miRNA142-3p targets Tet2 and impairs Treg differentiation and stability in models of type 1 diabetes

In type 1 diabetes, the appearance of islet autoantibodies indicates the onset of islet autoimmunity, often many years before clinical symptoms arise. While T cells play a major role in the destruction of pancreatic beta cells, molecular underpinnings promoting aberrant T cell activation remain poorly understood. Here, we show that during islet autoimmunity an miR142-3p/Tet2/Foxp3 axis interferes with the efficient induction of regulatory T (Treg) cells, resulting in impaired Treg stability in mouse and human. Specifically, we demonstrate that miR142-3p is induced in islet autoimmunity and that its inhibition enhances Treg induction and stability, leading to reduced islet autoimmunity in non-obese diabetic mice. Using various cellular and molecular approaches we identify Tet2 as a direct target of miR142-3p, thereby linking high miR142-3p levels to epigenetic remodeling in Tregs. These findings offer a mechanistic model where during islet autoimmunity miR142-3p/Tet2-mediated Treg instability contributes to autoimmune activation and progression.

4. The meaning "activated phenotype" should be specified for both mouse and human T cells for Fig.  1g-h. The peripheral blood usually contains very very few activated lymphocytes. 5. Functional blockade of miR-142 activity using nanoparticle delivery of inhibitors should be validated using an activity sensor (e.g. luciferase or fluorescent protein construct with miR-142 bindings sites) rather than PCR. PCR can be affected by inhibitor binding after lysis, and does not provide information about the fraction of cells that receive effective inhibition. It may be sufficient to cite a prior publication in which this technique for delivery has been fully assessed. 6. Why should miR-142 inhibitor and mimic have no effect on sorted naïve T cells from diabetic NOD mice, and in human subjects with recent onset or established T1D? Does this indicate some kind of extrinsic regulation of Treg differentiation that overrides the effect of miR-142 in diabetic individuals? 7. The presented data are not sufficient to assess the affect of miR-142 on Treg stability. The number of CD25+ and FoxP3+ cells in restimulated induced Treg cultures could be affected by differences in the growth and survival of the mixture of Tregs and other T cells in the culure, and/or it could be affected by changes in Treg stability. To test for the latter, FoxP3 lineage markers are needed (e.g. FoxP3;Rosa26-lsl-fluorescent protein mice, which are widely available). 8. Data availability section is incomplete. HITS-CLIP data "will be provided shortly", and there is no mention of NGS miRNA sequencing data. 9. Fig. 5a is unclear regarding chimeric reads. Does the bottom panel indicate the position of a single miR-142 chimeric read, while the middle panel indicates the read density for all chimeric reads? Also, where in this region are there predicted miR-142 binding site(s)? Are they conserved in mice? This is important since many key experiments were performed in mice. Miranda is a very permissive target prediction algorithm with many more false positive predictions than other algorithms such as Targetscan. 10. Further experiments are needed to validate Tet2 as a miR-142 target. Reciprocal miRNA and putative target gene expression are consistent with direct targeting, but could easily be due to indirect effects of the miRNA on other genes. This is especially true for inferences about anti-correlated miRNA and putative target gene expression patterns and related biology in vivo (e.g. Fig. 6). Is the 3' UTR of Tet (or the putative region bound by Ago2:miR-142 complexes) responsive to miR-142 inhibitor and/or mimics in 3' UTR reporter assays? 11. Do the inhibitors affect miR-142 activity in T cells in the mice that are dosed systemically? Or might they be working by alternate means (in other cells to which it is easier to deliver inhibitors, or by some indirect mechanism?). This type of delivery system requires rigorous validation, as previous efforts to block miRNAs in lymphocytes in vivo have not been so successful. Much better in vivo loss of function experiments could be performed using genetic miR-142-deficiency (e.g. Sun et, JCI 2015).
Reviewer #3, expert on Treg (Remarks to the Author): In the submitted paper the authors unveil a pathway that links miR142-3p expression in T cells to a reduction in Tet2 expression and hence Treg instability. The studies are based on type 1 diabetes in both patients and mice, and suggest that increased miR142-3p may play a pathogenic role in diabetes development. In humans, the relationship between the different observations is induced, while in mice it is formally demonstrated up to the point of T cell infiltrate (while impacts on diabetes development are not shown, possibly because such experiments may be impractical using the tools generated here). Overall, while the paper is written a bit roughly, the data are novel and potentially important.
Major points: -The higher number of Tregs with the miR-142-3p inhibitor could be increased stability, as stated in the text, but altered proliferation or apoptosis of either the Treg or Tconventional fractions was not excluded in Figure 4b.
-The studies on Foxp3 methylation patterns in Figure 4 should have been performed with and without miR-142-3p over-expression and inhibition. In the current format, it is not obvious what they contribute to the study.
-Are the changes to Foxp3 methylation in T1D patients compared to healthy control due to: 1) more contamination of transiently Foxp3+ T cells in the Treg isolation; 2) more Nrp1-pTreg within the Treg pool; or 3) partial methylation of tTreg? -The suggested direct link between miR-142-3p, CNS3 methylation and Tet2 could be further tested in human cells by Tet2 siRNA Minor points -It is stylistic, but I would suggest that the insertion of number, standard deviation and p values into the text reduces readability, while not providing information that is not already present on the figure. In addition, the paper would be well served by implementation of the standard paragraphing format -In Figures 4d and 4f, are the remethylated CNS2 Tregs at later time points still Foxp3+? The coexpression of Treg markers is stated in 4c and 4e, but a proper time course with the key markers would be more useful -In Figure 6c, are Tet2+ T cells reduced as a proportion of CD3 T cells? Or are just total T cell numbers down, which means Tet2+ cells will be down as a result?
Response: We thank the reviewer for mentioning these important studies and we agree that they provided considerable insight into the roles of miRNAs in regulating Treg function and T1D islet autoimmunity. However, below we would like to respectfully clarify how our novel data is relevant for the field and how it integrates in the existing literature. In addition, we have performed a series of additional in vitro and in vivo studies including cellular and molecular dissection of the functional implication of the identified miR142-3p/Tet2/Foxp3 axis in the regulation of islet autoimmunity.
Specially, we have included in vivo data in the setting of islet autoimmunity upon in vivo inhibition of miR142-3p demonstrating enhanced CD3 + Tet2 + T cells in the pancreas accompanied by improved stability of the Treg phenotype (see Fig. 8f-h). We have provided proof of principle studies to show miR142-3p inhibition in the setting of established T1D using PBMCs from individuals with T1D for the reconstitution of humanized mice, likewise in these humanized mice we identify increased numbers of CD3 + Tet2 + T cells in the pancreas following blockade of miR142-3p (now in Supplementary Fig. 11). In line with these findings, we have now also included novel analyses using T cells from individuals with recent onset of clinical T1D which presented with significantly reduced numbers of CD3 + Tet2 + T cells (now in Fig.7d and e). Of note, we have now performed novel experiments in the context of very early onset islet autoimmunity in NOD mice below 30 days of age. Importantly here, the identified increased Foxp3 CNS2 DNA methylation in NOD mice below 30 days of age with a very early onset of autoimmunity suggests a potential causative role of miR142-3p/Tet2 signaling in promoting autoimmune activation and progression (presented in Supplementary Fig. 9c). For further details please refer to the manuscript.
Below we discuss the indicated literature in the context of our novel findings: PMID 27438767 addresses how inflammatory cues induce expression or miR-17 which targets Eos, a co-regulatory molecule of Foxp3, diminishing the suppressive capacity of Tregs. While these findings are of considerable importance for the understanding of Treg function, the paper does not address the contribution of miRNAs to impaired Treg induction from naive T cells and their stability during onset of islet autoimmunity, which is the major focus of our manuscript where we provide mechanistic evidence for a direct targeting of Tet2 by miRNA142-3p. Specifically, we find that aberrant miR142-3p expression in CD4 + T cells acting via Tet2 repression functions as one mechanism by which dysregulated DNA methylation at the Foxp3 locus mediates impaired Treg homeostasis, and consequently contributes to autoimmune activation. PMID 26165721 identifies miR-31 as a regulator of Gprc5a (retinoic acid-inducible protein 3) expression, generation of peripheral Tregs and experimental autoimmune encephalomyelitis (EAE) severity. These findings are of interest for the regulation of Treg induction and its contribution to autoimmunity. However, the mechanisms described in our manuscript, in particular epigenetic remodeling of the Foxp3 CNS2 during onset of islet autoimmunity via miR142-3p/Tet2 signaling and the link to impaired Treg induction and stability are not addressed.
PMID 27791035 deals with TFH precursor cells, their role in the onset of islet autoimmunity and how their frequency is controlled by miR-92a. Even though this publication shows miRNA regulation of a T cell subset which is important for autoimmune pathogenesis, it does not address the crucial role of Tregs and specifically their induction and stability in autoimmunity.
PMID 29459719 identifies miR142-3p as a regulator of ATG16L1 and addresses its role in regulating autophagy, proliferation and function in thymic-derived Tregs as well as the implication of these findings for graft-versus-host disease. However, the manuscript does not address the impact of miR142-3p to de novo Treg induction from naive CD4 + T cells and stability of induced Tregs during the onset of islet autoimmunity. Furthermore, the identification of Tet2 as a direct target of miR142-3p and the impaired demethylation of the Fox3 CNS2 are important and novel findings, contributing to our understanding of regulation of Treg induction and stability in autoimmunity.
PMID 23650616 identifies miR142-3p as a regulator of GARP expression on CD25 + CD4 + T cells and, as a result, their expansion in response to activation, which provides insight into the control of Treg expansion. While these findings are of importance for the field, they do not address the role of miR142-3p for Treg induction or stability as well as the critical role of the miR142-3p/Tet2 axis for the onset of islet autoimmunity.
2. Tet2-mediated Foxp3 regulation for Treg differentiation and immune homeostasis has also been reported (PMID: 26275994). While the importance of Tet2 in T1D autoimmunity and progression is not well understood, the authors failed to provide experimental evidence for understanding in vivo functional role of Tet2 downstream miR-142-3p regulation.

