Scrt1, a transcriptional regulator of β-cell proliferation identified by differential chromatin accessibility during islet maturation

Glucose-induced insulin secretion, a hallmark of mature β-cells, is achieved after birth and is preceded by a phase of intense proliferation. These events occurring in the neonatal period are decisive for establishing an appropriate functional β-cell mass that provides the required insulin throughout life. However, key regulators of gene expression involved in functional maturation of β-cells remain to be elucidated. Here, we addressed this issue by mapping open chromatin regions in newborn versus adult rat islets using the ATAC-seq assay. We obtained a genome-wide picture of chromatin accessible sites (~ 100,000) among which 20% were differentially accessible during maturation. An enrichment analysis of transcription factor binding sites identified a group of transcription factors that could explain these changes. Among them, Scrt1 was found to act as a transcriptional repressor and to control β-cell proliferation. Interestingly, Scrt1 expression was controlled by the transcriptional repressor RE-1 silencing transcription factor (REST) and was increased in an in vitro reprogramming system of pancreatic exocrine cells to β-like cells. Overall, this study led to the identification of several known and unforeseen key transcriptional events occurring during β-cell maturation. These findings will help defining new strategies to induce the functional maturation of surrogate insulin-producing cells.


Results
To study the transcriptional regulation of pancreatic islet maturation, we performed high throughput sequencing of transposase-accessible chromatin of newborn and adult rat islets (Fig. 1a). Post-natal β-cell maturation in rat has been shown to occur on a slightly different time frame compared to mice 3,4,16 . Consistent with the data in the literature we previously observed that P10 rat β-cells have a high proliferative capacity compared to adult mature β-cells but are unable to secrete insulin in response to glucose 1 . Therefore, we elected to compare chromatin accessibility in 10 days-old rat islets (P10) and in adult rat islets. Bioinformatic methods for reads alignment, quality control, peak detection, differential accessibility analysis and motif finding were applied. For example, the Magnesium Transporter 2 (Mrs2) locus depicts two prominent ATAC-seq signal peaks, and the 3′ end peak shows an important increase of accessibility after the maturation process (Fig. 1b). Quality control of the samples (Table 1) showed an average of 370 million reads sequenced with 94.3% of reads with a MAPQ score above 30. The analysis of fragment size distribution showed the expected profile of ATAC-seq (Supplementary Fig. S1a) with usual oscillations due to the presence of nucleosomes. In addition, adult and P10 samples were well clustered, when evaluated by the correlation between samples ( Supplementary Fig. S1b) or using the first PCA component (Supplementary Fig. S1c). To detect accessible sites (ACS), we performed a peak calling using MACS2 17 (Methods), leading to the detection of ~ 102,000 ACS (Supporting Information 1). These sites were quantified in each sample separately and annotated with the closest transcription start site (TSS). Moreover, a differential accessibility analysis was performed using EdgeR 18 (Fig. 1c, top), and the values of the analysis reported in the source data (1). ACS were divided in three groups: Stable, significantly more accessible in P10 (Down) and significantly more accessible in adults (Up), with p-value < 0.05 and FDR < 0.2. About 20% of the ACS showed differential accessibility upon maturation with 11.8% of down ACS and 7.1% of up ACS (Fig. 1c, bottom).
Promoter and distal ACS depicted distinct accessibility patterns along islet maturation. To assess the impact of the ACS location in respect to the closest gene, we used ChIPseeker 19 to annotate our ACS and 100,000 randomly selected sites ( Supplementary Fig. S1d). ACS were enriched for exonic and intronic sequences and transcript start sites as opposed to random sites ( Supplementary Fig. S1e). Interestingly, ACS more accessible in adults (up), were enriched in distal sites and intron as compared to ACS less accessible in adults (down), which were enriched in promoters. We observed that 2/3 of the ACSs were distal sites and about 10% were located in the TSS or proximal class.

Identification of transcriptional regulators and chromatin remodelers of pancreatic islet maturation.
