Genes with epigenetic alterations in human pancreatic islets impact mitochondrial function, insulin secretion, and type 2 diabetes

Epigenetic dysregulation may influence disease progression. Here we explore whether epigenetic alterations in human pancreatic islets impact insulin secretion and type 2 diabetes (T2D). In islets, 5,584 DNA methylation sites exhibit alterations in T2D cases versus controls and are associated with HbA1c in individuals not diagnosed with T2D. T2D-associated methylation changes are found in enhancers and regions bound by β-cell-specific transcription factors and associated with reduced expression of e.g. CABLES1, FOXP1, GABRA2, GLR1A, RHOT1, and TBC1D4. We find RHOT1 (MIRO1) to be a key regulator of insulin secretion in human islets. Rhot1-deficiency in β-cells leads to reduced insulin secretion, ATP/ADP ratio, mitochondrial mass, Ca2+, and respiration. Regulators of mitochondrial dynamics and metabolites, including L-proline, glycine, GABA, and carnitines, are altered in Rhot1-deficient β-cells. Islets from diabetic GK rats present Rhot1-deficiency. Finally, RHOT1methylation in blood is associated with future T2D. Together, individuals with T2D exhibit epigenetic alterations linked to mitochondrial dysfunction in pancreatic islets.

with the decrease of insulin secretion.Therefore, what is the relationship between Rhot1 expression and insulin secretion in GK rats.It would be helpful to analyze the correlation between Rhot1 expression and insulin secretion.4. The authors found the important role of RHOT1 in maintaining mitochondrial function, but there was no data related to the mitochondrial function in GK rats.It would be better to provide the related results to analyze the relationship between Rhot1 expression and mitochondrial function in GK rats. 5.In Figure 3, RHOT1 expression was decreased by all four siRNA in human islets, such as FOXP1, TBC1D4, RHOT1 and CABLES1.Therefore, how does FOXP1, TBC1D4 and CABLES1 regulate the expression of RHOT1.6.In figure 4c and 5g, there was no GAPDH or ACTIN.How does the authors perform the quantification in WB analysis.Please explain that in detail.7.In some Figure legends, the sample size n was missed.
Reviewer #3 (Remarks to the Author): The manuscript presents a genome-wide DNA methylation study comparing healthy (75 samples) and T2D (25 samples) human islets.The authors also examine the correlation between differential DNA methylation and HbA1C using islets from an independent cohort (114 samples).They integrate the methylation data with RNA-seq, ATAC-seq, transcription factor and histone modifications' ChIP-seq data.They reported four T2D-associated methylation sites associated with lower risk of T2D in blood samples.Selected candidate genes are validated using siRNA knockdown in a rat beta cell line.The study highlights the potential role of Rhot1 in regulating mitochondria function and metabolites, as well as its deficiency in a rodent diabetes model associated with reduced insulin content.Although this study has multiple points, the logical flow between them is not coherent or compelling.Although the authors hint a mechanistic connection between DNA methylation and gene expression, the evidence is preliminary.A lot of results are only described in the text with a brief reference to an excel file without proper presentation or quantitative analysis.A key result about the prospective T2D risk analysis not adequately presented or discussed.Overall, I feel that this manuscript has major flaws and at least needs a lot of improvements to be acceptable for publication in NC.Here are some specific points.
1.I highly recommend including a panel in figure 1 (such as a heatmap) showing the changes the DNA methylation changes in every individual.This will help readers understand what kind of changes we are looking at, and how consistent the changes are between the control and T2D groups.Without a global picture, the readers may feel that the authors are only cherry-picking some best examples.1e is missing.The significance of the enrichment is not clear.

The p-values for the enrichment in Figure
3. The authors claimed some consistency between current study and previous studies (Dayeh et al.; Volkov et al.) but provide very little details.Only an excel file is provided.Summarization of the comparison should be included.For example, how many hyper-methylated sites from this study are hyper or hypo-methylated in the other studies; conversely, how many hypo-methylated sites from this study are hyper or hypo-methylated in the other studies.Quantitative statistical analysis should be also included.
4. The changes of DNA methylation shown in Fig. 2 examples all very subtle, making me wonder if these changes are meaningful.Figures 2a-2d and S2a-2c would benefit from being shown in a genome browser with larger range, with the information of other CpG sites and the results of nearby genes.5.It is unclear whether the differential methylation sites are only enriched in the DEG genes and how significant this is.
6.The section "DNA methylation associates with future T2D" is extremely brief.Again, the authors only refer the readers to an excel file.Here the authors test the putative CpG sites in blood samples.Should we worry about the consistency between islets and blood?This should be at least discussed.In fact, the authors could have screened for candidate predictive DNA methylation sites from blood samples in the beginning.Wouldn't that give better outcomes?The authors names 4 predictive sites, including one with RHOT1, with low T2D risk and low DNA methylation.But in Fig. 1, the RHOT1 CpG show low DNA methylation among T2D patients (i.e., low DNA methylation is associated with increased T2D risk).This is contradictory.7. Fig. 5g-k shows that in diabetic rat, the down regulation of Rhot1 is correlated with lower insulin content in the islets, suggesting a causal role.However, in human islets (Fig. S4a), siRHOT1 does not affect islet insulin content.This raise a serious question if and how much of the Rhot1 functional results from rat are conserved in human islets.
"Epigenetic dysregulation can lead to disease, and there is evidence of associations between the epigenome and T2D 2-3 .Smaller cohorts have shown that DNA methylation is different in human pancreatic islets from T2D cases versus non-diabetic controls 4-6 .Functional experiments in cultured βcells of identified candidate genes showing both differential DNA methylation and expression in islets from donors with T2D (e.g., CDKN1A, PDE7B, PARK2 and SOCS2) further linked epigenetic dysregulation in pancreatic islets to impaired insulin secretion 4,6 .Moreover, increased islet DNA methylation of T2D candidate genes, such as INS, PDX1, GLP1R and PPARGC1A, has also been associated with reduced expression of said genes and abrogated insulin secretion, further supporting a key role of epigenetic dysregulation in pancreatic islets from individuals with diabetes 7-10 .However, though these case-control studies identified differential DNA methylation of numerous genes in pancreatic islets from donors with T2D, they did not establish whether the identified epigenetic alterations predispose an individual to disease.Therefore, studies testing whether epigenetics predispose to T2D and whether epigenetic alterations can be found already in islets from individuals at risk for T2D are desirable.Such studies could provide support that epigenetic alterations in pancreatic islets may cause diabetes rather than being a consequence of the disease.In addition, if epigenetic alterations predispose an individual to diabetes, this information can be used for preventive care, delaying disease onset and progression, and reducing long-term complications and patient suffering.Moreover, larger case-control studies that robustly link epigenetics to T2D are warranted." 2. The star of significance should be larger in: -In Figures 2a (only in FOXP1 RNA-Seq panel), b, c, and d; -in Figures 3g and h Response: Thanks for this detailed scrutiny.Stars of significance have been changed according to the recommendations (now re-numbered as Fig. 3a-d, 4g-h).Please find new versions of all the figures being submitted.

