The long noncoding RNA Paupar modulates PAX6 regulatory activities to promote alpha cell development and function

SUMMARY

Many studies have highlighted the role of dysregulated glucagon secretion in the etiology of hyperglycemia and diabetes. Accordingly, understanding the mechanisms underlying pancreatic islet α cell development and function has important implications for the discovery of new therapies for diabetes. In this study, comparative transcriptome analyses between embryonic mouse pancreas and adult mouse islets identified several pancreatic lncRNAs that lie in close proximity to essential pancreatic transcription factors, including the Pax6-associated lncRNA Paupar. We demonstrate that Paupar is enriched in glucagon-producing α cells where it promotes the alternative splicing of Pax6 to an isoform required for activation of essential α cell genes. Consistently, deletion of Paupar in mice resulted in dysregulation of PAX6 α cell target genes and corresponding α cell dysfunction, including blunted glucagon secretion. These findings illustrate a distinct mechanism by which a pancreatic lncRNA can coordinate glucose homeostasis by cell-specific regulation of a broadly expressed transcription factor.

Graphical Abstract

IN BRIEF

Singer et al. identify temporally regulated lncRNAs by comparing the transcriptomes of mouse embryonic pancreas to adult islets. One lncRNA, Paupar, is enriched in glucagon-producing alpha cells where it confers the cell specific regulatory function of the transcription factor, PAX6, to promote the formation and function of alpha cells.

INTRODUCTION

Type 1 and type 2 diabetes mellitus (T1D and T2D) are chronic conditions with genetic, immunological, and environmental etiologies that occur due to the failure of the pancreatic islets to maintain glycemic control. Blood glucose homeostasis results from the coordinated, but opposing action of two pancreatic islet-derived hormones, insulin and glucagon. Nutrient ingestion stimulates insulin secretion from islet β cells, which promotes glucose uptake and suppresses liver glucose production. Hypoglycemia stimulates glucagon secretion from islet α cells, which promotes glucose production and release from the liver. While the majority of diabetes treatment options have focused on enhancing β cell function to meet insulin demands, more recent work has shown that α cell dysfunction and hyperglucagonemia also contribute to disease pathophysiology (Unger and Cherrington, 2012; Brissova et al., 2018). Thus, a better understanding of the regulatory mechanisms required for the development and function of these highly-specialized islet endocrine cells will provide important information that could be applied to therapeutic treatments.

The adult endocrine cell populations are all derived from Neurogenin 3 (Neurog3)-expressing endocrine progenitor cells and include insulin-producing β cells, glucagon-producing α cells, somatostatin-producing δ cells, and pancreatic polypeptide-producing PP cells. A large number of transcriptional regulators that are essential for islet cell lineage decisions have been identified and characterized. Many of these factors, including Pdx1 (Offield et al., 1996; Gao et al., 2014), Nkx2-2 (Sussel et al., 1998; Gutiérrez et al., 2017; Churchill et al., 2017), NeuroD1 (Anderson et al., 2009; Naya et al., 1997), and Glis3 (Kang et al., 2016), are expressed in several progenitor and/or islet cell populations and are continuously required in the adult for the maintenance of mature islet cell identity and function (Pan and Wright, 2011; Talchai et al., 2012; Gutiérrez et al., 2017; Ediger et al., 2017). One of the more well-characterized broadly expressed islet transcriptional regulators is the paired and homeodomain transcription factor, PAX6 (St-Onge et al., 1997; Sander et al., 1997). Mice deleted for Pax6 in pancreatic progenitor cells are born with severely reduced numbers of α, β, and δ cells, suggesting that PAX6 is necessary for the development of multiple islet endocrine cell types (Ashery-Padan et al., 2004; Hart et al., 2013). Conditional deletion of Pax6 in mature α or β cells demonstrated that PAX6 is also required to maintain the identity and function of mature islet endocrine cells (Ahmad et al., 2015; Gosmain et al., 2012; Mitchell et al., 2017; Swisa et al., 2017). Furthermore, a recent study showed that within mouse β cells, PAX6 directly activates critical β cell genes and represses genes that specify the alternative islet endocrine cell lineages (Swisa et al., 2017). PAX6 also has essential functions in the eye (Shaham et al., 2012) and central nervous system (Manuel et al., 2015). Studies in corneal and epithelial cell lines have shown that splice variants of PAX6 bind distinct DNA motifs (Epstein et al., 1994; Chauhan et al., 2004), yet the molecular mechanism mediating the tissue-, cell-, and gene-specific regulatory functions of PAX6, and other pancreatic transcription factors, remain poorly understood.

While the majority of transcriptional regulatory proteins are expressed in many different cell and tissue types, a class of non-protein coding RNAs (ncRNAs), called long noncoding RNAs (lncRNAs), tend to be highly tissue and cell type specific and temporally regulated (Carninci et al., 2005; Derrien et al., 2012; Guttman et al., 2011), giving them the potential to confer cell type specificity on broadly expressed transcription factors. In the pancreas, thousands of lncRNAs have been identified in human islets (Moran et al., 2012; Akerman et al., 2017; Fadista et al., 2014; Li et al., 2014), purified human α and β cells (Nica et al., 2013; Bramswig et al., 2013), mouse islets (Ku et al., 2012; Motterle et al., 2017), and purified mouse α and β cells (Ku et al., 2012; Benner et al., 2014), where they exhibit properties consistent with functional genes (Singer and Sussel, 2018). Studies on individual islet lncRNAs have provided evidence that lncRNAs play a role in islet function, primarily through the regulation of essential islet transcription factors, such as HI-LNC25 on GLIS3 (Moran et al., 2012), βlinc1 on Nkx2-2 (Arnes et al., 2016), and PLUTO on PDX1 (Akerman et al., 2017). These findings highlight a role for lncRNAs in islet biology, and their highly restricted expression patterns further suggest that they can influence islet cell specific functions.

To identify developmentally regulated pancreatic lncRNAs, we conducted comparative transcriptome analyses of embryonic mouse pancreas and adult mouse islets and identified 572 dynamically expressed lncRNAs. Our analyses uncovered several pancreatic lncRNAs that lie in close proximity to essential pancreatic transcription factors, including the lncRNA Pax6 Upstream Antisense RNA (Paupar), which mapped near the Pax6 genomic locus in mice and humans (Vance et al., 2014). Interestingly, we found that Paupar is enriched in glucagon-producing α cells where it promotes the alternative splicing of Pax6 to an isoform responsible for PAX6-dependent activation of essential α cell genes. Consistent with a role for Paupar in conferring the α-cell specific functions of PAX6, deletion of Paupar in mice resulted in dysregulation of PAX6 α cell target genes and α cell dysfunction. Our findings illustrate how lncRNAs can modulate transcription factor activities to achieve cell-specific regulation.