Accessibility analysis enables the detection of TFs or TFBS affecting the chromatin state and the transcription of nearby genes. In order to decipher the combinatorial code of transcription factor binding sites Figure 1. ATAC-seq successfully identified accessible sites (ACS) and transcription factor binding sites (TFBS) motif associated with pancreatic islet maturation. (a) Summary of the experimental design. Nuclei were extracted from the islets of 3 adult rats and 3 litters of rat pups at postnatal day 10 (P10) to perform the Tn5 reaction as described in 51 and prepare the library for sequencing. The computational pipeline involved a quality control of the sequencing data followed by read alignment to the rat reference genome (Rn5 assembly). ACS were identified using the peak calling tool MACS2 17 and quantified for each sample separately. The ACS sequences were scanned using FIMO 20 and analyzed to identify TFBS motifs that are implicated in the islet maturation process and in related pathways (See methods). (b) Example of identified ACS. The ACS nearby Mrs2 transcription end site is significantly higher in adult rats. This ACS contains several TFBS motifs. (c) Differential analysis of Trn5 integrations in accessible sites. On the top, the volcano plot representation of the log 2 fold-change of Trn5 integration between P10 and adult rat islets in the x-axis, and the FDR adjusted p-value in the y-axis. ACS more accessible in adults are represented in green (Up), those more accessible in P10 in red (Down), and those remaining stable in blue (p-value < 0.05, FDR < 0.2, n = 3). Below, the barplot of the number of ACS changing along postnatal maturation and the donut plot of the percentage of Up, Down and stable ACS.  www.nature.com/scientificreports/ that allows islet maturation, we scanned the sequence of each ACS using FIMO 20 together with the position weight matrices of Jaspar 2016 21 . With these sequence scans we could perform a motif set enrichment analysis for the accessibility in P10 or in adult islet cells using the FGSEA algorithm 22 (Supplementary Table 2, Fig. 1d). Thus, using the set of significantly changing ACS and their respective accessibility log 2 Fold Change, we were able to identify TFBS motifs that were either enriched in P10 or in adult rat islets. Several TFBS, previously implicated in islet maturation, were significantly enriched, such as MAF, FOX, FOS/JUN, NRF, and E2F [23][24][25][26][27][28] . Moreover, several motifs recognized by transcriptional repressors and insulators such as SCRT1 or CTCF were also enriched in P10 islets. To confirm these results and detect additional motifs playing a role in islet maturation, we applied a penalized linear model GLMnet 29 to all ACS with the matrix of motif match as predictors and the log 2 fold-change as output vector. With this method, an activity (β) for each motif was computed (Supplementary Table 3). This allowed to confirm most of the hits discovered using the FGSEA and to detect additional TFBS motifs such as RFX, SREB, NKX6, REL, MEIS, and TEAD3.
ACS significantly affected were located near genes involved in islet maturation. Next, we investigated if ACS changing in P10 versus adult rat islets control the expression of genes displaying significant differences upon maturation 30 . Of the 19,311 ACS differentially accessible, 11,372 (~ 60%) were located in the vicinity of differentially expressed genes. These ACS were annotated as enhancers or repressors, depending whether their accessibility was respectively correlated or anti-correlated with the expression changes in the nearby gene (Supporting Information 1). As previously described 30 , several KEGG pathways were enriched by analyzing significantly changing genes nearby differentially accessible ACS. For instance, insulin secretion, circadian rhythm, and calcium signalling were enriched in adult, while carbon metabolism, PI3-Akt, and prolif-   Table 3. qPCR Primer Sequences (mouse).   Table 4, Fig. 2a-c) near Syt4, Pax6 (two ACS), Mafb and NeuroD1 were cloned in a luciferase reporter construct driven by a minimal promoter. Interestingly, the inclusion of the ACS close to Syt4, MafB, and NeuroD1 resulted in an increase in luciferase activity, compared to the empty pGL3 vector. In addition, mRNA expression of NeuroD1 and MafB, tested by qPCR ( Fig. 2d), confirmed the higher expression of these genes in 10-day-old pups. MAFB was previously described as a regulator of the pancreatic β-cell function 32 and its expression to be decreased during β-cell maturation in mice 23 (reported in Supplementary Fig. S2b). Here we detected a cis-regulatory elements containing various TFBS motifs including Meis1,2 or Glis2 located more than 10Kbp away from the Mafb gene that was less accessible in adult rats ( Supplementary Fig. S2a). Thus, we concluded that the identified ACS are likely to be involved in transcriptional regulation of nearby genes.
