin delta 2 (CTNND2) DNA methylation in peripheral white blood cells in non-diabetic young women.Int. J. Mol. Sci. 20, 2928 (2019). 42. Horimoto, A. et al. Admixture mapping of chronic kidney disease and risk factors in Hispanic/Latino individuals from Central America country of origin. Circ. Genom. Precis. Med. https://doi. org/10.1161/CIRCGEN.123.004314 (2024). 43. Lu, D. et al. Abnormal glucose homeostasis and pancreatic islet function in mice with inactivation of the Fem1b gene. Mol. Cell. Biol. 25, 6570–6577 (2005). 44. Bruno, J., Brumfield, A., Chaudhary, N., Iaea, D. & McGraw, T. E. SEC16A is a RAB10 effector required for insulin-stimulated GLUT4 trafficking in adipocytes. J. Cell Biol. 214, 61–76 (2016). 45. Lin, W. H. Molecular scanning of the human sorbin and SH3-domain-containing-1 (SORBS1) gene: positive association of the T228A polymorphism with obesity and type 2 diabetes. Hum. Mol. Genet. 10, 1753–1760 (2001). 46. Walker, J. T. et al. Genetic risk converges on regulatory networks mediating early type 2 diabetes. Nature 624, 621–629 (2023). 47. Martinez-Lopez, J. A. et al. Single-cell mRNA-regulation analysis reveals cell type-specific mechanisms of type 2 diabetes.Nat. Commun. 16, 9475 (2025). 48. Buenaventura, T. et al. A targeted RNAi screen identifies endocytic trafficking factors that control GLP-1 receptor signaling in pancreatic β-cells. Diabetes 67, 385–399 (2018). 49. Plaisance, V. et al. MicroRNA-9 controls the expression of granuphilin/Slp4 and the secretory response of insulin-producing cells. J. Biol. Chem. 281, 26932–26942 (2006). 50. Tian, G., Sol, E. R. M., Xu, Y., Shuai, H. & Tengholm, A. Impaired cAMP generation contributes to defective glucose-stimulated insulin secretion after long-term exposure to palmitate. Diabetes64, 904–915 (2015). 51. Nagaraj, V. et al. Elevated basal insulin secretion in type 2 diabetes caused by reduced plasma membrane cholesterol. Mol. Endocrinol. 30, 1059–1069 (2016). 52. Tantama, M., Martínez-François, J. R., Mongeon, R. & Yellen, G. Imaging energy status in live cells with a fluorescent biosensor of the intracellular ATP-to-ADP ratio. Nat. Commun. https://doi. org/10.1038/ncomms3550 (2013). 53. Nagao, M., Lagerstedt, J. O. & Eliasson, L. Secretory granule exocytosis and its amplification by cAMP in pancreatic beta-cells.Diabetol. Int. 13, 471–479 (2022). 54. Langfelder, P. & Horvath, S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9, 559 (2008). 55. Li, Y. et al. Recent advances in pancreatic alpha-cell transdifferentiation for diabetes therapy. Front. Immunol. https:// doi.org/10.3389/fimmu.2025.1551372 (2025). 56. Brooks, E. P. et al. NKX2.2 and KLF4 cooperate to regulate alpha-cell identity. Genes Dev. 39, 242–260 (2025). 57. Akerman, I. et al. Human pancreatic beta cell lncRNAs control cell-specific regulatory networks. Cell Metab. 25, 400–411 (2017). 58. Singer, R. A. et al. The long noncoding RNA Paupar modulates PAX6 regulatory activities to promote alpha cell development and function. Cell Metab. 30, 1091–1106 (2019). 59. Bysani, M. et al. ATAC-seq reveals alterations in open chromatin in pancreatic islets from subjects with type 2 diabetes. Sci. Rep. 9, 7785 (2019). 60. Drucker, D. J. & Holst, J. J. The expanding incretin universe: from basic biology to clinical translation. Diabetologia 66, 1765–1779 (2023). 61. Koeck, T. et al. A common variant in TFB1M is associated with reduced insulin secretion and increased future risk of type 2 diabetes. Cell Metab. 13, 80–91 (2011). 62. Haythorne, E. et al. Diabetes causes marked inhibition of mitochondrial metabolism in pancreatic beta cells. Nat. Commun.10, 2474 (2019). 