figure 6 should be improved because of not very interpretive.
Response: Based on this valid comment, Fig. 6 (now Fig. 7) has been improved.We included this figure to summarize the main results and take-home message of this extensive study.As such, it is important that the figure is easy to understand.To improve this figure in accordance with this review comment, we have now changed the layout and structure, and also added a section where we summarize in a cartoon how we, based on our studies in β-cells and previous knowledge based on other cell types, view the normal function of RHOT1 in β-cells and how RHOT1 deficiency may lead to mitochondrial dysfunction and subsequently reduced insulin secretion.Please see Fig. 7 (previous Fig. 6) in the resubmitted version.Also, the Figure legend has been re-structured and made more informative to make the figure easier to interpret: page 30, line 1169-1185: "Figure 7. Study summary.a) An explorative epigenome-wide study of ~850,000 DNA methylation sites discovered 5,584 sites in human islets whose methylation associated with both T2D in the Islet T2D case-control cohort (25 cases, 75 controls) and HbA1c in the Islet HbA1c cohort (114 donors not previously diagnosed with T2D).Differential DNA methylation was found in regions of importance for gene regulation, and in genes differentially expressed in islets from donors with T2D vs. non-diabetic controls.b) The candidate gene evaluation found that methylation of some sites predicts T2D development when analyzed in blood (n=450), and that RHOT1 deficiency in human islets reduces insulin secretion and alters the expression of genes related to mitochondrial function.c) Metabolic profiling using β-cells silenced for Rhot1 confirmed reduced insulin secretion and discovered a reduced ATP/ADP ratio, mitochondrial mass, Ca 2+ and respiration, as well as altered metabolite levels.A rodent T2D model, the GK rat, exhibited RHOT1 deficiency.d) Based on our results, we suggest that RHOT1 is needed for proper mitochondrial movement in β-cells, creating a balance in mitochondrial dynamics to maintain functional mitochondria.Left: Functional mitochondria with balanced fission, fusion, and mitochondrial movement.Right: Impaired RHOT1 expression may affect mitochondrial movement with consequences for mitochondrial dynamics (unbalanced fusion/fission processes and altered mitophagy rate).A reduced number of functional mitochondria may impact ATP production and ultimately impair GSIS." Reviewer #2 (Remarks to the Author): The manuscript by Rönn et al. found that alteration of methylation in human islets associated with HbA1c in individuals not diagnosed with T2D.Importantly, RHOT1, as a key regulator of insulin secretion in human islets, was related to insulin secretion, ATP/ADP ratio, mitochondrial mass, Ca 2+ and respiration.Meanwhile, RHOT1-deficiency also regulated mitochondrial dynamics and metabolites, including L-proline, glycine, GABA and carnitines.Notably, RHOT1 methylation in blood associates with future T2D.Overall, although the results of this study are interesting, but several important concerns need to be carefully addressed.

Response:
We are grateful for the thorough revision by Reviewer #2 that has helped us improve our manuscript and we are very pleased that also this reviewer believes our study is interesting.

Point-by-point response to specific comments:
1.In the manuscript, the authors found that lower RHOT1 levels in β-cells led to deregulation of mitophagy and altered expression of proteins involved in mitochondrial dynamics and metabolism, resulting in mitochondrial dysfunction and perturbed GSIS.What are the possible mechanisms by which RHOT1 maintains mitochondrial homeostasis?The authors should discuss this issue in the discussion section.