Response:
We thank the reviewer for referring to this important study and we agree that it is of considerable importance for the understanding of Tet1-and Tet2-mediated regulation of Foxp3 expression, Treg differentiation and immune homeostasis. In addition to these important discoveries our novel findings provide mechanistic evidence for the role of Tet2 mediated Foxp3 CNS2 demethylation in the context of Treg induction from naive T cells in the setting of autoimmunity. Importantly, to dissect the potential causative contribution of impaired Treg stability in promoting autoimmune progression we now assessed Foxp3 CNS2 DNA methylation in NOD mice below 30 days of age with a very early onset of IAA + positivity. Of note, these very young IAA + NOD mice presented with distinctly increased Foxp3 CNS2 DNA methylation when compared to nonautoimmune prone BALB/c mice (presented in Supplementary Fig. 9c). These novel findings thereby support the concept that early onset Treg instability can be involved in promoting autoimmune activation and progression. In addition to these novel insights on Treg stability impairments in impacting autoimmune activation and progression we provide mechanistic evidence for a miRNA142-3p-Tet2 signaling axis and integrate their contribution in aberrant Treg induction and stability. Specifically, we identified miR142-3p, which is upregulated in autoimmunity, as an upstream regulator of Tet2, linking differential miRNA expression to reduced Tet2 abundance, incomplete Foxp3 CNS2 demethylation and impaired Treg induction and stability in autoimmunity.
In addition, we have performed a series of novel in vitro and in vivo experiments including loss of function models. Specifically, employing a combination of HITS-CLIP and various molecular and cellular approaches, including miR142-3p modulation and loss of function models, we have demonstrated that miR142-3p activity is linked to Tet2 abundance, Foxp3 CNS2 methylation, affecting Treg induction and stability. Concordantly, the inhibition of miR142-3p in NOD mice with ongoing islet autoimmunity significantly reduced Foxp3 CNS2 methylation in Tregs, increased the number of pancreatic Tregs and improved islet infiltration (now in novel Fig. 8). In addition we provide novel data showing that miR142-3p inhibition in vivo also leads to higher numbers of Tet2 + T cells in the pancreas (novel data in Fig. 8f and g) directly linking miR142-3p to Tet2 expression, epigenetic remodeling in Tregs and islet autoimmunity. In addition, we performed novel studies using humanized NSG mice. Humanized mice are immunodeficient mice that after reconstitution with human hematopoietic cells or tissues do develop a human immune system with a highly diverse TCR repertoire. These mice permit the assessment of human T cell responses in vivo. Specifically, we made use of the murine MHCII deficient, HLA-DQ8 transgenic NOD.Cg-Prkdc scid Il2rg tm1Wjl (NSG) mouse model 1 . To assess a potential relevance of miR142-3p inhibition in T cells from T1D phenotypes we employed PBMCs from individuals with recent onset T1D for a pilot experiment to reconstitute NSG mice for the establishment of humanized mice. Application of a miR142-3p inhibitor to these humanized NSG mice showed a trend towards increased frequencies of peripheral CD127 low CD25 hi Tregs and a significant increase in Tet2 expression in pancreatic T cells ( Supplementary Fig. 11).
3. In Fig. 7, the authors applied systemic delivery of the microRNA inhibitor to treat islet autoimmunity in vivo. Nevertheless, the authors did not provide evidence for miR-142-3p knockingdown efficacy in islet tissue. Furthermore, therapeutic effects of miR-142-3p inhibition on NOD diabetic phenotypes are lacking.