63. Abu Aqel, Y. et al. Glucokinase (GCK) in diabetes: from molecular mechanisms to disease pathogenesis. Cell Mol. Biol. Lett. 29, 120 (2024). 64. Roggli, E. et al. Changes in microRNA expression contribute to pancreatic beta-cell dysfunction in prediabetic NOD mice.Diabetes 61, 1742–1751 (2012). 65. Dirice, E. et al. Single-nucleus RNA-seq reveals singular gene signatures of human ductal cells during adaptation to insulin resistance. JCI Insight https://doi.org/10.1172/jci.insight.153877 (2022). 66. Rorsman, P. & Ashcroft, F. M. Pancreatic beta-cell electrical activity and insulin secretion: of mice and men. Physiol. Rev. 98, 117–214 (2018). 67. Engstrom, K. et al. Evaluating cell-specific gene expression using single-cell and single-nuclei RNA-sequencing data from human pancreatic islets of the same donors. Sci. Rep. 15, 36133, https://doi.org/10.1038/s41598-025-21595-1 (2025). 68. Vanhorenbeeck, V. et al. Role of the Onecut transcription factors in pancreas morphogenesis and in pancreatic and enteric endocrine differentiation. Dev. Biol. 305, 685–694 (2007). 69. Philippi, A. et al. Mutations and variants of ONECUT1 in diabetes.Nat Med. 27, 1928–1940 (2021). 70. Russ-Silsby, J. et al. The role of ONECUT1 variants in monogenic and type 2 diabetes mellitus. Diabetes 72, 1729–1734 (2023). 71. Kaplan, S. J. et al. CRISPR screening uncovers a long-range enhancer for ONECUT1 in pancreatic differentiation and links a diabetes risk variant. Cell Rep. https://doi.org/10.1016/j. celrep.2024.114640 (2024). 72. Bacos, K. et al. Blood-based biomarkers of age-associated epigenetic changes in human islets associate with insulin secretion and diabetes. Nat. Commun. https://doi.org/10.1038/ ncomms11089 (2016). 73. Hall, E. et al. Sex differences in the genome-wide DNA methylation pattern and impact on gene expression, microRNA levels and insulin secretion in human pancreatic islets. Genome Biol. 15, 522 (2014). ### Page 23 Nature Metabolism | Volume 8 | April 2026 | 957–980 979Resource https://doi.org/10.1038/s42255-026-01498-974. Guillam, M. T. et al. Early diabetes and abnormal postnatal pancreatic islet development in mice lacking Glut-2. Nat. Genet.17, 327–330 (1997). 75. McCulloch, L. J. et al. GLUT2 (SLC2A2) is not the principal glucose transporter in human pancreatic beta cells: implications for understanding genetic association signals at this locus. Mol. Genet. Metab. 104, 648–653 (2011). 76. Sansbury, F. H. et al. SLC2A2 mutations can cause neonatal diabetes, suggesting GLUT2 may have a role in human insulin secretion. Diabetologia 55, 2381–2385 (2012). 77. Schrader, S. et al. Novel subgroups of type 2 diabetes display different epigenetic patterns that associate with future diabetic complications. Diabetes Care 45, 1621–1630 (2022). 78. Patro, R., Duggal, G., Love, M. I., Irizarry, R. A. & Kingsford, C. Salmon provides fast and bias-aware quantification of transcript expression. Nat. Methods 14, 417–419 (2017). 79. Ewels, P., Magnusson, M., Lundin, S. & Käller, M. MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics 32, 3047–3048 (2016). 80. R Core Team R: A Language and Environment for Statistical Computing (R Core Team, 2022). 81. Soneson, C., Love, M. I. & Robinson, M. D. Differential analyses for RNA-seq: transcript-level estimates improve gene-level inferences. F1000Res. 4, 1521 (2016). 82. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. https://doi.org/10.1186/s13059-014-0550-8 (2014). 