RESULTS

Identification of developmentally regulated lncRNAs in the mouse pancreas

To identify temporally and spatially regulated lncRNAs in the pancreas, we performed RNA-sequencing (RNA-seq) on adult mouse islets and embryonic day 15.5 (e15.5) mice pancreas (EP). We analyzed these datasets using a computational pipeline (Pefanis et al., 2015) designed to identify putatively functional mouse pancreatic lncRNAs (Figure 1A) (Singer and Sussel, 2018). A set of stringent criteria was used to define 2728 pancreatic lncRNAs: (1) > 200 nucleotides (nt) in length; (2) no overlap with protein coding regions; (3) no overlap with pseudogenes (Karro et al., 2006); and (4) low predicted coding probability (Wang et al., 2013) (Figure 1B). We then filtered out genes with a Fragments Per Kilobase of transcript per Million mapped reads (FPKM) < 0.5 to enrich for lncRNAs amenable to molecular analyses (Figure 1C). These parameters yielded 572 high-confidence pancreatic lncRNAs that clustered according to developmental stage (Table S1; Figure 1D). Remarkably, comparative transcriptomics between the two stages showed that approximately half (279 or 48.7%) of all pancreatic lncRNAs were developmentally regulated: 108 lncRNAs were significantly enriched in EPs and 171 lncRNAs were significantly enriched in adult islets (Figure 1E). Given that the embryonic samples came from whole pancreas, we further verified that the majority of embryonically enriched lncRNAs were expressed in e15.5 Neurog3+ endocrine progenitor cells (Churchill et al., 2017; Table S1).

Figure 1. Systematic identification of developmentally regulated lncRNAs in the mouse pancreas.

(A) Overview of the lncRNA discovery pipeline from transcriptome analyses of e15.5 pancreas (EP) and adult isolated islets. n=3.

(B) Histogram of Coding Potential Assessment Tool (CPAT) scores of candidate lncRNAs. Dotted grey line indicates CPAT score < 0.364.

(C) Histogram of FPKM values of candidate lncRNAs. Dotted grey line indicates FPKM > 0.5.

(D) Heat map of 572 candidate lncRNAs with corresponding FPKMs, generated with Heatmapper. Z-scores calculated using Pearson distance measurements.

(E) Volcano plot of differentially expressed genes in EP compared to islets. Blue circles indicate lncRNAs enriched in EP; red circles indicate lncRNAs enriched in islets. Genes with p-value ≥ 0.05 (dotted grey line) are shown as black circles. Green square shows Paupar lncRNA.

(F) Overview of GREAT analysis.

(G) Gene ontology analysis on lncRNA-associated genes.

See also Figure S1 and Tables S1 and S2.

Previous studies have shown that subsets of lncRNAs regulate, or are co-regulated with, nearby protein-coding genes (Cabili et al., 2011; Moran et al., 2012). To gain insight into the function of these lncRNAs, we utilized Genomic Regions Enrichment of Annotations Tool (GREAT) (McLean et al., 2010) to identify nearby protein-coding genes (Figure 1F). Several lncRNAs mapped within 5 kb of a nearby transcriptional start site (TSS), suggesting they might function as bidirectional lncRNAs (Mercer, Dinger, and Mattick, 2009), while the majority of lncRNAs were located in intergenic regions (> 50 kb from a TSS) (Figure S1A). Gene ontology analyses of all neighboring genes showed a significant enrichment of genes involved in “endocrine pancreas development” and “pancreas development” (Figure 1G), including nine essential pancreatic transcription factors: Pax6, Nkx6.1, NeuroD1, Hes1, Foxo1, Mnx1, Foxa2, Meis2, and Hnf6 (Table S2).

Paupar is a nuclear lncRNA enriched in pancreatic α cells

One of the more differentially regulated lncRNAs (Figure 1E; green square) was the Pax6 Upstream Antisense RNA (Paupar), a previously characterized lncRNA located near the Pax6 locus (Vance et al. 2014) with an orthologous transcript in human islets (HI-LNC101) (Moran et al., 2012). Paupar is a 3482 bp gene transcribed 8 kb upstream and antisense from Pax6 and contained within the first intron of Pax6os1, an antisense RNA with no known function (Figure 2A). Paupar lies in a syntenically conserved region on chromosome 2 in mice and chromosome 11 in humans (UCSC LiftOver). Paupar is also highly conserved at the nucleotide level across mammals (89% in humans; Table S2), both in its gene body and regions corresponding to a putative promoter (Figure 2A), similar to many functional lncRNAs (Carninci et al. 2005). Comparative sequence analysis of Paupar across species did not reveal any conserved ORFs, reinforcing the computational prediction that Paupar lacks protein-coding potential (PhyloCSF - 6.5127; CPC 0.2666; CPAT 0.2086). In addition to being developmentally regulated and conserved in humans, we determined Paupar expression was reduced 2-fold in islets from db/db mice, a model of T2D, suggesting Paupar might play a role in islet dysfunction (Figure 2B).

Figure 2. Paupar is a nuclear lncRNA enriched in pancreatic α cells.

(A) UCSC mm10 genome browser of the Paupar locus.

(B-D) Paupar expression by qRT-PCR in, n=3-5 biological replicates, (B) WT and db/db mice,

(C) 12-week-old WT mouse tissues, and (D) whole pancreas at indicated ages and 12-week-old isolated islets.

(E) Expression analyses by qRT-PCR of purified α cells (green bars) and non-α islet cells (red bars) cells. n=3 (10-15 pooled mice).

(F) Paupar expression by qRT-PCR in αTCs following cell fractionation. Gapdh and Neat1 shown as nuclear (grey) and cytoplasmic (black) markers, respectively.

(G) Single molecule RNA fluorescent in situ hybridization (smFISH) images in αTCs showing Paupar (red) co-localized with antibody staining for Glucagon (green) and DAPI (blue). Scale bar indicates 10 μm. Images representative of 3 replicate experiments.

Quantification for (B) and others like it are presented as mean ± SEM, normalized to control values, t test, NS p > 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001.

See also Figure S2.

Paupar was previously identified as a single exonic lncRNA in the neuroblastoma N2A cell line, where it appears to have PAX6-dependent and independent functions (Vance et al., 2014). To more globally characterize Paupar expression, we performed extensive analysis of published RNA-seq datasets from 25 different mouse tissues and cell lines (Lin et al., 2014; Lawlor et al., 2017) and found that Paupar was expressed exclusively in the pancreas, eye, and brain (Figure S2A). These datasets also confirmed the findings of our initial screen, showing enrichment of Paupar in adult islets compared to embryonic pancreas (Figure S2A and Figure 1E). Furthermore, expression data from two commonly used immortalized islet cell lines, αTC-1 and βTC-6 cells (Powers et al., 1990; Hamaguchi and Leiter, 1990), that model α and β cells, respectively, suggested Paupar expression is enriched in α cells (Figure S2A). Surprisingly, in contrast to what was reported in N2A neuroblastoma cells, in the context of islets, αTCs, βTCs, and the eye, Paupar appears to contain three exons and two introns (Figures S2B and S2C).