Response:
We thank the reviewer for this comment. We used an LNA in vivo miRNA inhibitor which has been shown to be effective in a broad range of tissues. When delivered systemically, LNA miRNA inhibitors distribute broadly into most tissues, including hematopoietic tissues such as lymph nodes, spleen and bone marrow 2 . We agree that besides the accumulation in the tissues, the knocking-down efficacy in the relevant tissue is highly relevant. Here, the efficient knock-down of miR142-3p using systemic administration of a LNA miRNA inhibitor has been shown in splenocytes 3,4 . A list of studies using in vivo miRNA inhibition with LNA miRNA inhibitors can be found here: http://www.exiqon.com/ls/Documents/Scientific/mirna-inhibition-publications.pdf.
To further strengthen the evidence for successful delivery of the inhibitor to pancreatic tissue we performed novel experiments and upon application of the miR142-3p inhibitor validated the expression of a well-established miR142-3p target Tgfbr1 5 directly in pancreatic T cells.
Specifically, we demonstrated delivery to pancreatic tissue and local immune cells by showing increased expression of Tgfbr1 in pancreatic T cells when compared to NOD mice that had received a control inhibitor (Supplementary Fig. 10a and b).
Regarding the therapeutic effects of miR142-3p on diabetic phenotypes we would like to respectfully draw the attention of the reviewer to the in vitro Treg induction data presented in Fig. 2 and 3 and the in vivo miR142-3p inhibition data shown in Fig. 8 of the manuscript. We showed that miR142-3p inhibition in vitro results in increased Treg induction efficacy in NOD mice with and without islet autoimmunity, while there was only a trend towards a comparable improvement in NOD mice with established T1D. In human T cells miR142-3p inhibition improved Treg induction capacity in both subjects with recent onset of T1D and with long-term T1D. Here we wish to point to the differences of established T1D between NOD mice and humans. While in humans insulin replacement therapy controls blood glucose levels relatively well, in mice the lack of insulin can result in severe dysglycemia. We suggest that this difference also affects Treg induction which is in line with the finding that Treg induction capacity without any miRNA modulation differs significantly between IAA + and diabetic NOD mice, but it does not between human subjects with recent onset and longterm diabetes. Furthermore, we showed that inhibition of miR142-3p in IAA + NOD mice in vivo significantly improves islet autoimmunity, which is causative for the development of T1D. These improvements include reduced islet infiltration ( Fig. 8a and b), lower IAA levels (Fig. 8c), increased abundance of Foxp3 + Tregs directly in the pancreas ( Fig. 8d and e) and improved demethylation of the Foxp3 CNS2 in Tregs (Fig. 8h).
In addition we provide novel data showing that miR142-3p inhibition in vivo also leads to higher numbers of Tet2 + T cells in the pancreas ( Fig. 8f and g) directly linking miR142-3p to Tet2 expression, epigenetic remodeling in Tregs and islet autoimmunity. In addition, we performed novel studies using humanized NSG mice. As outlined above humanized mice are immunodeficient mice that after reconstitution with human hematopoietic cells or tissues do develop a human immune system with a highly diverse TCR repertoire. Specifically, we made use of the murine MHCII deficient, HLA-DQ8 transgenic NOD.Cg-Prkdc scid Il2rg tm1Wjl (NSG) mouse model 1 . To assess a potential relevance of miR142-3p inhibition in T cells from T1D phenotypes we employed PBMCs from individuals with recent onset T1D for a pilot experiment to reconstitute NSG mice for the establishment of humanized mice. Application of a miR142-3p inhibitor to these humanized NSG mice showed a trend towards increased frequencies of peripheral CD127 low CD25 hi Tregs and a significant increase in Tet2 expression in pancreatic T cells (novel data in Supplementary Fig. 11).
Furthermore, we have now also included novel analyses using T cells from individuals with recent onset of clinical T1D which presented with significantly reduced numbers of CD3 + Tet2 + T cells (now in Fig.7d  Otherwise, standard deviation (s.d) should be used for representing the error.

Response:
We thank the reviewer for this suggestion and agree on this point. We made changes in the manuscript, indicating the respective statistical test for multiple group comparison. In addition we specified how the experiments were repeated, as requested.

Minor concerns
1. In Fig. 1, whether miR-142-3p expression is correlated with T1D disease severity remains to be revealed.

Response:
We thank the reviewer for this comment. To address the question whether aberrant miR142-3p/Tet2 signaling together with alterations in Treg induction and stability is correlated with T1D disease severity we performed novel studies where we specifically focused on early onset of autoimmunity. Specifically, to dissect the potential causative contribution of impaired Treg stability in promoting autoimmune progression and thereby early onset and severity of clinical disease we assessed Foxp3 CNS2 DNA methylation in NOD mice below 30 days of age with a very early onset of IAA + positivity. Importantly, these very young IAA + NOD mice presented with distinctly increased Foxp3 CNS2 DNA methylation when compared to non-autoimmune prone BALB/c mice.
These novel findings are now presented in Supplementary Fig. 9c and support a potential causative role of miRNA142-3p/Tet2 signaling in promoting early onset Treg instability, autoimmune activation thereby impinging on progression and T1D disease severity.

Response:
We thank the reviewer for this suggestion. In order to confirm targeting of Tet2 by miRNA-142-3p we performed a series of novel mechanistic studies. Specifically, we performed luciferase reporter assays as requested. To this end, we co-transfected HEK-293 cells with a TET2 3'UTR reporter construct and a miR142-3p mimic and observed a lower Firefly/Renilla luminescence ratio compared to the transfection control. These novel data are now presented in Fig. 6e.
In line with the broad role of Tet2 in various cell types we first made use of the non-hematopoietic 3T3 fibroblast cell line, which was previously shown to lack miR142-3p expression almost completely 6 . As a second approach we dissected the miRNA142-3p-Tet2 axis using T cells from miR142 knockout (miR142 -/-) mice 7 . For further details please refer to the novel Supplementary Fig.   8.
Specifically, we transfected 3T3 fibroblasts with a miR142-3p mimic or a control mimic to introduce miRNA142-3p levels given the absence of endogenous miR142-3p expression. The introduction of miRNA142-3p expression in these cells following mimic application resulted in a significant decrease of Tet2 mRNA levels. These novel findings are now provided in Fig. 6f.
To further strengthen the mechanistic evidence for the direct miRNA142-3p-Tet2 relationship we used T cells from miRNA142 knockout animals. Stimulation of CD4+T cells from miRNA142 knockout mice showed significantly higher Tet2 mRNA levels when compared to T cells from miRNA142 competent mice, now presented in (Fig. 6g). The direct targeting of Tet2 by miRNA142 was also validated on the protein level. Specifically, T cells from miRNA142 deficient animals presented with significantly higher Tet2 protein expression following TCR stimulation when compared to T cells from miR142 +/+ mice (Fig. 6h). These findings provide compelling evidence for direct targeting of Tet2 by miRNA142.
3. Scale bars are missing in Fig. 6 and Fig. 7 islet images.

Response:
We thank the reviewer for this comment and apologize for the mistake. We have included scale bars in the micrographs in former Fig. 6 (currently Fig. 7) and former Fig. 7 (currently Fig. 8).

Reviewer #2, expert on micro-RNA and autoimmune disease (Remarks to the Author):
Scherm et al link miR-142 to Tet2 and the regulation of Treg differentiation in autoimmune disease.
The subject matter and proposed "axis" of regulation that could contribute to loss of immunologic self-tolerance are of high interest. However, the presented evidence do not strongly support the existence and importance of this 'axis' due to numerous technical issues, incomplete data presentation, and the lack of validating data for several key techniques.

Response:
We thank the reviewer for the positive comments, constructive criticism and interest in our studies. We are specifically grateful for the suggestion of additional experiments, which turned out to be extremely helpful in improving our dataset and revising this manuscript. In order to address the reviewer's concerns and strengthen the conclusions, we performed novel in vitro and in vivo experiments, reanalyzed data and provided additional information for clarification as requested.  Fig. 5a) indicates rather low read density and provides no further information about Ago2 binding to the transcript, which is a known target of miR-29. As the data are currently presented, they do not provide convincing support for miR-142 targeting of Tet2.

Response:
We thank the reviewer for this important comment and the suggestion to provide additional information to strengthen our HITS-CLIP dataset and especially the chimeric reads. In accordance with the information provided here, we expanded the description of the HITS-CLIP analysis and specific requested details as mentioned above in the methods section of our manuscript. to modify the presentation of the data accordingly in order to provide convincing support for miR142-3p targeting of Tet2. For further details please refer to the manuscript.
In relation to questions outlined above we also include additional specific information below: HITS-CLIP -sequencing and alignment statistics Two libraries, mRNA-and miRNA-enriched, were sequenced to 100bp on a HiSeq 2500. Two processing pipelines were used. The first considered only the non-chimeric reads in that the alignment program, bowtie, was looking for end-to-end alignments and did not allow significant parts of the read to 'dangle'. The second, applied to just the mRNA library used STAR to align to the transcriptome and did allow dangling ends. Dangling ends were then selected for lengths consistent with mature miRNAs and aligned to miRNA hairpins.