Response:
We appreciate this comment.The current knowledge of RHOT1 in pancreatic islets and βcells is limited.Hence, we performed this extensive study of mitochondrial function and metabolomics in β-cells silenced for Rhot1.Based on studies in neurons, Rhot1 (also called Miro1) is attached to the mitochondrial surface, where it connects to the motor proteins kinesin and dynein, which in turn connects to the axon filaments.Thus, Rhot1 is part of a motor/adaptor complex connecting the mitochondria to the axon filaments, and thereby important for mitochondrial movement (ref Mitochondrial Trafficking in Neurons, 2013, Schwarz).Our results from β-cells in combination with this knowledge, suggest that RHOT1 is needed for proper mitochondrial movement, thus creating a balance between fusion, fission and mitophagy, to maintain functional mitochondria.With impaired/reduced RHOT1 expression, mitochondrial movement is affected, with consequences for mitochondrial dynamics (unbalanced fusion/fission processes and altered mitophagy rate), which impact the number of functional mitochondria, resulting in reduced ATP production and ultimately lower glucose-stimulated insulin secretion.
This has now been included in the Discussion, page 12-13, line 467-473: "Studies in neurons have shown that RHOT1 connects the mitochondria via motor proteins to the axon filaments, thereby regulating mitochondrial movement (Schwarz 2013).Therefore, it is likely that RHOT1 is needed for proper mitochondrial movement also in human islets and β-cells, creating a balance in mitochondrial dynamics and maintaining functional mitochondria, as required to maintain the proper ATP production needed for GSIS.Thus, impaired RHOT1 expression alters mitochondrial dynamics, which impacts the number of functional mitochondria, resulting in reduced ATP production and insulin secretion (Fig. 7d)." We have also included a new Fig. 7 with a schematic drawing of RHOT1 function in normal and diseased β-cells (Fig. 7d).Page 30, line 1179-1185, Fig. 7d legend: "Based on our results, we suggest that RHOT1 is needed for proper mitochondrial movement in β-cells, creating a balance in mitochondrial dynamics to maintain functional mitochondria.Left: Functional mitochondria with balanced fission, fusion, and mitochondrial movement.Right: Impaired RHOT1 expression may affect mitochondrial movement with consequences for mitochondrial dynamics (unbalanced fusion/fission processes and altered mitophagy rate).A reduced number of functional mitochondria may impact ATP production and ultimately impair GSIS." 2. RHOT1-deficiency in β-cells causes reduced insulin secretion.How does RHOT1 regulate insulin secretion in β-cells?Please explain that.
Response: RHOT1 is needed for mitochondrial trafficking, hence deficiency leads to altered mitochondrial dynamics and eventually reduced number of functional mitochondria.Functional mitochondria are required to maintain proper ATP production, which in turn, is needed for glucosestimulated insulin secretion to occur.Please also see the response to comment 1, the inclusion of this to the Discussion part and the revised Fig. 7.
Discussion, page 12-13, line 467-473: "Studies in neurons have shown that RHOT1 connects the mitochondria via motor proteins to the axon filaments, thereby regulating mitochondrial movement (Schwarz 2013).Therefore, it is likely that RHOT1 is needed for proper mitochondrial movement also in human islets and β-cells, creating a balance in mitochondrial dynamics and maintaining functional mitochondria, as required to maintain the proper ATP production needed for GSIS.Thus, impaired RHOT1 expression alters mitochondrial dynamics, which impacts the number of functional mitochondria, resulting in reduced ATP production and insulin secretion (Fig. 7d)." We have also included a new Fig. 7 with a schematic drawing of RHOT1 function in normal and diseased β-cells (Fig. 7d).Page 30, line 1179-1185, Figure 7d legend: "Based on our results, we suggest that RHOT1 is needed for proper mitochondrial movement in β-cells, creating a balance in mitochondrial dynamics to maintain functional mitochondria.Left: Functional mitochondria with balanced fission, fusion, and mitochondrial movement.Right: Impaired RHOT1 expression may affect mitochondrial movement with consequences for mitochondrial dynamics (unbalanced fusion/fission processes and altered mitophagy rate).A reduced number of functional mitochondria may impact ATP production and ultimately impair GSIS." 3. The authors found an almost complete ablation of Rhot1 protein levels in islets of adult GK rats with the decrease of insulin secretion.Therefore, what is the relationship between Rhot1 expression and insulin secretion in GK rats.It would be helpful to analyze the correlation between Rhot1 expression and insulin secretion.
Response: Based on this valid comment, we correlated the levels of Rhot1 in islets with glucosestimulated insulin secretion in the adult GK rats.Although the Rhot1 levels are very low in the adult GK rats (see Fig. 6g), there is a trend (insignificant) showing a positive correlation between the Rhot1 levels and insulin secretion (see left figure below), indicating that lower Rhot1 levels are linked to decreased insulin secretion.Moreover, in the Wistar rats, where the Rhot1 levels are higher, we found a positive correlation between the levels of Rhot1 in islets and glucose-stimulated insulin secretion (p=0.035),further supporting this hypothesis (see right figure below).We added this information on page 10-11, line 388-390 of the revised ms: "The Rhot1 protein levels in islets correlated positively with GSIS in Wistar rats (r=0.90 and p=0.036), but the correlation was insignificant in adult GK rats (r=0.35, p=0.65), in which both Rhot1 and insulin secretion levels are very low." 4. The authors found the important role of RHOT1 in maintaining mitochondrial function, but there was no data related to the mitochondrial function in GK rats.It would be better to provide the related results to analyze the relationship between Rhot1 expression and mitochondrial function in GK rats.
Response: Thanks for this valid suggestion of investigating mitochondrial function in GK rats.To do this, we performed Western blot analysis of citrate synthase (a mitochondrial matrix TCA cycle enzyme and marker of intact mitochondria) and of components of all five complexes of the electron transport chain using protein available from the same GK and Wistar rats as were analyzed for Rhot1.Here, citrate synthase protein levels were significantly lower in adult GK versus Wistar rats and there was a nominal reduction in adult versus young GK rats (see Fig. 6l, Supplementary Fig. 7f).Moreover, components of all five complexes of the electron transport chain were significantly lower in adult GK versus Wistar rats and protein levels of four components were lower in adult versus young GK rats (see Fig. 6m-q, Supplementary Fig. 7g).
We added this information on page 11 (Results) and 12 (Discussion) of the revised ms, in the methods on page 22 and in the legend of Fig. 6 and Supplementary Fig. 7: Page 11, line 391-397: "Moreover, the protein level of citrate synthase, a mitochondrial matrix TCA cycle enzyme and marker of intact mitochondria, was significantly reduced in islets from adult GK versus Wistar rats, while the reduction in adult versus young GK rats was nominal (Fig. 6l, Supplementary Fig. 7f).We also found reduced protein levels of components from all five complexes in the electron transport chain in islets from adult GK versus Wistar rats, of which four were also reduced in adult versus young GK rats (Fig. 6m-q, Supplementary Fig. 7g).These results are in line with the data in Rhot1-deficient β-cells and support mitochondrial dysfunction in islets from the GK rat."Page 12, line 465-467: "We also found an age-related decline in Rhot1 and lower levels of several components of the electron transport chain in islets from diabetic GK rats, supporting the role of Rhot1 in islet function and T2D." Page 21, line 802-806: "For Western blot analysis, rat islets were lysed in RIPA buffer and the analysis performed as described above for INS-1 832/13 β-cells using the RHOT1 primary antibody (ab188029, abcam), cytochrome C antibody (#4272T; 1:1000, Cell Signaling Technologies) and total OXPHOS rodent antibody cocktail including antibodies against NDUFB8, SDHB, UQCRC2, MTCO1, and ATP5A (#ab110413-MS604, 1:250, abcam)."Page 30, line 1160-1167: "l) Citrate synthase protein levels were lower in islets from adult GK (n=4) vs. Wistar (n=5) male rats (p=0.02) and there was a nominal reduction in adult (n=4) vs. young (n=4) male GK rats (p=0.098)based on t-tests after log-transformation to reach normal distribution.Target signal was normalized to the total amount of protein loaded in each lane.m-q) Reduced protein levels of five and four components of the electron transport chain in islets from adult GK (n=4) vs. control Wistar rats (n=5) as well as in adult vs. young (n=4) GK rats, respectively, based on t-tests after log-6 transformation to reach normal distribution.Target signal was normalized to the total amount of protein loaded in each lane."Page 33, line 1269-1274: e-g) Western blot showing protein levels of RHOT1 (e), CS (f), and five components of the electron transport chain (NDUFB8, SDHB, UQCRC2, MTCO1, and ATP5A) (g) in adult (12 weeks of age; n=4) vs. young (8 weeks of age; n=4) GK rats and control Wistar rats (12 weeks of age; n=5) (top).Protein levels were normalized to the total amount of protein loaded in each lane (bottom) using ImageLab software version 6.0.1 (V3 Western Workflow, Bio-Rad).siNC: non-target control.
5. In Figure 3, RHOT1 expression was decreased by all four siRNA in human islets, such as FOXP1, TBC1D4, RHOT1 and CABLES1.Therefore, how does FOXP1, TBC1D4 and CABLES1 regulate the expression of RHOT1.
Response: Thanks for this question; it is indeed an interesting finding that some genes are affected by silencing of several other genes.For RHOT1 downregulation, different mechanisms may be involved, as suggested by information in different databases, FOXP1 has a binding site in the RHOT1 promoter, whereas TBC1D4 and RHOT1 may interact on the protein level.This has now been included in the Discussion, Page 12, line 452-454: "The reduction in RHOT1 may have different causes; for example, FOXP1 is a TF with a binding site in the RHOT1 promoter, whereas TBC1D4 and RHOT1 display a protein-protein interaction (Rouillard, 2016)." Based on this valid comment, we also performed a new analysis to dissect how FOXP1, TBC1D4 and CABLES1 potentially may regulate the expression of RHOT1 and included these results on page 8, line 284-297: "To characterize potential mechanisms underlying the reduced RHOT1 expression when silencing FOXP1, TBC1D4, and CABLES1 in human islets, we identified 619 TF binding sites in the vicinity of the RHOT1 gene (-1000, 0) using publicly available ChIP-seq data from the Gene Transcription Regulation Database (GTRD) (Kolmykov, Yevshin et al. 2021) (Supplementary Fig. 5e).Since most experiments in the GTRD were performed in non-islet cell types, we examined whether the investigated genomic regions are accessible in human islets, based on published ATAC-seq and histone modification data (Miguel-Escalada, Bonas-Guarch et al. 2019).We identified 362 TFs with 614 binding sites within regions defined as promoters in pancreatic islets (Supplementary Table 16).Next, we searched for overlaps between the 362 TFs that could bind 0-1000 bp upstream of RHOT1 TSS and differentially expressed genes after silencing FOXP1, CABLES1, or TBC1D4 in human islets (Supplementary Table 15, 16).We detected 7, 8 and 13 differentially expressed TFs, and among these, 6, 4 and 8 were downregulated after silencing FOXP1, CABLES1 or TBC1D4, respectively (Supplementary Table 16, Supplementary Fig. 5e).Although dysregulation of each of these TFs could contribute to the reduction of RHOT1, one TF, MLXIP (formerly known as MONDOA), was found to be dysregulated in all 3 silencing experiments.MLXIP is an important glucose-sensing TF in human β-cells (Richards, Rachdi et al. 2018)."6.In figure 4c and 5g, there was no GAPDH or ACTIN.How does the authors perform the quantification in WB analysis.Please explain that in detail.