83. Raine, A., Manlig, E., Wahlberg, P., Syvänen, A. -C. & Nordlund, J. SPlinted Ligation Adapter Tagging (SPLAT), a novel library preparation method for whole genome bisulphite sequencing.Nucleic Acids Res. 45, e36 (2017). 84. Krueger, F. Trim Galore! Babraham http://www.bioinformatics. babraham.ac.uk/projects/trim_galore/ (2019). 85. Krueger, F. & Andrews, S. R. Bismark: a flexible aligner and methylation caller for Bisulfite-Seq applications. Bioinformatics27, 1571–1572 (2011). 86. Cavalcante, R. G. & Sartor, M. A. annotatr: Genomic regions in context. Bioinformatics 33, 2381–2383 (2017). 87. Peters, T. J. et al. Characterisation and reproducibility of the HumanMethylationEPIC v2.0 BeadChip for DNA methylation profiling. BMC Genomics 25, 251 (2024). 88. Aryee, M. J. et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics 30, 1363–1369 (2014). 89. Pidsley, R. et al. A data-driven approach to preprocessing Illumina 450K methylation array data. BMC Genomics 14, 293 (2013). 90. Larsson, J. eulerr: Area-proportional Euler and Venn diagrams with ellipses. R package version 6.0.0. CRAN https://cran.r-project.org/ package=eulerr (2019). 91. Ofori, J. K. et al. Cell-specific DNA methylation in human alpha and beta cells regulates gene expression in type 2 diabetes.Zenodo https://doi.org/10.5281/zenodo.18656854 (2026). 92. Lindström, A. C-LingLab/WGBS_sorted_islets. Zenodo https://doi. org/10.5281/zenodo.18681937 (2026).AcknowledgementsWe thank the Nordic Network for Clinical Islet Transplantation (JDRF award 31-2008-413) for providing material for the study; the Human Tissue Laboratory at LUDC; T. Dayeh for experimental support; O. Korsgren, Uppsala University for support with the islet isolation infrastructure; C. Wollheim, University of Geneva for valuable discussions; Lund University Cell and Gene Therapy core facility for the support in virus production; and the Bioinformatics and Expression Analysis (supported by the Board of Research at the Karolinska Institute and the Research Committee at Karolinska Hospital) for microarray analysis. WGBS was performed by the SNP&SEQ Technology Platform in Uppsala, which is part of the National Genomics Infrastructure Sweden and Science for Life Laboratory, supported by the Swedish Research Council and the Knut and Alice Wallenberg Foundation. Research grants supporting the study were obtained from the Kungliga Fysiografiska Sällskapet i Lund (the Birgit and Hellmuth Hertz Foundation, to J.K.O.), Novo Nordisk foundation (to C.L.), the Swedish Foundation for Strategic Research (IRC15-0067, to C.L. and L.E.), the Swedish Research Council (2016-02486, 2018-02567, 2019-01406, 2014-2775, 2018- 02435, 2021-00628 and 2024-03597, to C.L.), the Region Skåne (Avtal om Läkarutbildning och Forskning (ALF), to C.L.), Strategic Research Area Exodiab (2009-1039, to C.L. and L.E.), the European Research Council (ERC-Paintbox, to C.L.), the Diabetes Foundation (to C.L. and L.E.), as well as Syskonen Svensson, Magnus Bergvall, Hjelt (to S.R.), Åke Wiberg, Påhlsson Foundations, Sigurd och Elsa Goljes memory Foundation (to S.R.) and Lars Hierta’s memory Foundation (to S.R.). The flow cytometry study was performed at the LUDC-Flow Cytometry Core Facility, supported by the Swedish Research Council, Strategic Research Area Exodiab (2009-1039), by the Swedish Foundation for Strategic Research (IRC15-0067), by Region Skåne/(ALF) and by the Infrastructure Grant of Lund University, 2018. This work was also supported by the SciLifeLab & Wallenberg Data Driven Life Science Program (KAW2024.0159, to C.L.).Author contributionsC.L. initiated the project. J.K.O., S.R., A.L., A.P., M.M., A.K., L.S., K.K., Å.N., M.D.