HITS-CLIP -non-chimeric statistics
The non-chimeric pipeline was described previously 8 . The non-chimeric pipeline trims adapter sequence using a program written in the Kaestner lab. There was some evidence of double ligation due to the enrichment for chimeric reads. We trimmed adapters from the 3' end that (partially) match GTGTCAGTCACTTCCAG or TGTCAGTCACTTCCAG. We kept sequences that were at least 16bp long. After trimming reads were aligned to the human genome, human RefSeq transcripts, and human miRNA hairpin sequences using bowtie allowing for multiple alignments. The mRNA library is enriched for long fragments, many of which will span introns or are chimeric, so the genomic alignment rate is lower than the miRNA library. RefSeq alignment for the mRNA fraction is also lower due to the chimeric fragments. The miRNA fraction of the miRNA library is relatively high.
Roughly 81 million mRNA fragments and 68 million miRNA reads where used in the following steps. We identified locations of RISC occupancy on RefSeq transcripts as described previously 8 .
Briefly, the alignments were processed to count the number of times each position in the transcripts occurred at the start of an alignment. Then these weighted positions were clustered (from heavy to light) into bins of 10bp to create the 5' ends of the RISC complex footprints. We identified 265,406 footprints on 28,693 transcripts. To quantify miRNA occupancy in the RISC-complex we counted the number of read that overlapped with the annotated locations (miRbase v20) on the miRNA hairpins. reads. The filtering rules were designed to select reads that were either non-chimeric well-aligned reads or chimeric reads liable to contain a miRNA fragment at the 5' end. Alignments had to be a clean end-to-end match, a match with 1 or 2 bp leading non-match, a long miRNA-like lead nonmatch (18bp or more), or an 8 bp trailing non-match. Alignments were considered bad if they had inserts or deletions in the alignment, or too short a match (less than 20bp). Of these reads, 3,912,266 had a potential miRNA at the 5' end. We then aligned the potential miRNA portion of the likelychimeric reads to miRNA hairpins using bowtie. We found 728,596 (18%) reads that aligned to 290 miRNAs. We then merged the STAR and bowtie alignment information for the reads to identify mRNA/miRNA pairs. Moreover, we evaluated the quality of the chimeric pairing by measuring the distribution of chimeric reads relative to mirRanda-predicted 9 miRNA binding sites for a set of the most frequent miRNAs.
The novel Supplementary Fig. 6 indicates that the chimeras generally agree well with miRanda predictions.

Response:
We thank the reviewer for this comment and apologize for the unclear labeling of Fig. 1d.
This was due to our HITS-CLIP pipeline which appends the 3p/5p suffix only if it needs to in order to infer the existence of a mature form that is not already included in the miRbase release we are using. We reanalyzed the data accordingly and added the suffix for all miRNAs.
3. Fig. 1F should indicate the variability in replicates (n=4 per group), and indicate whether and how read counts were normalized for quantification. Control miRNAs should be shown, and full sequencing data including sufficient information to assess data quality should be provided.

Response:
To mechanistically dissect impaired Treg induction during onset of islet autoimmunity, we determined miRNA expression profiles by next generation sequencing (NGS) using CD4 + T cells from non-diabetic children with or without islet autoimmunity. This pilot RNA-Seq experiment has been performed with pooled CD4 + T cell samples from four individuals with islet autoimmunity in comparison to four individuals without islet autoimmunity. To overcome this limitation validation experiments have been performed for miR142-3p with CD4 + T cells from children with recent onset of T1D and without T1D ( Fig. 1g: here box-and-whisker plots indicating minimum to maximum values to demonstrate data distribution have been used). We apologize for not making this clearer in the first version of our manuscript. The processing, normalization and statistical analysis of the sequencing data were performed as follows: Unwanted adaptor sequences were trimmed from small RNA reads using BTrim 10 and quality of sequencing was assessed for trimmed read data with a mean phred quality score of 38, referring to a base call accuracy of 99.99%. Read data was filtered of unwanted RNA fragments by mapping on rRNA, tRNA, snRNA and snoRNA sequences obtained from the Rfam database using bowtie 11 . Remaining reads were then mapped on mature human miRNA sequences obtained from mirBase (release 20) 12 and summed up to read count lists using SAMTools. mRNA read data was processed comparably without unnecessary trimming and filtering.
Raw read data was mapped on the human genome (build 37.2) using a gapped alignment for paired end data with bowtie2 13 . Finally, read count lists were created by HTSeqcount 47 14 . Differential expression of miRNA was evaluated using DESeq 15 , handling size factor correction and normalization. More information on NGS miRNA sequencing data, including a set of most abundant miRNAs relevant for T cell activation and/or Treg induction, can be found in a previous publication 16 . 4. The meaning "activated phenotype" should be specified for both mouse and human T cells for Fig. 1g-h. The peripheral blood usually contains very very few activated lymphocytes.

Response:
We thank the reviewer for this comment and apologize for the lack of clarity. Human activated T cells, as in Fig. 1f and g, were sorted as CD4 + CD3 + CD45RA -CD45RO + CD127 + CD25 intermediate and murine activated T cells, as in Fig. 1h, were sorted as CD4 + CD25 -CD44 high . The corresponding gating strategies are now shown in Supplementary Fig. 1a and b.
5. Functional blockade of miR-142 activity using nanoparticle delivery of inhibitors should be validated using an activity sensor (e.g. luciferase or fluorescent protein construct with miR-142 bindings sites) rather than PCR. PCR can be affected by inhibitor binding after lysis, and does not provide information about the fraction of cells that receive effective inhibition. It may be sufficient to cite a prior publication in which this technique for delivery has been fully assessed.

Response:
We thank the reviewer for this comment. The successful chitosan-coated PLGA nanoparticle-mediated miRNA uptake in CD4 + T cells including intracellular co-localization of the nanoparticles and the miRNA as well as effective miRNA inhibition have been shown for example in Serr et al. 2018 16 . We agree that qPCR can be affected by inhibitor binding after lysis and therefore 6. Why should miR-142 inhibitor and mimic have no effect on sorted naïve T cells from diabetic NOD mice, and in human subjects with recent onset or established T1D? Does this indicate some kind of extrinsic regulation of Treg differentiation that overrides the effect of miR-142 in diabetic individuals?

Response:
We thank the reviewer for this important and constructive comment. We showed that miR142-3p inhibition in vitro results in increased Treg induction efficacy in NOD mice with and without islet autoimmunity, while there was only a trend towards a comparable improvement in NOD mice with established T1D. In humans miR142-3p inhibition improved Treg induction capacity in both subjects with recent onset of T1D and with long-term T1D. Here we wish to point to the differences of established T1D between NOD mice and humans. While in humans insulin replacement therapy controls blood glucose levels relatively well, in mice the lack of insulin leads to severe dysglycemia. We suggest that this difference also affects Treg induction which is in line with the finding that Treg induction capacity without any miRNA modulation differs significantly between IAA + and diabetic NOD mice, but there is only a trend towards a difference between human subjects with recent onset and long-term diabetes. Regarding the mimic we agree with the reviewer and suggest that regulation induced by the onset of diabetes could lead to multifactorial impairments in Tregs and their induction, including an increase in signaling strength of stimulation, as reported previously 16  FoxP3;Rosa26-lsl-fluorescent protein mice, which are widely available).