Response:
We are sorry that the description of the method for Western blot normalization was not fully clear.We have now further explained how this was done, including a reference, in the method section: Page 17, line 658-663: "The antibody-specific signal was quantified and we then used total protein normalization where the detected antibody-specific signal is normalized to total protein in ImageLab software version 6.0.1 (V3 Western Workflow,.With total protein normalization, the antibody specific signal is normalized to the total amount of protein loaded in each lane.This method is more stable for different loading amounts than using house-keeping genes and avoids the risk of the internal control being affected by experimental conditions (Gurtler et al., 2013)." We also added more detailed info into the figure legends: page 29-30: "5c) RHOT1 protein levels in Rhot1-deficient β-cells with the target signal normalized to the total amount of protein in each lane" and "6g) Rhot1 protein in 8-and 12-week-old (young and adult, respectively; n=4) GK rats and 12week-old control Wistar rat (n=5) with the target signal normalized to the total amount of protein in each lane" For consistency, we also changed the legends for supplementary figures accordingly (Page 32, Figure legend for Supplementary Fig. 5h, 6a-c, and 7a-g).
7. In some Figure legends, the sample size n was missed.
Response: Thanks for this thorough review.We have now gone through all figure legends to ensure that sample sizes are included.Also, we also updated the figure legends to meet the Nature Communications formatting guidelines.See the updated Figure legends, page 28-33.