Scrt1 represses β-cell proliferation. We next focused on Scrt1, a transcriptional repressor involved in neuroendocrine development [33][34][35] but whose functions in β-cells remain to be determined. Binding sites for this transcriptional repressor were highly enriched in the chromatin regions that close upon β-cell maturation. Thus, our hypothesis was that Scrt1 binding in those sites would close the chromatin. In agreement with the lower chromatin accessibility for Scrt1 binding sites, Scrt1 expression is increased in adult rat islets compared to immature islets ( Fig. 2e, Supplementary Fig. S2d). Interestingly looking into the data-set published by Qiu et al. 23 , Scrt1 expression was also found to be augmented during maturation of mouse β-cells ( Supplementary  Fig. S2c). Transfection of adult rat islet cells with a set of siRNAs targeting Scrt1 led to a decrease in the expression of the repressor of about 70%. (Fig. 2f). Scrt1 knockdown neither affected glucose-stimulated insulin secretion ( Fig. 2g) nor insulin content (Fig. 2h). Apoptosis measured by Tunel assay in both control condition or in response to pro-inflammatory cytokines was also not affected ( Identification of Scrt1 targets involved in maturation. As Scrt1 appears to regulate the proliferative capacity of β-cells, we next aimed at finding its targets. For this purpose, adult rat islet cells were FACS-sorted to separate α-cells and β-cells. We observed that α-cells and β-cells express Scrt1 at similar levels (Supplementary Fig. S4a). Subsequently, an RNA-seq on β-cells exposed to a control siRNA or to siScrt1 was performed. Differential expression analysis between siScrt1 and control samples revealed that 168 genes were significantly impacted by silencing Scrt1 with a FDR adjusted p-value below 0.05 (Fig. 3a). Of these 168 genes, 111 were down-regulated and 57 were up-regulated. The significantly changing genes were enriched with ACS containing Scrt1 motif in their surroundings (hypergeometric test p-value < 10 -6 ). We then looked at the log 2 Foldchange distribution of ACS located at the TSS, proximal or distal (Fig. 3b). We observed that genes having a Scrt1 motif containing ACS at the TSS or proximal were increased upon siScrt1 while a substantial part of the distal Scrt1 ACS were decreased. As expected, among the potential targets of Scrt1 we found genes related to proliferation such as Notch1, Parp16, Ppp3r1, Ppp2r1b and Ywhag (Fig. 3c). Moreover, some genes related to glucose signalling and GSIS such as Syt4, or to sphingolipid metabolism as Ugt8 were down-regulated when knocking down Scrt1. Then, we compared the set of genes affected by Scrt1 silencing and the ones differentially expressed upon maturation (in postnatal 10-day-old (P10) versus adult rat islets). A common set of 62 genes was found to change in both data sets with an FDR adjusted p-value below 0.05 (Fig. 3c, Supplementary Table 7). In addition, a gene ontology (GO term) enrichment analysis for biological processes revealed that autophagy and oxygen sensing are over represented in these 62 genes (Supplementary Table 8). Interestingly, a significant anti-correlation (Pearson's correlation test, p-value = 0.013) was observed between the fold-changes from the comparison between siScrt1 versus siCtl in adult rat β-cells and P10 versus adult islets. The opposite change in the expression of Nfatc1, Nfatc2, Notch1 and Syt4 in response to Scrt1 downregulation versus islet maturation was also confirmed (Fig. 3f,g, Supplementary Fig. S5). Next, we looked at the enriched gene ontologies using the FGSEA algorithm on the RNA-seq of siScrt1 and siCtrl. Cell proliferation and pattern specification process were among the highest hits (Supplementary Table 5, Fig. 3d). FAC-sorted adult rat β-cells were transfected with a control siRNA (siCtl) or siRNAs directed against Scrt1 (siScrt1). RNA extraction and library preparation for RNA-seq were performed 48 h post-transfection. (a) Volcano plot of gene expression changes induced by Scrt1 knockdown. SCRT1 putative target genes (defined based on the presence of SCRT1 motif in ACS in the proximity of the gene) significantly altered are indicated with red dots and labels. In the top left, normalized counts from the RNA-seq data of Scrt1 expression is represented with a barplot. (b) log 2 fold changes distribution of significantly altered genes by siScrt1 containing a SCRT1 motif in an ACS either at the TSS (± 1Kbp), proximal (1-10Kbp) or distal (more than 10Kbp) from the gene. (c) Scatter plot of log 2 fold changes of differentially expressed genes from adult/P10 measured by microarray (n = 3) versus log 2 fold changes of siScrt1/control measured by RNA-seq (n = 5). Several genes of interest are represented with their label. (d) Enrichment plot for two significant annotations using the FGSEA algorithm. (e) Scheme describing the siScrt1 hypothesis. Briefly, the siRNA for Scrt1 will reduce the mRNA level of the Scrt1 gene and consequently the level of the protein. Thus, lower binding of SCRT1 will occur at target sites that will affect proliferation and specialization of β-cells. qPCR confirmation of gene expression in (f) FAC-sorted β-cells transfected with siCtrl or siScrt1 or in (g) P10 versus adult rat islets. Student T test, *p < 0.05, **p < 0.01. See also Supplementary Fig. S5  www.nature.com/scientificreports/ Taken together our results suggest that Scrt1 act as a repressor that regulate proliferation through de-repression of genes at their TSS or in cis-regulatory elements, while specialization seems to occur through distinct mechanisms involving distal ACS (Fig. 3e).

Scrt1 is a target of REST and is induced in exocrine to β-cell reprogramming. To gain insights
in to the mechanisms underlying Scrt1 regulation during β-cell maturation, we analysed previously published ChIP-seq data for hallmark β-cell transcription factors in mouse islets [36][37][38] . Interestingly, the Scrt1 locus is bound by PDX1, FoxA2, NeuroD1, INSM1 and NKX6-1 (Fig. 4a). Moreover, this locus is marked by H3K4Me3 in islets 39 , suggesting that the Scrt1 gene is active in the endocrine pancreas. Since REST (RE-1 silencing transcription factor) was previously reported to be an important repressor of β-cell genes that is silenced in adult islets 15,40,41 , we also reanalysed the binding of this transcriptional repressor 42 to the Scrt1 locus. Interestingly, REST binds the Scrt1 locus in non-islet mouse and human cells (ES and Panc1 cells) (Fig. 4a, Supplementary  Fig. S6), suggestive of a REST-mediated repression of the Scrt1 gene. Given the identification of a REST binding site upstream of Scrt1 gene (Fig. 4a), and a specific expression of Scrt1 in endocrine and not in exocrine cells (Fig. 4c), as well as the recently published involvement of REST in acinar to β-cell reprogramming 84 , we sought to test the regulatory role of REST on Scrt1 expression, in this biologically relevant and well controlled reprogramming system.