-P., A.W., K.B. and C.L. performed experiments and/ or analyses. J.J., T.R., B.K., M.R. and L.E. designed experiments and experimental tools, and interpreted data. J.K.O., K.B. and C.L. drafted the manuscript. J.K.O. performed functional work in human islets, rat islets and INS-1 beta cells. S.R. performed all epigenetic editing experiments and contributed to writing the manuscript. A.L. developed the web tool and pipeline for EPIC v2.0 data and performed bioinformatic analyses. A.P. developed bioinformatic pipelines and performed bioinformatic analyses. All authors designed some experiments and/or analyses of data. All authors read and edited the manuscript.FundingOpen access funding provided by Lund University.Competing interestsThe authors declare no competing interests.Additional informationExtended data is available for this paper at https://doi.org/10.1038/s42255-026-01498-9.Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s42255-026-01498-9.Correspondence and requests for materials should be addressed to Charlotte Ling.Peer review information Nature Metabolism thanks Maria Golson, Guoqiang Gu and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Revati Dewal, in collaboration with the Nature Metabolism team. Peer reviewer reports are available.Reprints and permissions information is available at www.nature.com/reprints. ### Page 24 Nature Metabolism | Volume 8 | April 2026 | 957–980 980Resource https://doi.org/10.1038/s42255-026-01498-9Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/. © The Author(s) 2026 1 Epigenetics and Diabetes Unit, Department of Clinical Sciences in Malmö, Lund University, Scania University Hospital, Malmö, Sweden. 2 SciLifeLab, Lund University, Lund, Sweden. 3 Center for Evolutionary Hologenomics, Globe Institute, University of Copenhagen, Copenhagen, Denmark. 4 Exodiab, Lund University Diabetes Centre, Lund University, Malmö, Sweden. 5 Department of Advanced Medical and Surgical Sciences, University of Campania ‘Luigi Vanvitelli’, Naples, Italy. 6 Institute of Pediatric Rare Diseases, Florida State University, Tallahassee, FL, USA. 7 University of Groningen, University Medical Center Groningen, Groningen, the Netherlands. 8 Islet Cell Exocytosis Unit, Department of Clinical Sciences in Malmö, Lund University Diabetes Centre, Lund University, Malmö, Sweden. 9 These authors contributed equally: Jones K. Ofori, Sabrina Ruhrmann. 10 These authors jointly supervised this work: Karl Bacos, Charlotte Ling. e-mail: charlotte.ling@med.lu.se ### Page 25 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 1 | See next page for caption. ### Page 26 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 1 | Characterization of the DNA methylome in sorted human α- and β-cells data related to Figs. 1 and 2. a, Gating strategy used in sorting α- and β-cells from human islets after staining with antibodies against glucagon and insulin. b, CpG-sites are mapped to Gene regions based on functional genome distribution, where the promoter is defined as the region between the transcription start site (TSS) and 1 kb upstream, and to CpG island regions based on CpG content. A CpG island is defined as a ≥ 200 bp stretch of DNA with a CG content of ≥50% and an observed CpG/expected CpG in excess of 0.6. Shores are 2,000 bp regions flanking CpG islands. Shelves are 2,000 bp regions outside the shores. The Open Sea is everything outside the shelves. c-h, Average DNA methylation levels in α- and β-cells in different gene regions; 1-5 kb from TSS, p = 8.36E-07 (low), p = 3.23E-04 (not) (c ), Promoters, p = 2.72E-04 (medium), p = 9.73E-09 (low), p = 0.0036 (not) (d ), 5’untranslated regions (UTRs), p = 3.83E-04 (medium), p = 3.41E-05 (low), p = 0.040 (not) (e ), Exons, p = 5.82E-05 (high), p = 0.008 (medium), p = 4.20E-06 (low), p = 1.11E- 08 (not) (f), Introns, p = 3.83E-05 (high), p = 2.10E-05 (low), p = 2.41E-06 (not) (g), and 3’UTRs, p = 2.17E-04 (high), p = 1.25E-04 (low), p = 1.08E-07 (not) (h ) of not expressed genes and expressed genes divided into low-, medium-, and high-expressed in α- and β-cells (n = 7 (α-cells) and 11 (β-cells) donors), *p < 0.05, **p < 0.01, ***p < 0.001 based on two-tailed Student’s t-tests. Data are presented as mean ± SEM. i, Coverage for the whole genome bisulfite sequencing (WGBS) data used for analysis of DMRs between sorted α- and β-cells from the same human islets (n = 7 donors). j, Manhattan plot displaying the 22,544 DMRs between α- and β-cells at a false discovery rate (FDR) less than 5% (q < 0.05 based on dmrseq25 ). k, Proportion of identified DMRs annotated to different gene and CpG island regions. l, m, Left; α- versus β-cell DMRs annotated to GATA6 (l) andSIX2 (m) in sorted α- and β-cell (n = 7 donors, dmrseq 25 ). Right; expression of respective gene in sorted α- versus β-cells (n = 16, ***q < 0.001, based on DESeq2 82 , presented as mean ± SEM, see Supplementary Table 6 for exact P values). Women and men are represented by filled and open circles, respectively. n, The top X-chromosome DMR (only males, n = 5) annotated to GPR119 (left) and GPR119expression (right) in sorted α- and β-cell (n = 10 males, ***q = 2.93E-35 based on DESeq282 , presented as mean ± SEM). The DMRs are analyzed and visualized by dmrseq25 . In DMR plots presented in l-n, locations of the methylation loci are shown with black lines at the bottom, and average, smoothed lines based on the dmrseq analysis are shown in pink (β-cells) or blue (α-cells), including standard deviation. The shaded area shows the DMR. Annotations to CpG island regions and genes are shown below each graph. The p-/q-values for j and l-n were corrected for multiple testing. ### Page 27 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 2 | α- versus β-cell DMRs in POU6F2 data related to Fig. 2. a-p, 16 significant DMRs annotated to POU6F2 identified in sorted α- versus β-cells (n = 7 donors) at a false discovery rate (FDR) less than 5% (q < 0.05, corrected for multiple testing). In all the DMR plots: locations of methylation loci are shown with black lines at the bottom and average, smoothed lines are shown in pink (β-cells) or blue (α-cells), including standard deviation. The shaded area shows the DMR. Annotations to CpG island regions and exons, if applicable, are shown below each graph. The DMRs were analyzed and visualized by dmrseq 25 . ### Page 28 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 3 | α- versus β-cell DMRs in PDX1/PLUT, data related to Fig. 2. a-j, Significant DMRs annotated to PDX1/PLUT identified in sorted α- versus β-cells (n = 7 donors) at a false discovery rate (FDR) less than 5% (q < 0.05, corrected for multiple testing). In all the DMR plots: locations of methylation loci are shown with black lines at the bottom and average, smoothed lines are shown in pink (β-cells) or blue (α-cells), including standard deviation. The shaded area shows the DMR. Annotations to CpG island regions are shown below each graph. The DMRs were analyzed and visualized by dmrseq 25 . ### Page 29 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 4 | DMRs, TF binding and DNA methylation of individual CpG sites in α- and β-cells, data related to Figs. 2 and 5. a, The longest α- versus β-cell