Response:
We thank the reviewer for this important comment and the suggestion to study the aspect of Treg stability in our restimulation experiments in more detail. We addressed this point by performing additional Treg induction and restimulation experiments. These studies provided novel data regarding cell viability and proliferation. In addition we would like to highlight that we used BALB/c Foxp3GFP reporter mice for the restimulation experiments enabling the sorting of highly pure CD4 + CD25 hi Foxp3 hi Tregs (purity > 95%) as a starting population for restimulation. As indicated in the novel Supplementary Fig. 4a and b, there were no significant differences in viability (percentage and number) or proliferation (Ki67 + cells) in the CD4 + or the CD4 + CD25 hi Foxp3 hi Treg fraction after Treg induction in presence of the miR142-3p inhibitor or the control inhibitor.
Similarly, the restimulation of the highly pure CD4 + CD25 hi Foxp3 hi Treg fraction in presence of the miR142-3p inhibitor or the control inhibitor did not affect cell viability or proliferation ( Supplementary Fig. 4c). These novel results support the notion that miR142-3p inhibition increases Treg stability in vivo.
8. Data availability section is incomplete. HITS-CLIP data "will be provided shortly", and there is no mention of NGS miRNA sequencing data.

Response:
We thank the reviewer for this comment and apologize for the incomplete data availability section which we completed as requested. All HITS-CLIP library sequencing data have been deposited into NCBI GEO under accession number GSE124264. More information on NGS miRNA sequencing data, including a set of most abundant miRNAs relevant for T cell activation and/or Treg induction, can be found in a previous publication 16 . 9. 6e.
In addition, we used two loss-of-function models to mechanistically validate Tet2 as a direct target of miR142-3p. In line with the broad role of Tet2 in various cell types we first made use of the nonhematopoietic 3T3 fibroblast cell line, which was previously shown to lack miR142-3p expression almost completely 6 . As a second approach we dissected the miR142-3p-Tet2 axis using T cells from miR142 knockout (miR142 -/-) mice 7 . For further details please refer to the novel Supplementary Fig.   8.
Specifically, we transfected 3T3 fibroblasts with a miR142-3p mimic or a control mimic to introduce miR142-3p levels given the absence of endogenous miR142-3p expression. The introduction of miR142-3p expression in these cells following mimic application resulted in a significant decrease of Tet2 mRNA levels. These novel findings are now provided in Fig. 6f.
To further strengthen the mechanistic evidence for the direct miR142-3p-Tet2 relationship we used T cells from miR142 knockout animals. Stimulation of CD4+T cells from miR142 knockout mice showed significantly higher Tet2 mRNA levels when compared to T cells from miR142 competent mice, now presented in (Fig. 6g). The direct targeting of Tet2 by miR142 was also validated on the protein level. Specifically, T cells from miR142 deficient animals presented with significantly higher Tet2 protein expression following TCR stimulation when compared to T cells from miR142 +/+ mice ( Fig. 6h). These findings provide compelling evidence for direct targeting of Tet2 by miR142-3p.
11. Do the inhibitors affect miR-142 activity in T cells in the mice that are dosed systemically? Or might they be working by alternate means (in other cells to which it is easier to deliver inhibitors, or by some indirect mechanism?). This type of delivery system requires rigorous validation, as previous efforts to block miRNAs in lymphocytes in vivo have not been so successful. Much better in vivo loss of function experiments could be performed using genetic miR-142-deficiency (e.g. Sun et, JCI 2015).

Response:
We thank the reviewer for this comment. We used an LNA in vivo miRNA inhibitor which has been shown to be effective in a broad range of tissues. When delivered systemically, LNA miRNA inhibitors distribute broadly into most tissues, including hematopoietic tissues such as lymph nodes, spleen and bone marrow 2 . We agree that besides the accumulation in the tissues, the knocking-down efficacy in the relevant tissue is highly relevant. Here, the efficient knock-down of miR142-3p using systemic administration of a LNA miRNA inhibitor has been shown in splenocytes 3,4 . A list of studies using in vivo miRNA inhibition with LNA miRNA inhibitors can be found here: http://www.exiqon.com/ls/Documents/Scientific/mirna-inhibition-publications.pdf.
To further strengthen the evidence for successful delivery of the inhibitor to pancreatic tissue we performed novel experiments and upon application of the miR142-3p inhibitor validated the expression of a well-established miR142-3p target Tgfbr1 5 directly in pancreatic T cells.
Specifically, we demonstrated delivery to pancreatic tissue and local immune cells by showing increased expression of Tgfbr1 in pancreatic T cells when compared to NOD mice that had received a control inhibitor. These novel data are now presented in Supplementary Fig. 10a.

Reviewer #3, expert on Treg (Remarks to the Author):
In the submitted paper the authors unveil a pathway that links miR142-3p expression in T cells to a reduction in Tet2 expression and hence Treg instability. The studies are based on type 1 diabetes in both patients and mice, and suggest that increased miR142-3p may play a pathogenic role in diabetes development. In humans, the relationship between the different observations is induced, while in mice it is formally demonstrated up to the point of T cell infiltrate (while impacts on diabetes development are not shown, possibly because such experiments may be impractical using the tools generated here). Overall, while the paper is written a bit roughly, the data are novel and potentially important.

Response:
We thank the reviewer for his/her enthusiasm for our work, constructive criticism and positive comments on our manuscript highlighting novelty and potential importance. In response to these comments, which were very helpful for this revision, we conducted a series of additional in vitro and in vivo experiments to strengthen our conclusions and improve the quality of the manuscript as suggested.
Major points: -The higher number of Tregs with the miR-142-3p inhibitor could be increased stability, as stated in the text, but altered proliferation or apoptosis of either the Treg or Tconventional fractions was not excluded in Figure 4b.

Response:
We thank the reviewer for this important comment and his/her helpful suggestion. We would like to highlight that we used BALB/c Foxp3GFP reporter mice for the restimulation experiments enabling the sorting of highly pure CD4 + CD25 hi Foxp3 hi Tregs (purity > 95%) as a starting population for restimulation. We performed novel Treg induction and restimulation experiments and analyzed cell viability and proliferation as requested. As indicated in Supplementary Fig. 4a and b, there were no significant differences in viability (percentage and number) or proliferation (Ki67 + cells) in the CD4 + or the CD4 + CD25 hi Foxp3 hi Treg fraction after Treg induction in presence of the miR142-3p inhibitor or the control inhibitor. Similarly, the restimulation of the highly pure CD4 + CD25 hi Foxp3 hi Treg fraction in presence of the miR142-3p inhibitor or the control inhibitor did not affect cell viability or proliferation ( Supplementary Fig. 4c).
These novel results support the notion that miR142-3p inhibition increases Treg stability in vivo.
-The studies on Foxp3 methylation patterns in Figure 4 should have been performed with and without miR-142-3p over-expression and inhibition. In the current format, it is not obvious what they contribute to the study.