We confirmed the RNA-Seq data using quantitative reverse transcription PCR (qRT-PCR) on RNA from 11 different mouse tissues to demonstrate that Paupar was significantly enriched in adult mouse islets, compared to the eye (~3.5-fold) and brain (~25-fold) (Figure 2C). Expression analyses on whole pancreata at several embryonic and postnatal developmental time points also revealed that onset of Paupar expression occurs between postnatal day 7 (P7) and P14 (Figure 2D), which is a critical window of postnatal development associated with the functional maturation of endocrine cells (Nishimura et al., 2006; Hang and Stein, 2011). Given that PAX6 has known regulatory roles in both α and β cells (Gosmain et al., 2010; Swisa et al., 2017), we wanted to confirm the αTC and βTC expression data that suggested Paupar expression was enriched in the α cell lineage (Figure S2A). We isolated islets from mice expressing the Glucagon-Venus reporter gene (Reimann et al., 2008) and used FACS to purify α cells from the other islet cell types. Expression analysis of FACS purified mouse α cells and non-α islet cells confirmed Paupar was highly enriched in α cells compared to other islet cell types, in contrast to Pax6, which is expressed equally in both cell populations (Figure 2E). Expression levels of Gcg and Ins2, essential α and β cell genes, respectively, are shown to demonstrate the purity of the samples (Figure 2E).

Given that the Paupar locus is highly conserved at the sequence level across placental mammals, we wanted to determine whether PAUPAR expression is conserved in human islets. Analyses of published RNA-seq datasets (Blodgett et al., 2015) showed that human PAUPAR was significantly enriched in adult human islets compared to fetal pancreas (Figure S2D), similar to Paupar expression in mice (Figure 2D). However, unlike its restricted expression in mouse islet α cells, PAUPAR was expressed in both human α and β cells (Figure S2E). This may reflect the different gene expression patterns observed between mouse and human islet cells, similar to MafB/MAFB, which is α-cell specific in mice but is expressed in both α and β cells in humans (Hang and Stein, 2011) (Figure S2F). Consistently, we determined that PAUPAR expression was positively correlated with both MAFB (Figure S2G) and GCG (Figure S2H) expression in human islets. Since Paupar is downregulated in db/db islets, we evaluated human PAUPAR expression in islets from T2D patients, as well as islets from healthy patients cultured in low or high glucose concentrations (Figures S2I and S2J). Expression analyses did not reveal a significant difference in PAUPAR in T2D islets or in human islets exposed to a glucose challenge (Figures S2I and S2J), which could be due to considerable gene expression variability between human donors or a result of PAUPAR having unique regulatory functions in human α versus β cells.

LncRNA localization within a cell can also yield important mechanistic insight; nuclear lncRNAs often regulate transcription and pre-mRNA processing, while cytoplasmic lncRNAs more likely influence mRNA stability and translation (Batista and Chang, 2013). Expression analyses on αTCs following cell fractionation revealed a significant enrichment of Paupar in the nucleus compared to the cytoplasm (Figure 2F). Nuclear localization of Paupar in αTC cells was confirmed by single-molecule fluorescent in situ hybridization (smFISH) using oligonucleotide probes targeting the full length Paupar transcript (Figure 2G). In comparison, glucagon protein is localized to the cytoplasm (Figure 2G). Analysis of Paupar smFISH images determined that the average copy number for Paupar is 54 ± 31 transcripts per αTC cell (Figure S2K). As a comparison, Paupar smFISH in MIN6 beta cell line showed no specific staining above background levels (Figure S2L). Taken together, these results demonstrate Paupar is a conserved lncRNA predominantly localized to the nuclei of mature pancreatic α cells.

Paupar regulates Pax6 α cell target genes

The expression profile of Paupar prompted us to investigate its regulatory function in α cells. Three unique sets of antisense oligonucleotides (ASO) successfully downregulated Paupar RNA in αTC cells by an average of 57% (Figure 3A). In contrast to what has been previously reported in N2A cells, we observed no change in Pax6 expression following Paupar knockdown (Figure 3A), suggesting that Paupar RNA does not influence Pax6 transcription. We were surprised then, to discover that Paupar knockdown led to the downregulation of several canonical Pax6 α cell target genes, including Gcg, MafB, Arx, and NeuroD1 (Gosmain et al., 2010) (Figure 3A). Furthermore, the amount of Paupar KD was positively correlated with the reduction in both Gcg (Figure 3B) and MafB (Figure 3C) expression. These findings suggest that PAX6-mediated regulation of α cell genes is sensitive to the level of Paupar present in αTCs. The expression of Foxa2, an α cell gene that is not a target of PAX6, was not altered by loss of Paupar, further demonstrating that the activity of Paupar is restricted to PAX6 target genes (Figure 3A). Consistently, co-expression analysis of human islets (Fadista et al., 2014) showed PAUPAR expression was positively correlated with the expression of both MAFB (Figure S2G) and GCG (Figure S2H), suggesting Paupar may have similar regulatory functions in human islets.

Figure 3. Paupar lncRNA regulates PAX6 α–cell target genes and interacts with several nuclear proteins involved in alternative splicing.

(A) Gene expression analysis by qRT-PCR from αTCs following control or Paupar knockdown (KD). n=4 for each condition.

(B, C) Correlation plots showing % Paupar KD compared to % KD of Gcg (B) and MafB (C) KD. P-values calculated using Pearson’s correlation.

(D) CHART-qRTPCR analysis on eluate from experiments done with control COs or Paupar COs showing percent RNA retrieval relative to input for Paupar, Gapdh, and TBP.

(E) Gel images corresponding to (D) with an added lane showing Paupar COs plus RNase.

(F) Diagram showing output from STRING analysis of Paupar interacting proteins found by CHART-MS. Colors correspond to the categories highlighted in (G).

(G) GO analyses of Paupar interacting proteins from (F).

(H) Depiction of results from RBPmap showing z-scores for predicted SRSF binding sites mapped to Paupar (green dots), Nkx2-2 (blue dots), and Malat1 (red dots) loci. Graph is aligned to UCSC screenshot showing the Paupar locus. Grey rectangle highlights the region of Paupar with high sequence conservation in placental mammals.

See also Figure S3 and Table S3.

Paupar lncRNA interacts with nuclear proteins involved in alternative splicing

To investigate the molecular mechanism by which Paupar regulates α-cell PAX6 target genes, we performed Capture Hybridization Analysis of RNA Targets (CHART) (Simon et al., 2011) to identify the Paupar interactome. We used two unique sets of biotinylated capture oligos (COs) to pull down Paupar, in addition to control COs (scrambled and Paupar sense). Both sets of Paupar COs retrieved Paupar RNA from αTC nuclear extract (Figures 3D and 3E). Importantly, Paupar COs did not pull down Gapdh or TBP, nor did control COs retrieve Paupar (Figures 3D and 3E). The lack of Paupar enrichment with sense COs, which cannot directly target the RNA but have sequence identity to the DNA locus, demonstrates that the CHART-based enrichment is RNA mediated (Figure S3A).