DMR with most CpG sites is annotated to MAFA/ZC3H3. b, c, Some of the most significant α- versus β-cell DMRs on the X chromosome in males (n = 5 donors) are annotated to USP27X (b) and PRKX (c). Expression of respective gene in sorted α- and β-cells is shown on the right (n = 10 donors, data presented as mean ± SEM, ***q < 0.001, based on DESeq2 82 ). In DMR plots presented ina-c, locations of the methylation loci are shown with black lines at the bottom, and average, smoothed lines based on the dmrseq analysis are shown in pink (β-cells) or blue (α-cells), including standard deviation. The shaded area shows the DMR. Annotations to CpG island regions and genes are shown below each graph. The p-/q-values for a-c were corrected for multiple testing. d, A Venn diagram and table showing DMRs in α- versus β-cells on the X chromosome in males overlapping with binding sites for five islet specific transcription factors: FOXA2, MAFB, NKX2.2, NKX6.1 and PDX1 (ref. 24). Overlaps are enriched with p ≤ 0.001 based on region-based permutation tests. For individual p-values, see Supplementary Table 4D. e-g, pre-T2D/T2D-associated differences in the absolute DNA methylation level (%) of individual CpG sites annotated to INS (e), PDX1/PLUT (f), and MAFB (g) in α-cells. *p < 0.05, **p < 0.01 based on two-tailed Wilcoxon rank-sum test in pre-T2D/T2D (n = 3) versus non-diabetic controls (n = 7), and data are presented as mean ± SEM. For exact p-values, see Source data file for Extended Data Fig. 4. Women and men are represented by filled and open circles, respectively. ### Page 30 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 5 | See next page for caption. ### Page 31 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 5 | Epigenetic regulation of T2D candidates, data related to Fig. 7. a, Venn diagram showing T2D-associated differentially expressed genes (DEGs) in human α- and β-cells overlapping with DEGs identified in whole human islets by Bacos et al. 26 b, c, Examples of T2D-associated DEGs found in β-cells (n = 15 non-diabetic donors and 7 donors with T2D) that overlap with T2D-associated DEGs found in human islets by Bacos et al. 26 : GABRA2, SOCS1,SYT1, PAX5, CABLES1 and TBC1D4 (*p < 0.05, **p < 0.01 based on DESeq2 82 ).d-j pre-T2D/T2D-associated differences in the absolute DNA methylation level (%) of individual CpG sites annotated to ADCY9 (d), ZNF431 (e), ZSWIM9 (f), GNAL(g), ONECUT2 (h), SNX1 (i), and TRAPPC13 (j), respectively in β-cells. *p < 0.05, **p < 0.01 pre-T2D/T2D (n = 3 donors) versus non-diabetic controls (n = 11 donors) based on two-tailed Wilcoxon rank-sum test. k, GNAL, NTN1, and PRKCH are T2D-associated DEGs in both sorted α- and β-cells, as well as in whole islets from Bacos et al. 26 . Data shown is for β-cells from pre-T2D/T2D (n = 7 donors) versus non-diabetic controls (n = 15 donors), *p < 0.05, ***p < 0.001. In b-k, women and men are represented by filled and open circles, respectively. Data are presented as mean ± SEM. For exact p-values, see Supp. Table 10 (panels b, c, and k) and Source data files (panels d-j). ### Page 32 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 6 | See next page for caption. ### Page 33 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 6 | Selection of T2D candidates, data related to Figs. 7 and 8. a, Candidate gene selection for functional follow-up from RNA-seq data in sorted α- and β-cells from donors with pre-T2D/T2D versus controls. Flowchart shows the differentially expressed genes (DEGs) in pre-T2D/T2D versus non-diabetic controls in α- and β-cells, respectively, and selection