To address the potential repression of Scrt1 by REST in pancreatic cells, a series of experiments were performed on mouse exocrine, β-, and reprogrammed β-like cells, as described in Fig. 4b. First, analysis of previously published transcriptomic data 15 , shows that Scrt1 (Fig. 4c) and Rest expression (Fig. 4d) is inversely correlated in exocrine cells and in β-cells. A high expression of Scrt1 in β-cells was observed, while Rest was expressed only in exocrine cells. Next REST was overexpressed in the mouse insulinoma cell line MIN6 and gene expression was analysed by qPCR for Rest, Ins2, Pdx1, Snap25 and Scrt1 three days post infection (Fig. 4e). While Ins2 and Pdx1 (not direct targets of REST) were unaffected by Rest overexpression, Snap25 (a known target of REST 40 ) and Scrt1 expression were significantly reduced. Opposite results were obtained in primary exocrine cells infected with Ad-mCherry control or Ad-DN-REST virus (encoding a dominant negative REST protein 43 ). At 5 days postinfection, Scrt1 expression was significantly higher when Rest was silenced using the dominant negative mutant (Fig. 4f). These results are consistent with a REST-mediated repression of the Scrt1 gene. Next, we examined Scrt1 and Rest expression levels in a model of β-like cells undergoing reprogramming from primary exocrine cells 3 and 6 days post-infection 6,15 . A continuous accumulation of Scrt1 mRNA and a parallel reduction in Rest were observed (Fig. 4g,h). Lastly, Rest overexpression in FACS-sorted reprogrammed cells resulted in a marked decrease in Scrt1 expression (Fig. 4i). Taken together, our results demonstrate that Scrt1 is controlled by REST. Thus, REST represses Scrt1 in exocrine cells and Scrt1 increases during reprogramming in parallel with a concomitant decline in Rest expression.

Discussion
Postnatal islet maturation is a critical process to achieve proper β-cell function. Immediately after birth, β-cells are not fully functional and have to undergo a major gene reprogramming to acquire the ability to secrete adequate amounts of insulin in response to glucose 44 . Our group and others 7,23,30,45 have shown a large-scale rewiring of transcriptional programs occurring during the neonatal period. However, little is known on cis-regulation of gene expression at the chromatin level before and after weaning. Chromatin accessibility of human islets on a genome-wide scale has been previously produced using FAIRE-seq 31,46 . More recently, several studies took advantage of ATAC-seq together with GWAS to identify causal variants of T2D in cis-regulatory elements in human [47][48][49][50] . Another study identified cell-type specific accessible sites and transcription factor binding sites in α, β and acinar cells 12 . However, the transcriptional regulation of postnatal islet maturation at the chromatin level has not been reported so far. In this project, we employed ATAC-seq 13,51 to produce a global map of accessible  (g,h). Primary β-cells are the GFP + FAC-sorted fraction of islets isolated from MIP-GFP mice. Results are mean normalized counts ± SD, and ***p < 0.001, n = 4. (e) Min6 cells were infected with Ad-mCherry control virus or Ad-hREST and 3 days post infection gene expression was measured by qPCR and normalized by housekeeping gene (Tbp). Results are presented as relative expression versus Ad-mCherry control and are mean ± SD; ** p < 0.001 on t-test, n = 4. (f) Primary exocrine cells were infected with Ad-mCherry control or Ad-DN-REST viruses and 5 days post-infection Scrt1 expression was analyzed by qPCR and normalized by Tbp. Results are presented as relative expression levels versus Ad-DN-REST group and are mean ± SD; n = 4. Normalized counts from RNA-seq data showing Scrt1 (g) and Rest (h) expression in reprogrammed β-like cells from primary exocrine cells. Primary exocrine cells isolated from MIP-GFP mice were infected with Ad-M3-mCherry virus. Day 3 and 6 Cherry/GFP fractions were FAC-sorted as indicated. Cherry + represents the M3-infected fraction and the GFP + fraction enriched for reprogrammed β-like cells. Results are mean normalized counts ± SD, *p < 0.05, n = 4. (i) FACsorted day 5 reprogrammed exocrine cells expressing M3-mCherry and LacZ or REST. mCherry fractions were FAC-sorted and Scrt1 expression was analyzed by qPCR and normalized by Tbp. Results are presented as relative expression levels versus mCherry + fraction of Ad-mCherry + Ad-LacZ control virus group, and are mean ± SD; *** p < 0.001, n = 4. See also Supplementary Fig. S6 and Supplementary Table 9. www.nature.com/scientificreports/ sites (ACS) in the islets of 10-day-old pups and in adult rats. This permitted to detect more than 100,000 ACS, among which about 20% were differentially accessible before and after the functional maturation of β-cells. Interestingly, the ACS with the most significant p-value and a large fold-change was located on the 3′UTR of the Mrs2 gene. This gene is encoding a magnesium transporter at the surface of the mitochondria 52 . Genetic variants in this magnesium-related ion channel were previously associated with type 2 diabetes and pancreatic cancer 53,54 .