To identify Paupar-interacting proteins, we performed CHART followed by mass spectrometry (CHART-MS), and identified 56 Paupar binding proteins that were enriched > 2-fold over control COs (Table S3). Unlike in N2A cells, we were not able to detect an interaction between Paupar and PAX6 protein (Figure S3B). Analysis of Paupar interacting proteins using a database of known and predicted protein-protein interaction networks (Search Tool for the Retrieval of Interacting Genes/Proteins (STRING); Szklarczyk et al., 2017) showed that 25 out of the 56 Paupar-interacting proteins were predicted to interact with at least one other Paupar-interacting protein (Figure 3F). These protein complexes were significantly enriched in several pathways (Figure 3G), including RNA splicing (Figure 3F, blue circles), regulation of gene expression (Figure 3F, red circles), and DNA binding (Figure 3F, green circles). Since the largest complex of interacting proteins contained canonical regulators of alternative splicing (Figure 3F), we compared the set of Paupar interacting proteins with those retrieved by NEAT1 and MALAT1, two nuclear lncRNAs with established roles in alternative splicing (West et al., 2014). This analysis identified 50.6% and 51.7% overlap with NEAT1 and MALAT1 interacting proteins, respectively (Figure S3C). In contrast, similar comparison to proteins or lncRNAs that are not involved in alternative splicing mechanisms, MAFA (Scoville et al., 2015), NEUROD1 (Romer et al., 2019) and Xist (Chu et al., 2015), showed limited overlap (Figure S3C).

Consistent with a direct interaction between Paupar and 8 of the 12 annotated serine and arginine rich splicing factors (SRSFs) (Figure 3F; Table S3), we used a computational tool (RBPmap; Paz et al., 2014) to predict that the full-length Paupar transcript had 1685 putative SRSF binding sites above a stringent z-score threshold (> 2). Notably, several of the most significant SRSF sites were located within a region of the Paupar locus that has high sequence conservation across placental mammals (Figure 3H, grey rectangle). To address the possibility that SRSF binding occurs because Paupar is a spliced transcript, we compared the z-scores for Paupar SRSF binding sites to Nkx2-2, an mRNA that is spliced but does not regulate RNA splicing, as well as Malat1, a lncRNA that is not spliced but regulates SRSF-mediated RNA splicing. We found that Paupar had more SRSF binding sites with higher z-scores than both Nkx2-2 and Malat1 (Figure 3H). SRSF binding sites for the full ~7 kb Malat1 locus is shown in Figure S3D. Taken together, these findings suggest that Paupar functions in the α cell to regulate alternative splicing.

Paupar promotes the alternative splicing of Pax6 to the isoform required for activation of Pax6 α cell target genes

Previous studies have demonstrated that the regulatory role of Paupar in N2A cells is partially mediated through PAX6 (Vance et al., 2014; Pavlaki et al., 2017). The identification of interactions between Paupar and the SRSF family of proteins, and the knowledge that Pax6 has two well-characterized isoforms with distinct regulatory functions (Epstein et al., 1994; Chauhan et al., 2004; Kiselev et al., 2012; Sasamoto et al., 2017) (Figure 4A), led us to examine a possible role for Paupar-mediated alternative splicing of Pax6. The two major Pax6 isoforms, termed Pax6 (Figure 4A, blue lines) and Pax6 5a (Figure 4A, red lines), differ from each other by an alternatively spliced exon, “5a”, that adds 14 amino acids to the paired DNA-binding domain of PAX6 protein and alters DNA binding recognition (Kiselev et al., 2012). Since PAX6 regulates distinct target genes in α versus β cells (Gosmain et al., 2010; Gosmain et al., 2012), we hypothesized that its cell-specific regulatory activities could be mediated through its different isoforms. Consistently, computational analysis of alternative splicing events in α versus β cell transcriptomes (DiGruccio et al., 2016) identified the Pax6 5a isoform as significantly enriched in α cells (p < .0001; data not shown). While non-quantitative RT-PCR analysis detects both isoforms in α cells (Figure 4B), we could only quantify expression of Pax6 5a, given that amplification of the shorter Pax6 isoform produces both isoform products (Figure 4A, black arrows), which confounds qRT-PCR analysis. To directly test whether Paupar promotes alternative splicing of Pax6, we performed qRT-PCR on RNA from Paupar-deficient αTCs. Strikingly, while Paupar KD did not induce a change in total Pax6 mRNA levels, we did observe a significant and specific reduction (45%) of Pax6 5a (Figure 4C). We also determined that Paupar-mediated regulation of Pax6 splicing occurs directly though RNA-RNA interaction, rather than indirectly through SR splicing factors, since the association between Paupar and Pax6 RNA could be partially ablated by RNase treatment (Figure 4D).

Figure 4. Paupar promotes the alternative splicing of Pax6 to the isoform required for activation of PAX6 α-cell target genes.

(A) Schematic of the Pax6 genomic locus and Pax6 pre-mRNA with two isoforms, Pax6- red lines and Pax6 5a- blue lines. Black filled-in arrows are RT-PCR primers used in (B) to amplify both isoforms. White arrows are qRT-PCR primers used in (C) and (D) to specifically amplify Pax6 5a.

(B) RT-PCR analysis of αTC cDNA showing amplification of both Pax6 isoforms.

(C) qRT-PCR analysis of Paupar, total Pax6 RNA, and Pax6 5a from αTCs following control or Paupar KD.

(D) CHART qRT-PCR showing the enrichment of Pax6 5a in experiments using Paupar COs or Paupar COs with RNaseA treatment, compared to control COs.

(E) Volcano plot showing PAX6 αTC ChIP-seq peaks. Differentially bound peaks (control vs Paupar KD) with p-value < 0.01 are shown as red dots. Black dots correspond to peaks with p-value ≥ 0.01. n=2 biological replicates for each condition.

(F) Genomic distribution of all regions bound by PAX6 and regions differentially bound by PAX6 following Paupar KD.

(G) Gene ontology (GO) analyses of genes associated with differentially bound PAX6 peaks.

(H) Venn diagram showing genes associated with differentially bound PAX6 peaks compared to genes enriched > 2-fold in α cells compared to other islet cell types (DiGruccio et al., 2016). P-value calculated using Fisher’s exact test.

A study in non-pancreatic cell lines that stably expressed Pax6 or Pax6 5a showed that the two proteins have different DNA binding specificities (Kiselev et al., 2012). To determine whether Paupar mediated-alternative splicing of Pax6 5a is necessary for direct PAX6 regulation of α cell target genes, we performed PAX6 ChIP-sequencing (ChIP-seq) in αTC cells with normal or reduced amounts of Paupar and assessed changes in PAX6 global occupancy. In control α cells, PAX6 was physically associated with 26244 genomic sites; notably, approximately 10% of the PAX6 sites (2046 peaks) showed differential binding following Paupar KD (Figure 4E). Interestingly, regions that exhibited differential binding upon Paupar KD were underrepresented in regions < 1 kb from promoters and more predominantly located in intergenic regions (> 3 kb from 5’ or 3’ UTR) (Figure 4F). Gene ontology (GO) analysis of the genes located within 100kb of the 2046 differentially bound PAX6 sites showed a significant enrichment in biological processes associated with α cell function, including: glucose homeostasis, regulation of ion transport, and cellular response to glucose stimulus (Figure 4G).