Reviewer #3 (Remarks to the Author):
The manuscript presents a genome-wide DNA methylation study comparing healthy (75 samples) and T2D (25 samples) human islets.The authors also examine the correlation between differential DNA methylation and HbA1c using islets from an independent cohort (114 samples).They integrate the methylation data with RNA-seq, ATAC-seq, transcription factor and histone modifications' ChIP-seq data.They reported four T2D-associated methylation sites associated with lower risk of T2D in blood samples.Selected candidate genes are validated using siRNA knockdown in a rat beta cell line.The study highlights the potential role of Rhot1 in regulating mitochondria function and metabolites, as well as its deficiency in a rodent diabetes model associated with reduced insulin content.Although this study has multiple points, the logical flow between them is not coherent or compelling.Although the authors hint a mechanistic connection between DNA methylation and gene expression, the evidence is preliminary.A lot of results are only described in the text with a brief reference to an excel file without proper presentation or quantitative analysis.
A key result about the prospective T2D risk analysis not adequately presented or discussed.Overall, I feel that this manuscript has major flaws and at least needs a lot of improvements to be acceptable for publication in NC.Here are some specific points.

Response:
We appreciate the valuable comments provided by Reviewer #3, which helped us improve our study and manuscript considerably.We have addressed all review comments, both from the above section, and the specific points below.We believe the revised manuscript and our novel data are of importance and of immediate interest to the broad readership of Nature Communications.

Response to the more general comments above:
"Although this study has multiple points, the logical flow between them is not coherent or compelling" Response: We want to thank Reviewer 3 for this comment.To make research findings easily available for the scientific community, the flow within the manuscript is indeed very important.To improve readability and to make the message and results clearer, we worked together with a professional proofreading and editing company (San Francisco Edit), to optimize the manuscript before re-submission.
We believe this editing substantially improved the manuscript and answered this as well as other concerns.Please see changes marked with track-changes throughout the re-submitted manuscript.