Response:
We thank the reviewer for this comment and apologize for the lack of clarity. To illustrate the importance of the findings presented in Fig. 4c Fig. 5a-d). These findings support the importance of this early phase of Treg induction and point to an involvement of the DNA demethylation machinery.
Regarding the requested studies on methylation patterns with and without miR142-3p inhibition we would like to refer to the limitations of our in vitro Treg induction system. Specifically, the rapid remethylation of the Foxp3 CNS2 in vitro interferes with the precise analysis of relatively small changes in DNA methylation. However, in addition we provide evidence by demonstrating that miRNA142-3p inhibition in vivo impinges on Foxp3 CNS2 methylation, as outlined now in Fig.8h.
-Are the changes to Foxp3 methylation in T1D patients compared to healthy control due to: 1) more contamination of transiently Foxp3+ T cells in the Treg isolation; 2) more Nrp1-pTreg within the Treg pool; or 3) partial methylation of tTreg?

Response:
We thank the reviewer for this important question. To address this question we performed novel experiments to analyze the composition of the Treg pool in T1D patients compared to healthy controls with regard to thymus-derived and peripheral Tregs. Specifically, we measured the expression of Helios, which is a marker of thymic-derived Tregs 19 , in Tregs isolated from peripheral blood of human subjects without T1D and with recent onset of T1D. Helios expression did not differ between the two groups ( Supplementary Fig. 9d), suggesting that the observed differences in Foxp3 CNS2 DNA methylation are not affected by the composition of the Treg pool. Minor points -It is stylistic, but I would suggest that the insertion of number, standard deviation and p values into the text reduces readability, while not providing information that is not already present on the figure.
In addition, the paper would be well served by implementation of the standard paragraphing format Reviewers' comments: Reviewer #1 (Remarks to the Author): The authors have appropriately addressed this reviewer's concern and questions.
Reviewer #2 (Remarks to the Author): The updated manuscript contains some improvements, and new data that strengthen some parts of the study. However, my previous comments about the HITS-CLIP data and the validation of miRNA antagonism in T cells (especially in tissues in vivo) have not been addressed adequately. These are important issues, and both techniques require further rigorous analysis and validation by orthogonal methods.
Evidence for miR-142 targeting of Tet2 remains unconvincing, and the HITS-CLIP data do not appear robust. After considerable effort, I was able to find the location of the 5 apparently identical miR-142 chimeric reads from the HITS-CLIP dataset. Could these 5 reads be amplified from the same cloning event? I see no mention of the use of unique molecular identifiers. In any case, there is no evidence for a mir-142 binding sequence in this region using miRANDA (at microrna.org) or Targetscan. Instead, this region contains one of several deeply conserved mir-29 binding sites. The new figure highlights just how sparse CLIP reads are in this region of the Tet2 3' UTR as compared to other parts of the mRNA, although it remains unclear whether those reads coincide with binding sites for other miRNAs that have been previously validated to target Tet2. Overall, the data provided indicates that the mRNA part of the HITS-CLIP data are not of sufficient quality to support specific miRNA:target interactions. It's not clear that any known miRNA binding sites are detected, much less how well these data agree with target predictions and prior data on miRNA/Ago binding to human transcripts. Supplementary Figure 6 shows only 6 unnamed regions with miRanda predicted biding sites.
The updated manuscript still lacks evidence for delivery of functional antagomirs in vivo and even in vitro. If this reagent is working as proposed, very efficient delivery must be achieved, since miRNA miR-142 is the most abundant miRNA in T cells. Measurement of two known targets is not sufficient, as these genes may be affected by the treatment through indirect pathways. I am aware that this method for antagomir delivery to T cells was used in a prior publication, but that reference also lacks evidence for direct effects on specific miRNA activity. An activity sensor should be used to determine what fraction of miR-142 activity in T cells is relieved by delivery of the miR-142 specific antagomir compared with controls. Genetic deletion of miR-142 can be achieved with existing mutant mice. They would provide a better means of loss of function analysis.
Reviewer #3 (Remarks to the Author): The authors have adequately addressed all of the concerns raised in my initial review. The paper is of sufficient scope and quality to be published at Nature Communications. The authors have appropriately addressed this reviewer's concern and questions.

Response:
We thank the reviewer and are delighted that we have addressed all his/her concerns and questions satisfyingly and that the reviewer finds our newly added data critical to strengthen and clarifying our manuscript.

Reviewer #2, expert on micro-RNA and autoimmune disease (Remarks to the Author):
The updated manuscript contains some improvements, and new data that strengthen some parts of

Response:
We thank the reviewer for his comments. Our understanding is that in the view of reviewer 2 the HITS-CLIP data and the quality of the accompanied analyses need further clarification (1) and that within the HITS-CLIP data analysis the evidence for miR142-3p targeting of Tet2 is not fully convincing (2). Please find our responses to these concerns detailed below: (1) While we understand the reviewer's concern regarding the absence of a predicted miR142-3p binding site in the Tet2 3'UTR, we would like to respectfully draw the attention of the reviewer to Supplementary Fig. 6, showing the quality of the chimeric pairing by measuring the distribution of chimeric reads relative to mirRanda-predicted miRNA binding sites for a set of the most frequent miRNAs. The Supplementary Fig. 6 indicates that the chimeras generally agree well with miRanda predictions, however, we also found a few miRNAs where the distribution of chimeric reads did not correlate with predicted binding sites. To the best of our knowledge miRNA target prediction tools first of all rely on the well characterized 6 nucleotide seed sequence at the 5′ end of the miRNA  Fig. 3c and d).
The delivery of functional miR142-3p inhibitors in vivo was demonstrated using an LNA in vivo miRNA inhibitor which has been shown to accumulate in a broad range of tissues and to efficiently inhibit its targets in several independent studies 13,14 . In order to provide experimental evidence for the delivery of functional miR142-3p inhibitors specifically to CD4 + T cells in vivo we used a fluorescently-labeled miR142-3p inhibitor. Importantly, the successful delivery of the inhibitor to To support the notion that miR142-3p directly targets the Tet2 mRNA, we co-transfected HEK-293 cells with a TET2 3'UTR reporter construct and a miR142-3p mimic and observed a significant reduction in luciferase activity compared to the transfection control (Supplementary Figure 3b).
Furthermore, to support the direct miR142-3p -Tet2 target relationship, we employed two loss-offunction models, 3T3 fibroblasts and miR142 knockout (miR142 -/-) mice. As virtually all nonhematopoietic cells the 3T3 fibroblast cell line lacks miR142-3p expression almost completely ( Supplementary Figure 8a), providing a suitable experimental system to study the effect of miR142-3p in absence of endogenous miR142-3p expression. Here, the introduction of miR142-3p in these cells via transfection with a miR142-3pc mimic resulted in significantly decreased Tet2 mRNA levels ( Figure 6f).
Next, we used T cells from miR142 deficient animals to validate the direct targeting of Tet2 by miR142-3p directly in a mouse model. The stimulation of CD4 + T cells from miR142 -/mice resulted in significantly increased Tet2 mRNA levels when compared to T cells from miR142 +/+ mice ( Figure   6g). In line with the increased expression of Tet2 mRNA, Tet2 protein levels were likewise elevated in T cells from miR142 deficient mice following TCR stimulation when compared to T cells from mice expressing miR142 ( Figure 6h).
As described above, stimulation of BALB/c T cells in presence of a miR142-3p inhibitor resulted in increased expression of Tet2 (Figure 6b) when compared to a control inhibitor. To confirm that these changes are due to a direct effect of miR142-3p inhibition and that Tet2 is a direct target of miR142-  Taken together, we show the successful miRNA uptake in CD4 + T cells including intracellular colocalization of the nanoparticles and the fluorescently-labeled miRNA ( Figure 1 for the reviewers).
Moreover, we now demonstrate reduced miR142-3p activity in T cells by delivery of the miR142-3p inhibitor compared with a control inhibitor using experiments with a miR142-3p activity sensor as suggested ( Figure 2 for the reviewers / Supplementary Figure 3b in the revised manuscript).