To determine whether Paupar KD caused a global shift in PAX6-mediated regulation of α cell genes, we compared genes associated with the differentially bound PAX6 sites with a published dataset of α cell enriched genes (DiGruccio et al., 2016). We observed a statistically significant overlap of 60 genes between the datasets (Figure 4H), which likely underrepresents the number genes given that the ChIP-seq peak annotation strategy excludes some distal regulatory elements. Importantly, the overlapping 60 genes contained several genes misregulated by Paupar KD, including Gcg and MafB (Figure 3A). We further confirmed these findings using ChIP-qPCR for PAX6 on the promoters of Gcg and MafB (Gosmain et al., 2007; Gosmain et al., 2010). As expected, in control αTCs there was a greater than 30-fold enrichment of PAX6 over IgG on Gcg and MafB regulatory DNA (Figure S3E), while Paupar KD induced a 34% and 58% reduction in PAX6 occupancy on the Gcg and MafB promoters, respectively (Figure S3E). These results, combined with our finding that Paupar KD causes specific downregulation of Pax6 5a but not total Pax6 mRNA (Figure 4E), directly demonstrate that Paupar confers the α-cell specific regulatory function of Pax6 5a via alternative splicing.

Paupar knockout mice have impaired α cell development and function

To determine whether Paupar’s regulation of Pax6 isoform selection affected α cell function in vivo, we generated Paupar null (KO) mice by replacing the endogenous Paupar locus with the histone-fusion GFP (H2B:GFP) reporter gene (Kanda et al., 1998) (Figures S4A and S4B). Expression analysis of islets from Paupar KO mice confirmed the complete loss of Paupar RNA (Figure S4C), and immunofluorescence analyses indicated that the GFP reporter recapitulated endogenous Paupar expression specifically in the glucagon-producing α cells (Figures S4D and S4E).

Paupar KO mice are viable, fertile, and indistinguishable from their WT littermates with respect to weight (Figure S5A), ad libitum blood glucose (Figure S5B), and glucose tolerance (Figure S5C-E). These findings are not surprising given the restricted expression of Paupar in α cells; several studies have shown that mice are highly resistant to perturbations in α cell function (Furuta et al., 2001; Shiota et al., 2003; Heller et al., 2004; Hancock et al., 2010; Wilcox et al., 2013). To overcome this limitation of the mouse model, hypoglycemia, a physiological trigger for glucagon secretion from α cells, can be used to interrogate α cell function in vivo. Consistent with an α cell defect, we showed that 6-week-old Paupar KO mice were significantly hypoglycemic compared to WT controls following an overnight fast, suggesting Paupar KO mice may exhibit impaired glucagon secretion in response to fasting-induced hypoglycemia (Figure S5F). We further utilized an in vivo insulin tolerance test (ITT) to assess the ability of Paupar KO mice to respond to acute hypoglycemia. Strikingly, throughout the ITT, 6-week-old Paupar KO mice were significantly hypoglycemic compared to WT mice, suggesting KO mice were less efficient at returning to baseline blood glucose levels (Figures 5A and 5B; Figure S5I). To rule out the possibility that hypoglycemia observed in Paupar KO mice during an ITT was due to enhanced insulin sensitivity, we calculated the Homeostatic Model Assessment of Insulin Resistance (HOMA-IR), a well-established measure of insulin resistance (Matthews et al., 1985), in fasted Paupar WT and KO mice. As expected, given the absence of Paupar expression in peripheral metabolic tissues (Figure S2A), there was no difference in HOMA-IR in Paupar KO mice compared to control mice (Figure S5J). Alternatively, the inability of Paupar KO mice to respond to hypoglycemia could be due to impaired glucagon secretion. To test this, we measured plasma glucagon before and during an ITT. Thirty minutes after an insulin injection, Paupar KO mice had ~2.5-fold decrease in plasma glucagon relative to WT mice (Figure 5C), demonstrating that Paupar KO mice secrete significantly less glucagon in response to hypoglycemia.

Figure 5. Paupar knockout mice have impaired α cell development and function.

(A) Graph of insulin tolerance tests Paupar WT (black) and Paupar KO (red) mice. n=7-8 for each genotype.

(B) Area under the curve (AUC) calculations for (A).

(C) Plasma glucagon levels (pg/mL) measured by ELISA during ITTs.

(D) The amount of glucagon secreted (pg/islet/hour) by isolated islets in response to indicated stimuli. Values are relative to the average secretion during initial 20 mM glucose incubation. n=4 mice per group, duplicate batches of 20 islets per mouse.

(E) Average glucagon content per islet (pg/islet) measured by ELISA. n=4.

(F-M) Immunofluorescence on Paupar WT (F, I, L) and Paupar KO (G, J, M) pancreata showing α cells (F, G), β cells, (I, J), and merged channels with DAPI (blue) (L, M). Scale bar indicates 50 μm. Images are representative of 3 replicate experiments

(H) Quantification of (F) and (G). Values shown are average from 15-20 islets per mouse, n=3 mice per genotype.

(K) Same as (H) but showing quantification of (I) and (J).

(N) Quantification of (L) and (M) showing average islet area (mm²) in Paupar WT (plain grey bar) and KO mice (grey striped bar).

See also Figures S4-S7.

The physiological response to an insulin challenge is complex and involves several non-pancreatic tissues, including liver, muscle, and fat. Although Paupar is not expressed in those tissues (Figure S2A), to eliminate the possibility that the impaired response of Paupar KO mice to hypoglycemia was due to increased uptake of glucose by the peripheral tissues, we performed ex vivo glucagon secretion assays. Cultured islets from 6-week-old Paupar KO mice secreted ~60% less glucagon than WT islets in response to low glucose (2 mM) and ~47% less glucagon than WT islets in response to 10 mM arginine, both potent glucagon secretagogues (Gerich, Charles, and Grodsky, 1974) (Figure 5D). To determine whether this blunted glucagon secretion was due to reduced islet glucagon content, we measured total islet glucagon content and found that Paupar KO islets had dramatically less (71%) total glucagon than controls (Figure 5E). Interestingly, when we then normalized ex vivo glucagon secretion (Figure 5D) to islet glucagon content (Figure 5E), we discovered that Paupar KO islets cells inappropriately hypersecrete glucagon in high glucose conditions (Figure S5K). These findings demonstrate that impaired α cell physiological function in Paupar KO islets is at least partially due to decreased glucagon content; however, it does not rule out defects in glucose sensing, membrane depolarization, glucagon production, or exocytosis.

Consistent with the physiological phenotype, morphometric analysis on 6-week-old Paupar WT and KO pancreata revealed that while α cells made up ~13% of Paupar WT islets (Figure 5F), they comprised only ~6% of Paupar KO islets (Figure 5G), corresponding to an average 2.25-fold decrease in the α cell population (Figure 5H). We also observed a modest increase in β cell area relative to islet area in Paupar KO mice (~79%) compared to WT mice (~71%) (Figures 5I-5K). There was no measurable difference in average islet size between Paupar WT and KO mice (Figures 5L-5N). Examination of islet morphology in 7-month-old WT and KO mice showed that Paupar KO mice had significant α cell (Figures S6A and S6B) and islet hyperplasia (Figure S6C) due to increased α cell proliferation (Figures S6D-S6F). We further showed that these hyperplasic α cells are functional, given that 3- and 6-month-old Paupar KO mice were no longer hypoglycemic compared to controls in response to an overnight fast (Figures S5G and S5H). Furthermore, when we conducted ITTs on older (>6 months) mice, we discovered that throughout the assay, Paupar KO mice were hyperglycemic compared to control mice (Figures S7A-C), which was the opposite of the phenotype in 6-week-old Paupar KO mice (Figures 5A and 5B; Figure S5I). This compensation phenomenon has been well documented in several mouse models of α cell dysfunction (Gelling et al., 2003; Conarello et al., 2007; Hayashi et al., 2009; Courtney et al., 2013; Solloway et al., 2015) and provides further evidence that Paupar functions similarly to canonical α cell regulatory proteins. Taken together, these results demonstrate that Paupar is required for normal α cell development and function.