of eight DEGs in β-cells for follow-up. b, Gene manipulation in EndoC- βH1 β-cells often gives different phenotypes compared to human islets and INS-1 β-cells. Figure shows that RHOT1 knockdown (left, ***p = 0.0002) resulted in increased glucose-stimulated insulin secretion (right, **p = 0.0041) in EndoC-βH1 β-cells, which should be compared to reduced insulin secretion seen in human islets and INS-1 β-cells treated similarly (Figs. 4f and 5b in previous publication 12 ). n = 6 biological replicates and paired t-tests. c, Reduced ADCY9 protein levels in INS-1 β-cells after siRNA transfection, n = 4 biological replicates, *p = 0.037 versus negative control (siNC) based on one-tailed paired Student’s t-test.d, Increased ONECUT2 protein levels in INS-1 β-cells after lentiviral transduction, n = 6 biological replicates, **p = 0.0079 versus GFP (control) based on paired Student’s t-test. e, No significant effect on oxygen consumption rate in siAdcy9 INS-1 β-cells, versus siNC, n = 4 biological replicates, p > 0.05 based on paired Student’s t-test. f, Reduced ATP5A protein levels in OE-ONECUT2 INS-1 β-cells, n = 6 biological replicates, *p = 0.049 versus GFP based on paired Student’s t-test.g, Microarray showed decreased expression of NDUFA1 (**p = 0.0022), NDUFA7(**p = 0.0022), and COX10 (**p = 0.0043) in OE-ONECUT2 islets (n = 6 biological replicates/donors) based on Mann-Whitney U-tests, presenting unlogged data. h, Expression of NRID1 (**p = 0.0022), PTPRN (**p = 0.0022), and SLC8B1(**p = 0.0087), contributing to the enrichment of the down-regulated gene set ‘regulation of insulin secretion’ in GSEA of OE-ONECUT2 human islets versus GFP (Mann-Whitney U-tests, presenting unlogged data). n = 6 biological replicates/ donors. i, emapplot visualizing GSEA results of transcriptome data in human islets transfected with siADCY9 versus siNC. n = 5 biological replicates and GSEA was corrected for multiple testing. For full data, see Supplementary Table 11. j, ONECUT2 expression, and k, insulin secretion, presented as fold change (secretion at 16.7 mM glucose divided by secretion at 2.8 mM glucose), when silencing ONECUT2 using siRNA (siONECUT2) versus siNC in EndoC-βH1 β-cells exposed to high glucose (19 mM glucose) plus palmitate (1 mM palmitate, HGP) or control media for 72 h. n = 6 biological replicates, data analyzed by one-way ANOVA (j, p = 0.0293, and k, p = 0.0108) with paired Student’s t-tests as post hoc.j, *p = 0.023: control+siNC versus HGP+siONECUT2, ***p = 4.38E-05: HGP+siNC versus HGP+siONECUT2, and k, *p = 0.017: control+siNC versus HGP+siNC, *p = 0.049: HGP+siNC versus HGP+siONECUT2. n.s.: Not significant. l, Cartoon summarizing key results from this study; created in BioRender; Ofori, J. https:// BioRender.com/1w54ey4 (2026). Data in b-h and i-j are presented as mean ± SEM. Tests were two-tailed. ### Page 34 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 7 | Heatmaps of the WGBS samples, data related to Figs. 2, 5 and 6. Heatmaps showing correlations between all a, α-cell samples (n = 10 donors),b, β-cell samples (n = 14 donors), c, α- and β-cells samples from donors with T2D (n = 6 donors), and d, α- and β-cell samples from the non-diabetic controls (n = 18 donors) for the WGBS data. ### Page 35 Nature MetabolismResource https://doi.org/10.1038/s42255-026-01498-9Extended Data Fig. 8 | Dendrogram of the WGBS samples, data related to Figs. 2, 5 and 6. Dendrogram including all α- (n = 10 donors) and β-cell (n = 14 donors) samples from the WGBS. ### Page 36 ### Page 37 α βα βα βαβ ### Page 38 ### Page 39 β ββ ββββ β ### Page 40 α βα β