"Although the authors hint a mechanistic connection between DNA methylation and gene expression, the evidence is preliminary"
Response: We appreciate this comment and, in our opinion, the connection between DNA methylation and gene expression is well established within the epigenetic research community.For example, we have several times shown functionally that increased DNA methylation within gene promoters directly regulates (reduce) the transcriptional activity and we already referred to some of these studies in our Results, page 5, line 157-158: "DNA methylation may interfere with transcriptional activity, as proven by in vitro experiments 4,6 and causal inference test 17 ".This notwithstanding, to again prove a mechanism for epigenetic gene regulation, we designed a Luciferase assay by cloning the RHOT1 promoter region into a vector lacking CpG sites.By methylating the promoter sequence in vitro, we were able to downregulate the transcriptional activity in clonal β-cells.This data is now added on page 8-9, line 305-310, and in the methods on page 16, line 617-627, and in Supplementary Fig. 5f.
"First, we investigated the direct impact of increased promoter DNA methylation on the transcriptional activity of RHOT1, using a luciferase assay (Supplementary Methods).The CpG-free luciferase reporter construct including 1540 bp of the RHOT1 promoter was methylated by SssI and transfected into INS-1 832/13 β-cells.Methylation of the promoter reduced the transcriptional activity (Supplementary Fig. 5f), supporting that altered DNA methylation can influence gene expression.""Luciferase assay 1540 bp of the RHOT1 promoter (Supplementary Methods) was inserted upstream of TSS of a CpG-free luciferase reporter vector (pCpGL-basic, kindly provided by Dr Klug and Dr Rehli 74 ).SssI, a DNA methyltransferase (New England Biolabs, Ipswich, MA, USA), was then used for methylation of the construct.INS-1 832/13 β-cells were co-transfected with 25 ng methylated pCpGL-vector including the RHOT1 promoter together with 4 ng of pRL renilla luciferase control reporter vector (Promega, Madison, WI, USA).Firefly luciferase luminescence, as the value of transcriptional activity, was measured using the Dual-Luciferase® Reporter Assay System (Promega) and an Infinite® M200 PRO multiplate reader (Tecan Group Ltd., Männedorf, Switzerland).Cells transfected with an empty pCpGLvector were used as background control for firefly luciferase results, and untransfected cells were used as a background for renilla results."control and T2D groups.Without a global picture, the readers may feel that the authors are only cherry-picking some best examples.
Response: Thanks for this comment.We agree that heatmaps can be a good way to visualize global changes in large datasets.However, you can only visualize a certain number of datapoints within one figure, within the magnitude of a few thousands, but to include and properly visualize ~850,000 rows is not possible.Doing that would require to collapse several methylation sites into one data point, thereby losing the intention of studying a large amount of methylation sites with possible specific effects, rather than just measure global, overall methylation.
Nevertheless, to show the consistency between individuals from the T2D cases and control groups, we have now included a Heatmap in Fig. 1, showing DNA methylation levels for each individual, for the 100 CpG sites with the highest absolute differences in methylation between groups (Results, Page 3, line 86-88; Fig. 1e; and page 28, line 1088-1090 Figure legend 1e)."A heatmap of the 100 sites with largest differences in methylation between T2D cases and controls visualizes the consistency between individuals in each group (Fig. 1e)" and "Hierarchical clustering heatmap showing the DNA methylation pattern for the 100 CpG sites with the largest differences in the Islet T2D case-control cohort.Samples are shown on the x-axis, and methylation sites on the y-axis."Also, Fig. 2a-d clearly show that for specific sites, methylation levels are consistent between the T2D cases and control groups.Moreover, Supplementary Table 1, 3, and 4 present all our epigenome-wide associations and in these tables, we present the regression coefficients together with the standard error as well as mean and SD for respective group and differences between the groups, both absolute differences and fold-change.We believe we are transparent with all our data/results, and our goal was to avoid cherry picking.To generate epigenome-wide associations, we used robust statistical analyses, correction for multiple testing and correction for cell composition as presented in the methods.1e is missing.The significance of the enrichment is not clear.

The p-values for the enrichment in Figure
Response: Thanks for letting us know that this needs to be better explained.We have now updated the legend for Figure 1e (now Fig. 1f), page 28, line 1090-1093: "f) Selected KEGG pathways associated with T2D based on genome-wide DNA methylation data from the Islet of T2D case-control cohort (dark blue; adj. P-values <0.05).Most pathways were also associated with HbA1c, based on genome-wide DNA methylation data from the Islet HbA1c cohort (light blue; adj.P-values <0.05)." Additionally, we have added all individual adjusted p-values into Fig.1f (All unique p-values were previously only found in Table S2).
3. The authors claimed some consistency between current study and previous studies (Dayeh et al.; Volkov et al.) but provide very little details.Only an excel file is provided.Summarization of the comparison should be included.For example, how many hypermethylated sites from this study are hyper or hypo-methylated in the other studies; conversely, how many hypo-methylated sites from this study are hyper or hypo-methylated in the other studies.Quantitative statistical analysis should be also included.