Paupar regulates essential α cell genes in vivo

To validate our in vitro studies which used an RNA knockdown approach to disrupt Paupar activity in αTC cells, we performed global transcriptome analyses on 6-week-old isolated islets from Paupar WT and KO mice which revealed 3106 differentially expressed genes (DEGs) (p-value < .05) (Figure 6A). Consistent with the α-cell specific phenotype, the DEGs represented > 10% (75/689) of all documented α cell enriched genes (DiGruccio et al., 2016) (Figure 6B). Furthermore, the DEGs included many factors required for α cell function and many known PAX6 α-cell targets, including transcription factors (Arx, MafB, Irx1, and Irx2) (Table S4), voltage-gated ion channels (Kcnq2, Kcnip3), and exocytotic machinery (Sytl5 and Sytl2) (Figure 6C). In addition, the majority (63/75) of the dysregulated α-cell specific genes were downregulated in Paupar KO mice, indicating that Paupar predominantly regulates gene activation in α cells. Of note, Paupar KO islets had normal levels of Pax6os1, as well as genes within 100 kb of Paupar. These findings, along with experiments showing similar transcriptional changes in Paupar KO islets and Paupar-deficient αTCs, demonstrate that the Paupar KO mouse phenotype is not due to deletion of an important DNA regulatory element.

Figure 6. Paupar regulates essential α cell genes in vivo.

(A) Volcano plot showing differentially expressed genes (DEGs) in Paupar KO versus WT mice. Dot colors correspond to p-values and log2 fold changes as indicated in legend above graph.

(B) Venn diagram comparing DEGs from (A) to genes enriched > 2-fold in α cells compared to β cells (DiGruccio et al., 2016). P-value calculated by Fisher’s exact test.

(C) Plots showing log2 fold change (Paupar KO/WT) values for 75 genes in middle of Venn diagram in (B) grouped by functional category.

(D) qRT-PCR analysis of FACS purified α cells from Paupar WT (green bars) and Paupar KO (green patterned bars). n=10-15 pooled mice.

(E) Venn diagram showing PAX6 peaks with altered binding following Paupar KD (Figure 4E, red dots) compared to DEGs shown in (A).

See also Table S4.

To specifically assess the regulatory function of Paupar in α cells, we isolated islets from Paupar WT and KO mice expressing the Glucagon-Venus reporter gene (Reimann et al., 2008) and used FACS to purify α cells from the other islet cell types. Consistent with our in vitro studies, α cells from Paupar KO mice showed a significant downregulation of Pax6 5a, as well as several PAX6 α-cell target genes, while total Pax6 mRNA levels were unchanged (Figure 6D). These analyses provide further evidence that Paupar-mediated regulation of PAX6 α-cell target genes is at the level of Pax6 splicing rather than transcription.

To expand our understanding of Paupar-mediated regulation of PAX6 target genes, we determined the overlap between genes regulated by Paupar in islets (3056 genes; p-value < 0.05) and genes associated with altered PAX6 binding following Paupar KD in αTCs (2056 genes; p-value < 0.01). This analysis revealed a statistically significant overlap of 328 genes between the datasets (Figure 6E), including many essential α cell genes, such as MafB, Arx, Irx1 and Irx2, whose misregulation could directly explain the Paupar KO mouse phenotype. Based on these cumulative studies, we propose a model that Paupar promotes α cell function via alternative splicing of the Pax6 5a isoform required for regulation of PAX6 α cell target genes (Figure 7).

Figure 7. Model of Paupar mediated regulation of PAX6 α-cell target genes.

Paupar (red RNA molecule) is enriched in glucagon-producing α cells where it interacts with SR proteins (green circles) to promote the alternative splicing of Pax6 pre-mRNA (blue line) to the Pax6 5a (pink lines with black dot) isoform. We demonstrate that Pax6 5a is required for activation of the essential α cells genes, including Gcg and MafB. Deletion of Paupar in vivo resulted in dysregulation of PAX6 α-cell target genes and impaired α cell development and function.

DISCUSSION

During the past 20 years, a major focus of diabetes research has been directed towards identifying the complex transcription factor networks required for the specification and function of pancreatic islet cells. Yet, how these broadly expressed transcription factors acquire unique regulatory functions at different stages of pancreas development and in different islet cell types remains poorly understood. In this study we uncover a novel mechanism by which the lncRNA, Paupar, confers cell specific regulatory function on the essential pancreatic transcription factor, PAX6, by promoting the alternative splicing of Pax6 to the Pax6 5a isoform that is required for the regulation of downstream α cell target genes (Figure 7). We have shown that the loss of Paupar blunts the production of Pax6 5a isoform, causing misexpression of PAX6 target genes and impaired glucagon-mediated glucose homeostasis (Figure 7). These findings uncover a novel layer of islet gene regulation and provide further evidence that lncRNAs are fundamental players in islet development and function.

Paupar was first identified in neuroblastoma N2A cells as a single exon lncRNA that regulated genes independently and through direct interaction with PAX6. Surprisingly, our expression analyses showed that Paupar was most highly enriched in pancreatic islets compared to all other tissues examined, including the eye, brain, and N2A cells. Furthermore, unbiased mapping of the adult islet transcriptome showed evidence of splicing within the Paupar locus; cloning and sequencing confirmed that the Paupar transcript has three exons and two introns. We also discovered that Paupar does not directly interact with PAX6, nor does it regulate Pax6 expression at the transcript level. These discrepancies between our findings and previous studies in N2A cells possibly reflect tissue-specific regulation and functional activities of Paupar. Additionally, although Paupar is also expressed in the mouse eye and brain, we did not observe any gross abnormalities or phenotypes associated with these tissues in Paupar KO animals. It is also not likely that the loss of Paupar in the brain contributes to the α cell phenotype, since we observed impaired glucagon secretion and a significant decrease in glucagon content in isolated islets of Paupar KO mice. Furthermore, Paupar-mediated regulation of α cell genes in vitro was consistent with the observed in vivo phenotypes.