miR142-3p inhibitor delivery in vivo
For the inhibition of miR142-3p in vivo we used an LNA in vivo miRNA inhibitor which has been shown to accumulate in a broad range of tissues and to efficiently inhibit its targets in several  independent studies 13,14 . Specifically, important studies using LNA miRNA inhibitors to successfully inhibit miRNAs in vivo, among others in T cells, are listed in Table 1 and an even more extensive list of studies can be found here: http://www.exiqon.com/ls/Documents/Scientific/mirnainhibition-publications.pdf. Table 1. In vivo miRNA inhibition using LNA miRNA inhibitors. List of selected publication. [REDACTED] We now performed additional novel experiments using a fluorescently-labeled miR142-3p inhibitor in order to provide further experimental evidence for the delivery of functional miR142-3p inhibitors specifically to CD4 + T cells in vivo. Importantly, we confirmed the successful delivery of the inhibitor to CD4 + T cells in relevant draining lymph nodes including liver-draining lymph nodes, mesenteric lymph nodes as well as pancreatic lymph nodes and directly in pancreas-residing CD4 + T cells as assessed after 4 hours (Figure 5a for the reviewers / Supplementary Figure 10a in the revised manuscript) and 24 hours (Figure 5b for the reviewers / Supplementary Figure 10b in the revised manuscript).

Figure 5 for the reviewers. miR142-3p inhibitor accumulates in CD4 + T cells in vivo. (a)
Representative histograms showing the accumulation of the FAM-labelled miR142-3p inhibitor in CD4 + T cells isolated from relevant lymph nodes and the pancreas 4 hours after inhibitor application.  In addition and to confirm that the observed effects of miR142-3p inhibitor application on Treg induction, stability and Tet2 in NOD mice in vivo were directly mediated by reduced miR142-3p activity, we now performed novel in vivo experiments and applied the miR142-3p inhibitor to miR142 -/mice. As expected, the inhibitor had no effect on Treg frequency ( Figure 7a   These results clearly indicate that the effect observed in vitro and in vivo are due to the specific inhibition of miR142-3p by means of the miR142-3p inhibitor. Altogether we provide compelling evidence for a miR142-3p/Tet2/Foxp3 axis in murine and human CD4 + T cells that during islet autoimmunity interferes with the efficient induction of Tregs and leads to impairments in Treg stability. Importantly, we strengthen the claims that Tet2 is a direct target of miR142-3p, and that inhibitor delivery in vivo and in vitro is sufficiently robust to account for the observed effects.

Reviewer #3 (Remarks to the Author):
The authors have adequately addressed all of the concerns raised in my initial review. The paper is of sufficient scope and quality to be published at Nature Communications.

Response:
We thank the reviewer and are delighted that we have addressed his/her concerns satisfyingly and that the reviewer finds our revised manuscript of sufficient scope and quality to be The authors decline to provide further analysis of the HITS-CLIP data or to discuss specific criticisms of the sparse and possibly jackpotted data at the Tet2 3' UTR site featured in the paper. In the current form, the manuscript does not provide sufficient information to support the inclusion of the HITS-CLIP data.
The authors provide some additional experiments that are consistent with their assumption that in vivo delivered antagonists act in T cells to functionally inhibit miR-142 in those cells with downstream biological consequences. However, there plausible alternative explanations for these observations. The authors decline to directly test the activity of miR-142 in T cells in vivo (or even in vitro). Cited prior publications that used this technique and suggested direct effects in T cells also failed to provide this absolutely critical validatory data.
More detailed responses to the authors's rebuttal is provided in the attached file (all new material is in black, bold print).

Reviewer #3 (Remarks to the Author):
Excellent job, all remaining issues from reviewer 2 dealt with. Adrian Liston.

Results: "Moreover, a miR142-3p mimic induced a significant reduction in luciferase activity in HEK-293 cells transfected with a wildtype TET2 3'UTR reporter construct
while there was no effect of the mimic in cells transfected with a reporter construct containing the TET2 3'UTR with mutated miR142-3p binding sites (Fig. 5e,   Supplementary Fig. 10a)." Pages 11 and 12 of the manuscript.
In addition, a more detailed response to the additional comments of reviewer #2, including the references to the respective figures and sections of the updated manuscript, is provided below.

Reviewer #3 (Remarks to the Author):
Excellent job, all remaining issues from reviewer 2 dealt with. Adrian Liston.

Response for revision 3:
We thank the reviewer for his comment and are delighted that, in his opinion, we have addressed all concerns satisfyingly and that the reviewer finds our revised manuscript of sufficient scope and quality to be published at Nature Communications.

Additional comments of Reviewer #2
Comments

Response of authors provided in Revision 2:
We thank the reviewer for the positive comments and constructive criticism. In response to the reviewer's remaining comments and questions, which were very helpful for this revision, we here provide additional information for clarification as requested. Moreover, to further strengthen the main claims of our manuscript: that Tet2 is a direct target of miR142-3p, and that inhibitor delivery in vivo and in vitro is sufficiently robust to account for the observed effects, we directly addressed these comments of the reviewer by performing a series of additional novel in vitro as well as in vivo experiments.
Evidence for miR-142 targeting of Tet2 remains unconvincing, and the HITS-CLIP data do not appear robust. After considerable effort, I was able to find the location of the 5 apparently identical miR-142 chimeric reads from the HITS-CLIP dataset. unnamed regions with miRanda predicted biding sites.

Response of authors for revision 2:
We thank the reviewer for his comments. Our understanding is that in the view of reviewer 2 the HITS-CLIP data and the quality of the accompanied analyses need further clarification (1) and that within the HITS-CLIP data analysis the evidence for miR142-3p targeting of Tet2 is not fully convincing (2). Please find our responses to these concerns detailed below: (1) While we understand the reviewer's concern regarding the absence of a predicted miR142-3p binding site in the Tet2 3'UTR, we would like to respectfully draw the attention of the reviewer to Supplementary Fig. 6, showing the quality of the chimeric pairing by measuring the distribution of chimeric reads relative to mirRandapredicted miRNA binding sites for a set of the most frequent miRNAs. The Supplementary Fig. 6 indicates that the chimeras generally agree well with miRanda predictions, however, we also found a few miRNAs where the distribution of chimeric reads did not correlate with predicted binding sites. To the best of our knowledge miRNA target prediction tools first of all rely on the well characterized 6 nucleotide  Fig. 3c and d).