Response:
We appreciate this comment and have included the requested information on page 4, line 121-128 of the revised ms: "We also compared the current EPIC results with a previous study using 450K arrays and islets of 15 T2D cases and 34 controls (Dayeh, Volkov et al. 2014).There was a small overlap of islet donors with the current study (5 cases, 18 controls).Importantly, of sites reported by Dayeh et al. to have altered methylation in islets from T2D cases (q<0.05 and Δβ≥5%), we could replicate 813 sites with consistent differences between cases and controls (Supplementary Table 7), including sites annotated to CACNA1C, CDKN1A, GLP1R, HDAC4, HDAC7, IL6R, KCNQ1, PDE7B, and THADA (Supplementary Fig. 1a-i).In both studies, 98.5% of these 813 sites were hypomethylated in islets from T2D cases versus controls (Supplementary Table 7)." Importantly, here, we only consider results consistent when the different studies show changes in the same direction, e.g., in the comparison with Dayeh et al., sites are hypomethylated both in that study and in the current EPIC T2D case-control cohort, or sites are hypermethylated in both studies.This has now been highlighted in the manuscript and is also clearly shown in Supplementary Table 7.
Regarding statistical enrichment analysis between the current study and the study by Dayeh et al.; we do not believe it is correct to do such an analysis due to the overlap of some samples between the different studies, meaning that the different cohorts are not fully independent.Hence, we present numbers (within the text) and specific sites (Supplementary Table 7), to highlight that these sites/genes have been significantly associated with T2D repeatedly.Moreover, the following sentence was included on page 4: "There was a small overlap of islet donors with the current study (5 cases, 18 controls)."It should be noted, however, that a Chi 2 -test found the overlap enriched (p<0.05,not included in the revised ms).
While it is possible to include all the requested information for the study by Dayeh et al., which investigated DNA methylation of individual CpG sites using the 450k Illumina array, it is not possible to do the same for the study by Volkov et al., which studied differentially methylated regions (DMRs) using whole genome bisulfite sequencing (WGBS).For DMRs, methylation differences presented are based on the average of multiple sites, as described on page 4, line 129-136.Nevertheless, some of the requested information was added here.Moreover, on page 3, line 78-86, and in Fig. 1b-d, we present the largest differences in DNA methylation of individual CpG sites: "The largest absolute methylation difference was 21.1%, at cg15132295 in DNAH17 (Fig. 1b).Lower DNAH17 methylation has been linked to hepatic cancer (Fan, Guo et al. 2019).For increased methylation in T2D cases, the largest absolute difference was 19.5% (cg09972436 in LCE3C, Fig. 1c), a gene linked to autoimmune diseases (Docampo, Rabionet et al. 2010).We also calculated the fold change in DNA methylation, as the impact of absolute differences may depend on whether basal methylation levels are low or high.We found percentage differences up to 80%, meaning that the number of cells being methylated at a specific site was almost doubled in one group (Table S1).This was true for a site in DSCR3, with 12.9% methylation in controls and 23.3% methylation in T2D cases (Fig. 1d).DSCR3 methylation is a potential biomarker for preeclampsia (Kim, Kim et al. 2015)." And on page 3-4 we present pathway analyses, further pointing to enrichment of pathways linking islet DNA methylation to risk of T2D: Supplementary Fig. 4 Line 89-94: "To further explore this comprehensive data and its biological context, we used the methylGSA package to perform a pathway analysis on the whole methylation data set (Ren and Kuan 2019)., including the MAPK, Calcium, and Insulin signaling pathways, Type II diabetes mellitus, Pancreatic secretion, and Purine metabolism (Supplementary Table 2).Fig. 1f visualizes some key pathways, supporting a strong link between islet methylation and T2D." Line 108-112 "A pathway analysis on the whole methylation data set based on HbA1c associations revealed 53 enriched pathways (adj. p-value <0.05).The result resembled that of T2D associations, and 47 pathways overlap with pathways enriched in T2D, including Type II diabetes mellitus and Pancreatic secretion (Supplementary Table 5, Fig. 1f)." 5.It is unclear whether the differential methylation sites are only enriched in the DEG genes and how significant this is.
Response: We appreciate this question.However, we do not believe it is appropriate to calculate significance for enrichment of differential methylation in specific genes in a study based on the EPIC array.The array covers ~850,000 CpG sites, however, these are not evenly distributed throughout the genome.Different genes are represented by a highly variable number of CpG sites, and it also differs where in the gene those CpG sites are located.Moreover, the array is based on 50 bp probes and singlebase extension, which further limits the possibility to cover all genomic features.The overlap of DEG and altered methylation in our study varies between 1-30 significant CpG sites, which could be weighted in different ways (e.g.gene location, number of CpGs in the region, presence of CpG island etc).Hence, we present the overlap in the manuscript "75% of genes with differential expression in islets of T2D cases versus controls also display one or more sites with differential methylation", however, we do not argue that this is a significant enrichment, nor do we believe this is an appropriate test to do with this study design.
Anyway, we performed a Chi 2 -test as this reviewer specifically asked for significance, and indeed, there is an enrichment of differentially methylated sites in DEGs (p=2x10 -21 ).
We will leave it to the reviewer and editor to decide if we should include the significance of enrichment, or if it is misleading, based on the discussion above.
6.The section "DNA methylation associates with future T2D" is extremely brief.Again, the authors only refer the readers to an excel file.Here the authors test the putative CpG sites in blood samples.Should we worry about the consistency between islets and blood?This  risk).This is contradictory.