Within the pancreas, Pax6 is expressed in several islet cell types, while Paupar expression is restricted to mature α cells. Although extremely low levels of Paupar transcript could be detected in β cells, Paupar knockdown experiments in the MIN6 beta cell line resulted in minimal gene expression changes (this study- GSE132072), confirming its primary role in α cells. Intriguingly, the onset of Paupar pancreatic expression between P7 and P14 corresponds to a critical developmental window during which cells acquire mature transcriptional profiles. For example, during this postnatal window the transcription factors MafA and MafB become restricted to mature β and α cells, respectively. These findings, along with the dramatic reduction in α cells seen in 6-week-old Paupar KO mice, and experiments showing reduced PAX6-activation of MafB in Paupar-deficient αTC cells, demonstrate that Paupar is required for the differentiation of mature α cells by promoting PAX6-mediated activation of MafB.

Previous studies have shown that PAX6 functions as a transcriptional activator and a repressor; however, the mechanism that mediates this dual capability in a single cell type is unknown. The discovery that Paupar expression specifically in α cells promotes the alternative splicing of Pax6 to an isoform with altered DNA binding specificity raises the interesting possibility that the different PAX6 isoforms have unique functions. Since were able to detect both the Pax6 and Pax6 5a isoforms within α cells, it is possible that the ratio of each isoform is crucial for proper gene regulation. This idea is supported by our finding that reduction of the Pax6 5a isoform in Paupar-deficient α cells corresponded to reduced expression of several canonical PAX6-activiated α cell genes, including glucagon and MafB (Gosmain et al., 2007; Gosmain et al., 2010), while ghrelin, the only gene known to be repressed by PAX6 in α cells (Ahmad et al., 2015), was not upregulated in Paupar KO islets or in Paupar-deficient αTC cells (Figure 6C and data not shown). Furthermore, the finding that Paupar KO mice phenocopy mice deleted for Pax6 in α cells (Ahmad et al., 2015) suggests the primary mechanism of Paupar in the islet is to regulate the PAX6 α cell specific transcriptional program.

Two recent studies examined the transcriptome of purified α cells from T1D donors (Brissova et al., 2018) and single islet cells from T2D donors (Segerstolpe et al., 2016). Although these datasets confirmed PAUPAR is expressed in human α cells, PAUPAR expression was not altered in the α cells derived from either T1D or T2D patients. A caveat of these studies was the small sample size analyzed; however, consistent with PAUPAR regulatory function in mice, MAFB and GCG expression were also not affected in these datasets. It is therefore possible that PAUPAR plays a conserved regulatory role in human α cells, but we still lack the comprehensive understanding of normal human α cell function that is necessary to make conclusions about the role of PAUPAR in α cell dysfunction.

While several lncRNAs have been shown to influence alternative splicing through their direct interactions with splicing factors (Romero-Barrios et al., 2018), this study is the first to demonstrate a specific interaction between a cell-restricted lncRNA and its cis-related gene that in turn influences the production of a distinct protein isoform. In particular, the presence of Paupar in islet α cells skews the alternative splicing of Pax6 to favor an isoform required for α-cell specific gene regulation. These results highlight an important mechanism through which tissue restricted lncRNAs influence transcription factor target selection to confer cell specific activities. The identification and characterization of additional cell restricted lncRNAs will be instrumental in determining the extent of this gene regulatory mechanism. We predict that additional evidence of cis lncRNA-regulation of tissue-specific mRNA splicing will emerge, since many lncRNAs function in the same pathway as their neighboring gene, even though there is often no evidence of regulation at the transcriptional level. In summary, our extensive molecular and functional characterization of an α cell enriched lncRNA suggest that lncRNAs could represent important tissue and/or cell restricted therapeutic targets to regulate the production and function of cell-specific isoforms of more widely expressed proteins.

LIMITATIONS OF STUDY

Our finding that Paupar is predominantly enriched in alpha cells presented several experimental challenges for the molecular analyses since many biochemical assays still require cell numbers that are orders of magnitude beyond what can be reasonably obtained in vivo. Given the relative paucity of alpha cells in mouse islets and their further reduction in the Paupar KO mice, it was necessary to rely on an immortalized alpha cell line for several of the mechanistic experiments. When possible, we corroborated our in vitro findings on purified alpha cells from Paupar WT and KO mice. Furthermore, our efforts to investigate human PAUPAR functions were limited by the absence of a human alpha cell line, as well as our finding that single cell transcriptome data from human islets often lacked the sequencing depth needed to detect low abundance transcripts, such as PAUPAR and the majority of tissue-specific lncRNAs. The development of additional experimental models and improved techniques to analyze human diabetic alpha cells would increase the clinical impact of our present findings.

STAR METHODS

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for resources and reagents should be directed to, and will be fulfilled by, the Lead Contact, Dr. Lori Sussel (lori.sussel@cuanschutz.edu). All unique reagents and mouse lines generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell line and cell culture

αTC1 clone 6 cells (ATCC, CRL-2934) were cultured in DMEM supplemented with 10% FBS, 15 mM Hepes, 0.1 mM non-essential amino acids, 0.02% BSA, 1.5 g/L sodium bicarbonate, 2.0 g/L glucose, and 1% antibiotic-antimycotic. MIN6 cells (Miyazaki et al., 1990) were cultured in DMEM supplemented with 10% FBS and 1% Penicillin-Streptomycin. All cell lines were passaged and maintained following standard techniques at 37°C in 5% CO2 and 95% air.

Mouse studies

We generated the Paupar knockout allele using BAC recombineering. First, Gibson assembly was used to clone H2BGFP (Kanda et al., 1998) and a polyA sequence and insert them together into the pL451 plasmid containing flox Neo flox (FNF) (Nam and Benezra, 2009). Short (80 bp) arms homologous to the regions flanking Paupar were then added to H2BGFPpA-FNF by PCR. The BAC clone (RP23-465J7; BAC-PAC Resources) was modified by insertion of the H2BGFPpA-FNF into the Paupar locus using Cre from SW106 cells, which was then retrieved into pMCS-DTA (a gift from Dr. Kosuke Yusa) with a 2 kb 5’ arm and a 5 kb 3’ rm. Positive clones were validated by PCR and DNA sequencing, and a correctly modified BAC was electroporated into mouse embryonic stem cells (129SV background) at Columbia University (Herbert Irving Comprehensive Cancer Center Transgenic Shared Resource). Potentially recombined clones were screened by PCR and two positive clones were used to generate chimeric mice that resulted in germline transmission. Chimeras were bred with FLPe transgenic mice (Jackson Laboratories) to excise the neomycin resistance cassette. Paupar +/− mice were backcrossed 10 generations into the C57BL/6J (Jackson Laboratories) background. Glucagon-Venus mice, in which cells that express proglucagon are labeled by YFP Venus, have been described previously (Reimann et al., 2008).

All mice were maintained on a mixed C57BL6/129SV genetic background. Mice were group-housed in a 12-hour light/dark cycle (light between 07:00 and 19:00) at 22°C with free access to water and food (LabDiet 5053) and maintained according to protocols (AAAQ3403) approved by the Columbia University Institutional Animal Care and Use Committee. Cages and beddings were changed once per week. Mice were monitored regularly for their health status, and there were no viral and parasitic infections during this study. Euthanasia was performed by CO2 inhalation. Both male and female mice were used for experiments except when noted. Mouse aged ranged from embryonic day 15.5 to 12 months. Specific ages are indicated in the figure legends or methods. Genotyping for the Paupar WT or KO allele was performed using primers P1, P2, and P3 (Figure S4A). These primer sequences, as well as the genotyping primers for Glucagon-Venus mice, are listed in Table S5.