Response of authors for revision 3:
We thank the reviewer for this comment and apologize for any lack of clarity.
The application of the miR142-3p mimic to 3T3 fibroblasts was added in revision 1, the results can be found on page 12 of the manuscript and in Figure 5f.
The miR142 -/mice experiments, showing increased Tet2 mRNA and protein abundance following TCR stimulation, compared to miR142 +/+ mice were newly added in revision 2, the results can be found on page 12 of the manuscript and in Figure 5 g and h.
The experiments showing no effect of the miR142-3p inhibitor on Tet2, Tgfbr1 or ATG16L1 expression in miR142 -/mice were added in revision 2, the results can be found on pages 12 and 13 of the manuscript and in Supplementary Figure 10

Response of authors for revision 3:
We thank the reviewer for this comment and would also like to use this comment to respectfully disagree. As described above and in the reviewer's comments in revision 2 the reviewer had suggested using an activity sensor to test the activity of miR142-3p in T cells. According to his  We would like to emphasize that the luciferase reporter plasmid used in this activity sensor experiment does not contain the Tet2 3'UTR. We delivered, by plasmid transfection, a luciferase reporter gene with an artificial 3'UTR containing the perfect complementary sequence to miR142-3p, as suggested by the reviewer in this revision and revision 2. We used this sensor plasmid in Jurkat T cells to validate the functional blocking of miR142-3p by the inhibitor directly in T cells.
Taken together, we show the successful miRNA uptake in CD4 + T cells including intracellular co-localization of the nanoparticles and the fluorescently-labeled miRNA ( Figure 1 for the reviewers). Moreover, we now demonstrate reduced miR142-3p activity in T cells by delivery of the miR142-3p inhibitor compared with a control inhibitor using experiments with a miR142-3p activity sensor as suggested ( Figure 2 for the reviewers / Supplementary Figure 3b in the revised manuscript).

miR142-3p inhibitor delivery in vivo
For the inhibition of miR142-3p in vivo we used an LNA in vivo miRNA inhibitor which has been shown to accumulate in a broad range of tissues and to efficiently inhibit its targets in several independent studies 13,14 . Specifically, important studies using LNA miRNA inhibitors to successfully inhibit miRNAs in vivo, among others in T cells, are listed in Table 1 and an even more extensive list of studies can be found here: http://www.exiqon.com/ls/Documents/Scientific/mirna-inhibition-publications.pdf. [REDACTED] We now performed additional novel experiments using a fluorescently-labeled miR142-3p inhibitor in order to provide further experimental evidence for the delivery of functional miR142-3p inhibitors specifically to CD4 + T cells in vivo. Importantly, we confirmed the successful delivery of the inhibitor to CD4 + T cells in relevant draining lymph nodes including liver-draining lymph nodes, mesenteric lymph nodes as well as pancreatic lymph nodes and directly in pancreas-residing CD4 + T cells as    In addition and to confirm that the observed effects of miR142-3p inhibitor application on Treg induction, stability and Tet2 in NOD mice in vivo were directly mediated by reduced miR142-3p activity, we now performed novel in vivo experiments and applied the miR142-3p inhibitor to miR142 -/mice. As expected, the inhibitor had no  Here we wish to respectfully highlight that all the above mentioned analyses ( Figure 6 for the reviewers and Figure 7 for the reviewers of revision 2, added in revision 2) were all performed in CD4 + T cells and the same applies for the in vivo experiments.
Therefore we would like to highlight that these experiments provide compelling evidence for a direct effect of the inhibitor in T cells.
Response of authors for revision 2: These results clearly indicate that the effect observed in vitro and in vivo are due to the specific inhibition of miR142-3p by means of the miR142-3p inhibitor. Altogether we provide compelling evidence for a miR142-3p/Tet2/Foxp3 axis in murine and human CD4 + T cells that during islet autoimmunity interferes with the efficient induction of Tregs and leads to impairments in Treg stability. Importantly, we strengthen the claims that Tet2 is a direct target of miR142-3p, and that inhibitor delivery in vivo and in vitro is sufficiently robust to account for the observed effects. Response of authors for revision 3: As described above in more detail, we provided an extensive set of novel experiments to validate the effect of the miR142-3p inhibitor. These additional data have been included in response to revision 1-3.

Response
We confirmed its functionality in vitro using a miR142-3p activity sensor as requested by the reviewer (page 7 of the manuscript, Supplementary Figure 4b). We showed that the miR142-3p inhibitor increased the abundance of validated miR142-3p My prior concerns about this manuscript were focused on two major issues: The evidence for miR-142 targeting of Tet2, and the evidence that in vivo treatments act through miR-142 in T cells.
The first of these concerns has been adequately addressed by removal of the HITS-CLIP crosslinking data that seemed to suggest miR-142 interaction with a non-canonical binding site. New luciferase assays added in this revision indicate that two different canonical predicted miR-142 binding sites can mediate miR-142 regulation of the Tet2 3'UTR. I applaud the authors persistence in investigating how miR-142 affects Tet2 expression.
On the second point, the authors and I will have to agree to disagree. I eagerly await a clear demonstration that miRNA inhibitors can be effectively delivered to T cells and alter miRNA activity in vivo. However, I don't dispute that the reported treatments had the described effects in vivo, be it through inhibition of miRNA activity in T cells or other cells, or through some other unexpected mechanism.
Overall, the revised manuscript will be of significant value to the scientific community.

Nature Communications Manuscript number NCOMMS-18-24478C
miRNA142-3p targets Tet2 and impairs Treg differentiation and stability in models of type 1 diabetes

REVIEWERS' COMMENTS:
Reviewer #2 (Remarks to the Author): My prior concerns about this manuscript were focused on two major issues: The evidence for miR-142 targeting of Tet2, and the evidence that in vivo treatments act through miR-142 in T cells. Response: We thank the reviewer for his comments and constructive criticism during the previous rounds of revision, which were very helpful to improve the quality of the manuscript. We are pleased that we have addressed his concerns sufficiently as outlined below.
The first of these concerns has been adequately addressed by removal of the HITS-CLIP crosslinking data that seemed to suggest miR-142 interaction with a non-canonical binding site. New luciferase assays added in this revision indicate that two different canonical predicted miR-142 binding sites can mediate miR-142 regulation of the Tet2 3'UTR. I applaud the authors persistence in investigating how miR-142 affects Tet2 expression. Response: We thank the reviewer for his comment and are delighted that we adequately addressed his concerns regarding the miR142-3p -Tet2 -target relationship. As detailed in revision #3 we removed the chimeric reads part of the HITS-CLIP analysis and now focused on molecular and cellular approaches to confirm Tet2 as a direct target of miR142-3p. We are happy that the reviewer agrees that the newly added luciferase assay data convincingly confirm Tet2 as a direct target of miR142-3p in line with the two predicted binding sites mediating regulation of Tet2 by miR142-3p.
On the second point, the authors and I will have to agree to disagree. I eagerly await a clear demonstration that miRNA inhibitors can be effectively delivered to T cells and alter miRNA activity in vivo. However, I don't dispute that the reported treatments had the described effects in vivo, be it through inhibition of miRNA activity in T cells or other cells, or through some other unexpected mechanism. Response: We thank the reviewer for this comment and are glad that we agree on the in vivo effects of the miR142-3p inhibitor, described in our study. Regarding the delivery and efficacy, we would like to refer to the extensive set of experiments we had provided in revisions 1 and 2 in order to strengthen the successful inhibitor delivery in vitro and in vivo. Using the FAM-labeled LNA inhibitor, we demonstrated the successful in vivo delivery of the miR142-3p inhibitor to CD4 + T cells in relevant draining lymph nodes including liver-draining lymph nodes, mesenteric lymph nodes as well as pancreatic lymph nodes and directly in pancreas-residing CD4 + T cells. In addition, we clearly showed that the miR142-3p inhibitor blocks miR142-3p activity in vitro using various experimental approaches, including the suggested miR142-3p activity sensor and the increased abundance of two validated miR142-3p target genes. These results, in addition to the confirmed localization of the inhibitor to CD4 + T cells in vivo and the likewise increased expression of a miR142-3p target gene in vivo, provides compelling evidence for the functionality of the miR142-3p inhibitor both in vitro and in vivo.