Using two different computational approaches and the Jaspar PWM database, we could find known and unforeseen transcriptional regulators potentially involved in the maturation process. We identified many DNAbinding proteins affecting chromatin accessibility, including MAF, FOX, FOS/JUN, NRF, E2F, CTCF, RFX, SREB, NKX6, REL, MEIS, TEAD and SCRT1. Several of these transcription factors have been already implicated in pancreas development and postnatal islet maturation 55 . For instance, E2F1 is an established cell survival and proliferation activator in β-cells 28 and we have recently shown that this transcription factor controls the expression of the long non-coding RNA H19 and is profoundly down-regulated during the postnatal period. Moreover, we obtained evidence suggesting that E2F1 and H19 may contribute to the decrease in β-cell mass in the maternal low protein diet offspring model 45 .
Next, we integrated these accessibility maps with mRNA micro-array data from 30,45 in the same model and we found that ~ 60% of the differentially accessible sites were nearby differentially expressed genes, suggesting that these cis-regulatory elements are involved in the transcriptional regulation of gene expression. Several pathways previously related to the maturation process such as insulin secretion or circadian rhythms are under the control of these accessible cis-regulatory elements. For instance, several components of the core clock are not rhythmic in 10 days old pups but are consistently oscillating in β-cells after weaning 56 . It has been recently demonstrated that circadian rhythms trigger islet maturation via clock-controlled metabolic cycles 27 . These cisregulatory elements may be controlled by transcriptional enhancers or repressors. Indeed, we confirmed that 3 out of the 5 tested ACS that are located nearby genes important for proper β-cell function had a significant enhancer activity. These ACS were located nearby Syt4, Neurod1, and Mafb. We showed that an enhancer site and the promoter of Mafb depicted a decreased accessibility along maturation. Mafb is required at the early stage of β-cells maturation and decreases in later stages 32 . We observed that an enhancer near Neurod1 was significantly more accessible and Neurod1 expression was higher in mature β-cells. Another group showed that overexpression of Mafa results in increased Neurod1 expression 57 , and that Mafa is crucial to induce glucose-responsive insulin secretion in neonatal rat β-cells. We showed that Syt4 is a potential target of NFAT, FOX and RFX TFs family ( Supplementary Fig. S5). Syt4 expression was correlated with its enhancer accessibility and was higher in fully mature β-cell. Syt4 is known to modulate Ca 2+ sensitivity of insulin granules and consequently enhance GSIS 58 .
One of the identified transcriptional regulators, SCRT1, is of particular interest. Scrt1 was previously shown to be implicated in brain development 33,34 but its role in β-cells remained to be determined. While this manuscript was under revision, Scrt1 was suggested to act as a regulator of insulin expression and secretion in response to glucose in combination with a cAMP raising agent (IBMX) 59 . Here, downregulation of Scrt1 in adult rat islets did not impact insulin release in response to only high concentration of glucose but induced β-cell proliferation, suggesting that reduced expression of Scrt1 in P10 islets may favour the acquisition of an appropriate β-cell mass during the neonatal period. Indeed, transcripts related to proliferation were significantly impacted by the downregulation of Scrt1 in FACS-sorted adult β-cells. Interestingly, a significant increase of Scrt1 in adult islet cells was also reported in a recent study analyzing age-dependent gene expression and chromatin changes in human 60 . RNA-seq analyses suggest that Scrt1 affects Syt4, Notch1 and Nfatc1 and Nfatc2 gene expression and consequently may control the calcineurin/NFAT pathway, an important regulator of β-cell growth and function 61,62 . Functional enrichment analysis showed that Scrt1 also influences the expression of genes related to oxygen sensing and autophagy. Accordingly, previous studies have demonstrated a role for the hypoxia-inducible factor HIF1a in normal β-cell function 63 , and altered β-cell autophagy in human T2DM 64 .