Response:
We agree with reviewer 3 that this section is brief and based on this comment we added more information in the result section on page 7, in Supplementary Table 14 and in the discussion on page 11-12.
Regarding consistency between the epigenetic pattern in pancreatic islets and blood, the epigenome has a main function in regulating cell specific gene expression and shows large differences between different cell types (Roadmap Epigenomics Consortium, Nature, 2015, PMID: 25693563).Nevertheless, certain CpG sites can have the same epigenetic pattern and regulation under certain conditions in several different cell types.For example, we previously demonstrated that ageing is associated with alterations in DNA methylation of the same sites/genes (e.g., ELOVL2, FHL2, KLF14) in several different cell types such as human adipose tissue, blood, pancreatic islets and the liver (Rönn et al Hum Mol Gen, 2015, PMID: 25861810, Bacos et al Nature Com, 2016, PMID: 27029739, and Bysani et al Epigenomics, 2017, PMID: 27911095), hence supporting that methylation of some sites in blood may mirror the methylation pattern in pancreatic islets.
The overall goal of this study was to better understand the role of epigenic dysregulation in the pathogenesis of T2D, and since pancreatic islets is a key tissue influencing the development of diabetes due to its secretion of insulin, we chose to study DNA methylation in human pancreatic islets.Blood has a less important role in the pathogenesis of T2D, and therefore blood was not the focus of this study.However, if one can identify some sites with differential DNA methylation in pancreatic islets from donors with T2D also showing similar regulation in blood, this may be used as prognostic biomarker.This is the reason for why we also analyzed DNA methylation in blood of a prospective cohort of CpG sites with differential methylation in islets from donors with T2D.It should also be noted that some studies have previously analyzed DNA methylation in blood using either the Infinium 450k array or the EPIC array, and together these studies only found a few methylation sites associated with T2D (ref 27-33 in the ms).Hence, we do not believe screening for candidate predictive DNA methylation sites from blood samples in the beginning would have been better.
However, based on this valid comment, we performed a search of previous T2D DNA methylation studies carried out in blood, summarized these results, and then checked for overlap with the methylation sites identified in islets from the present study.(Chambers, Loh et al. 2015, Florath, Butterbach et al. 2016, Cardona, Day et al. 2019, Juvinao-Quintero, Marioni et al. 2021, Domingo-Relloso, Gribble et al. 2022, Fraszczyk, Spijkerman et al. 2022, Fraszczyk, Thio et al. 2022)." Page 11-12, line 428-435: "Moreover, a previous study found 18 methylation sites in blood associated with future T2D (Cardona, Day et al. 2019).Of these, two were among our T2D-associated methylation sites in human islets (cg00574958/CPT1A and cg02711608/SLC1A5), and the first one was also associated with HbA1c.Two additional studies identified T2D-associated methylation of cg00574958/CPT1A in blood (Juvinao-Quintero, Marioni et al. 2021, Fraszczyk, Spijkerman et al. 2022).Other studies performed in blood identified methylation sites associated with T2D, and several of these sites were annotated to the same genes, but for different CpG sites, as were differentially methylated in human islets of T2D donors in the present study (Chambers, Loh et al. 2015, Florath, Butterbach et al. 2016, Cardona, Day et al. 2019, Juvinao-Quintero, Marioni et al. 2021, Domingo-Relloso, Gribble et al. 2022, Fraszczyk, Spijkerman et al. 2022, Fraszczyk, Thio et al. 2022).Epigenetic alterations common for several tissues may provide therapeutic or predictive biomarkers." Finally, regarding the degree of methylation of the four predictive sites identified in our study: Odds ratio <1 means lower T2D risk with increasing DNA methylation (as seen in the EPIC-Potsdam), which is in accordance with lower DNA methylation in the islet T2D cases.To clarify the different ways to interpret the outcomes, depending on type of analysis, we have now included results for both linear and logistic regression for the Islet T2D case-control cohort in Table 2, page 27: "For islet DNA methylation, both absolute difference (cases-controls) based on linear regressions and odds ratios based on logistic regressions are shown."And as can be seen, the results in blood and pancreatic islets are in agreement, although two separate tissues: "Table 2. DNA methylation associated with T2D in the Islet T2D case-control cohort within genes differentially expressed in islets from T2D cases vs. controls that also associated with future T2D based on blood DNA methylation in the prospective EPIC-Potsdam T2D case-control study.For islet DNA methylation, both absolute difference (cases-controls) based on linear regressions and odds ratios based on logistic regressions are shown." 7. Fig. 5g-k shows that in diabetic rat, the down regulation of Rhot1 is correlated with lower insulin content in the islets, suggesting a causal role.However, in human islets (Fig. S4a), siRHOT1 does not affect islet insulin content.This raise a serious question if and how much of the Rhot1 functional results from rat are conserved in human islets.
Response: We thank Reviewer 3 for this question.Human islet experiments were performed in islets from previously healthy/non-diabetic donors, where RHOT1 is being silenced in vitro for 72h, and where insulin secretion/insulin content was measured and shown to be reduced.Similarly, there is a reduction in GSIS in the GK-rat islets when comparing insulin secretion/insulin content (2.8 ± 0.7 vs. 5.2 ± 0.7, p=0.007).Hence, RHOT1 impairs the insulin secretion process independently of changes in insulin content in both human islets in vitro and in islets from the GK rat.It is true that insulin content is reduced in the GK-rat islets, but Rhot1 is probably one of several factors, which together contribute to diabetes and/or reduced insulin content in the GK rats (see e. g.: Nagao, M. et al. 2020.Selectively Bred Diabetes Models: GK Rats, NSY Mice, and ON Mice.In: King, A. (eds) Animal Models of Diabetes.Methods in Molecular Biology, vol 2128.Humana, New York, NY. https://doi.org/10.1007/978-1-0716-0385-7_3).Moreover, we previously found reduced insulin content in human pancreatic islets from Differentially expressed gene, with one or more differentially methylated CpG sites within +/-10 kb should be at least discussed.In fact, the authors could have screened for candidate predictive DNA methylation sites from blood samples in the beginning.Wouldn't that give better outcomes?The authors names 4 predictive sites, including one with RHOT1, with low T2D risk and low DNA methylation.But in Fig.1, the RHOT1 CpG show low DNA methylation among T2D patients (i.e., low DNA methylation is associated with increased T2D (Chambers, Loh et al. 2015, Florath, Butterbach et al. 2016s showing differential DNA, Juvinao-Quintero, Marioni et al. 2021, Domingo-Relloso, Gribble et al. 2022ry Table1), with published DNA meth, Fraszczyk, Thio et al. 2022 or incident T2D studies performed in blood(Chambers, Loh et al. 2015, Florath, Butterbach et al. 2016, Cardona, Day et al. 2019, Juvinao-Quintero, Marioni et al. 2021, Domingo-Relloso, Gribble et al. 2022, Fraszczyk, Spijkerman et al. 2022, Fraszczyk, Thio et al. 2022).
blood were different from the methylation sites identified in islets (Supplementary Table1, 14)