Intriguingly, Ackermann and others 5,12 pointed out that many poised genes in α cells contain a signature of functional β-cells and, consequently, could be of use for α-to-β cell reprogramming. This suggests plasticity and specialization of the different cell types along maturation and rewiring of transcriptional programs at the chromatin level. In fact, reprogramming of acinar cells into insulin-producing cells has been accomplished using adenoviral gene delivery of PDX1, NGN3 and MafA 6 . A subsequent study characterized these changes using RNA-seq and ChIP-qPCR and revealed that loss of Rest combined with Pdx1 expression leads to activation of endocrine genes and correlates with epigenetic modifications of local chromatin 15 . In this study, we provide evidence indicating that Scrt1 is a direct target of REST in different mouse and human cell models. Moreover, overexpression of Rest inhibits Scrt1 expression in β-cells and in β-like cells reprogrammed from exocrine cells. Islet plasticity is of interest for the design of future treatments of diabetes through the production of mature surrogate insulin-producing cells 65 . Thus, understanding the role of Scrt1 in the reprogramming process will be important to improve the generation of mature surrogate insulin-producing cells. REST downregulation has been suggested to be involved in postnatal β-cell maturation in response to thyroid hormones 57 and it is tempting to speculate that Scrt1 derepression might also contribute to this process.
In this study, ATAC-seq and the mRNA microarray analyses have been performed in whole islets and not in FACS-sorted cells or in single cells. Several recent studies have demonstrated a large heterogeneity of gene expression at a single cell level using smFISH or single cell RNA-seq 23,66-68 . Taking advantage of the heterogeneity of gene expression together with single cell ATAC-seq could shed more light on the transcriptional control of islet maturation and the different cell types involved.
Overall, we produced a high-resolution map of chromatin accessible sites in islets of 10-day-old pups and adult rats (Fig. 5a). These genome wide accessibility maps are an important resource to study cis-regulation of gene expression along islet cell maturation. Using these maps, we discovered a new important transcriptional repressor implicated in the maturation process, namely Scrt1, which controls β-cell proliferation and function. Scrt1 is controlled by REST and its expression is increased along acinar to β-cell reprogramming (Fig. 5b). Islet isolation, dispersion and sorting. Islets were isolated by collagenase digestion of the pancreas 69 . In more details, P5 and P10 pancreases of 3-5 pups were collected, pooled and digested by hand shaking in 3 ml of collagenase P solution (Sigma) at 1 mg/ml for 3-5 min, while 3 or 10 ml of collagenase P solution were injected in the pancreatic duct of P20-23 pup and adult rats, respectively, followed by 8 and 15 min digestion in a warm bath. After digestion, islets were separated from the exocrine tissue using an histopaque density gradient followed by hand-picking and were incubated for 2 h in RPMI 1640 GlutaMAX medium (Invitrogen) containing 11 mM glucose and 2 mM l-glutamine and supplemented with 10% fetal calf serum (Gibco), 10 mM Hepes pH 7.4, 1 mM sodium pyruvate, 100 μg/mL streptomycin and 100 IU/mL penicillin. Dissociated islet cells were obtained by incubating the islets in Ca 2+ /Mg 2+ free phosphate buffered saline, 3 mM EGTA and 0.002% trypsin for 5 min at 37 °C. For some experiments, islet cells were separated by Fluorescence-Activated Cell Sorting (FACS) based on β-cell autofluorescence, as previously described 70,71 . Sorted islet cells were seeded on plastic dishes coated with extracellular matrix secreted by 804 G rat bladder cancer cells (804 G ECM) 72 . Enrichment of α-and β-cells was evaluated by double immunofluorescence staining using polyclonal guinea pig anti-insulin In the present study we investigated the chromatin accessibility along the post-natal pancreatic islet maturation process. (a) SCRT1 motif is enriched in ACS that are closing upon maturation. We found that Scrt1 is a direct target of REST and that the Scrt1 motif is present in TSS/proximal sites as well as distal sites. The inhibition of Scrt1 lead to increased proliferation and has additional effects on other processes such as specialization. (b) In an in vitro reprogramming system, Scrt1 expression raises after PNM infection and is anti-correlated with REST levels. www.nature.com/scientificreports/ (dilution 1:40, PA1-26938 Invitrogen) and polyclonal mouse anti-glucagon (dilution 1:1000, Abcam Ab10988) antibodies, followed by goat anti-guinea pig Alexa-Fluor-488 and goat anti-mouse Alexa-Fluor-555 (diluted 1:400, Thermofisher A11073 and A21422, respectively) secondary antibodies. On average, β-cell fractions contained 99.1 ± 0.9% insulin-positive cells and 0.6 ± 0.6% glucagon-positive cells and α-cell-enriched fractions contained 10.6 ± 8.2% insulin-positive cells and 88.8 ± 8.2% glucagon-positive cells.
ATAC-seq sample preparation. ATAC-seq libraries were prepared as previously described 51 using dissociated islet cells from 3 adult male rats and from a mix of few P10 pups of 3 different litters. Briefly, 100′000 islet cells were resuspended in 50 μl of cold lysis buffer (10 mM Tris-HCl pH7.4, 10 mM NaCl, 3 mM MgCl2 and 0.1% IGEPAL CA-630) and centrifuged at 500 g for 10 min at 4 °C. The pellet was resuspended in the transposase reaction mix (Nextera kit, Illumina). The transposition reaction was performed at 37 °C for 30 min and was followed by purification of the samples using the Qiagen MinElute PCR purification kit (Qiagen). Transposed DNA fragments were amplified for 11 cycles using the NEBnext high-fidelity PCR master mix and the Ad1_noMX and Ad2.1-2.6 barcoded primers from 51 . Amplified libraries were purified with AMPure XP beads (Beckman Coulter) to remove contaminating primer dimers. Library quality was assessed using the Fragment Analyzer and quantitated using Qubit. All libraries were sequenced on Illumina HiSeq 2500 using 100 bp paired-end reads.
ATAC-seq data quality control and analysis. Fastq Table 6). These microarray data are available in the GEO database under the accession number GSE106919.
Luciferase assay. Luciferase activities were measured in INS 832/13 cells using the Dual-Luciferase Reporter Assay System (Promega). Firefly luciferase activity was normalized to Renilla luciferase to minimize experimental variabilities. Experiments were performed in triplicates.
RNA extraction, quantification and sequencing. RNA was extracted using miRNeasy micro kit (Qiagen) followed by DNase treatment (Promega). Gene expression levels were determined by qPCR using miScript II RT and SYBR Green PCR kits (Qiagen) and results were normalized to the housekeeping gene Hprt1. Data were analyzed using the 2 −ΔΔC(T) method. Primer sequences are provided in Table 2. For mRNA-sequencing, the RNA was converted into a sequencing library using the Illumina TruSeq RNA-sequencing kit and standard Illumina protocols. Single-end, 151 nt long reads were obtained using a HiSeq 4000 instrument. Reads were aligned to the transcriptome (Rnor6) with Kallisto 82 and presented as transcripts per million (TPM) and EST pseudo counts (Suppementary table 5). Subsequently, differential expression was calculated using Sleuth 83 . Finally, bio- www.nature.com/scientificreports/ logical function and pathway analyses were performed using Cluster profiler 78 . RNA-seq raw data were deposited in the GEO database under the accession number GSE130651.
Insulin secretion and content. Transfected