Mouse islets from both male and female mice were isolated by perfusion of the pancreas with Collagenase P (Roche) through the common hepatic bile duct at a concentration of 1 mg/ml in M199 medium (Invitrogen). The pancreas was dissected out and dissociated at 37°C for 16 minutes. After several washes in M199 supplemented with 10% FBS (Gemini Bio Products), islets were then filtered (500 μm; Fisher NC0822591), hand-picked to avoid exocrine contamination, and processed for downstream applications.

Human tissue

Human islets from both male and female donors (3 non-diabetic and 3 with T2D) were obtained through the NIH-supported Integrated Islet Distribution Program via the University of Wisconsin, the University of Alberta, and the Southern California Islet Cell Resources Center. The islets were harvested from deceased donors without any identifying information with informed consent and IRB approval at the islet isolation centers. Donors ranged in age from 27 to 71 years old (mean 54.3). Mean BMI was 34.76 (range 27.3–40.1). Islet purity ranged from 55-90% (mean 78%). Upon arrival, islets were cultured in media (CMRL with 5.5 mM glucose, 10% FBS, 1% Glutamax, 1% Pen/Strep) at 37°C in 5% CO2 and 95 % air. For Figure S2J, islets were cultured in either 2.8 mM or 16.7 mM glucose for 5 days prior to RNA extraction.

METHOD DETAILS

LncRNA Identification

To identify novel lncRNAs, we implemented an established pipeline for identifying lncRNAs from RNA-Seq data (Arnes et al., 2018; Pefanis et al., 2015). Briefly, rRNA-depleted total RNA from n=3 e15.5 embryonic mouse pancreata and 12-week-old mouse islets was prepared using the Ribo-Zero rRNA removal kit (Epicentre). Biological replicates indicate that each RNA-seq data set was generated from individual mice (islets) or 3-4 pooled e15.5 embryonic pancreata. Libraries were prepared with Illumina TruSeq RNA sample preparation kit and then sequenced with 60 million, 2 × 100 paired reads on an Illumina HiSeq 2000 V3 instrument at the Columbia Genome Center. To reconstruct the transcriptomes, we first mapped all reads of total RNAs to the mouse reference genome (mm9) with TopHat v1.3.2 (Trapnell et al., 2012). Cufflinks v1.2.1 (Trapnell et al., 2012) was subsequently applied to assemble the whole transcriptome and to identify all possible transcripts. Then, the six transcriptomes from all samples was merged into a single gene transfer format (GTF) file with quantified gene RPKM values for each biological replicate. LncRNAs were then identified from the GTF file by removing transcripts if any of the following criteria were met: (1) they were overlapped with genes annotated in Ensembl and not annotated as ‘lincRNA’, ‘non_coding’, ‘anti-sense’, ‘3prime_overlapping_ncrna’, ‘processed_transcript’, ‘miRNA’, ‘misc_RNA’, or ‘pseudogene’; (2) overlapped with RefSeq genes (mm9) annotated as protein coding, where the RefSeq ID began with ‘NM’; (3) overlapped with pseudogenes from Pseudogene.org; (4) less than 200 nt in length; (5) predicted to have protein coding potential by the Coding Potential Assessment Tool (CPAT; coding probability >0.364); (6) gene-level maximum FPKM was less than 0.5. The 572 remaining transcripts are high-confidence pancreatic lncRNAs. See Table S1 for the following information on the 572 pancreatic lncRNAs: transcript ID, gene ID, genomic location (mm9), fold change (islets/EP), transcript length, FPKM values in islets and EP. The last column of Table S1 ‘Detectable reads in e15.5 Ngn3+ cells’ contains manual annotation of expression of each lncRNA an RNAseq dataset from e15.5 Ngn3+ cells (GSE80444; Churchill et al., 2017). This annotation provides readers with additional confirmation that expression of a candidate lncRNA in e15.5 whole pancreas samples corresponds to Ngn3+ endocrine progenitor cells.

RNA extraction and quantitative RT-PCR analysis

Total RNA was extracted from tissue or cells using the RNeasy Plus Micro Kit (Qiagen) or the RNeasy Plus Mini Kit (Qiagen), depending on the sample. Purified RNA was quantified by Nanodrop (Thermo Fisher), which also determined RNA quality. 250 ng of total RNA was used to synthesize cDNA with the SuperScript III First-Strand Synthesis System (Invitrogen) and random hexamer primers. Resultant cDNA was diluted in water and 25 ng was used in each qRT-PCR reaction. Reactions were run on a Bio-Rad CFX96 Real Time System using either gene specific primers with iQ Sybr Green Supermix (Biorad) or TaqMan probes (Applied Biosystems). Expression levels were normalized to TATA-binding protein (TBP) and quantified using the ΔΔCT method. Sybr Green primers and AODs used are listed in Table S5.

Cytoplasmic and nuclear RNA fractionation

αTC cells were grown to confluency, detached by trypsinization, and pelleted. Half of the pellet was used for total RNA isolation, and the other half was used for nuclear and cytoplasmic isolation using the PARIS kit (Ambion) following the manufacturer’s instructions. Assessment of Paupar expression in each compartment was done using qRT-PCR with Gapdh and Neat1 used for cytoplasmic and nuclear controls, respectively.

Single-molecule FISH (smFISH)

Oligonucleotide FISH probes were designed to target the full length of Paupar transcript. All probes were designed as 20-nt long, with CG content of 40-60%, no self-repeats and inner loop structures. Probes are labeled with Alexa647 at the 3’ end and purchased from Integrated DNA Technologies. The sequence information of all FISH probes is provided in Table S5. The hybridization experiments were performed as previously described protocols (Raj et al., 2008; Lubeck and Cai, 2012; Cui et al., 2018). Cells were seeded onto collagen-coated, glass-bottom petri dishes (#1.5 thickness, MatTek) and grown to 70% confluence. Then cells were fixed with fresh 4% paraformaldehyde and quenched with 0.1% sodium borohydride. Fixed cells were permeabilized and stored with 70% ethanol until final use. FISH probes were diluted to 10 nM in hybridization solution (10% dextran sulfate, 10% formamide, 2X SSC, 0.02% RNase-free BSA, 2 mM ribonucleotide vanadyl complex). Hybridization was performed at 37°C overnight in a humid chamber. For cells co-stained with immunofluorescence, primary antibody was 1:1000 diluted and added to the hybridization solution. The next day, cells were thoroughly rinsed with 10% formamide in 2X SSC, followed by staining with secondary antibody and DAPI. Detailed antibody information is listed in the Key Resources Table. Images were taken from an Olympus IX71-based single-molecule microscope equipped with 405 nm, 488 nm, 542 nm, 594 nm and 640 nm solid lasers. A 100X oil immersion objective lens (NA 1.4) was used and images were captured with an EMCCD camera (Andor iXon Ultra 897). A z-stack scanning was performed to cover the whole cell volume (200 nm scanning step, 30-40 frames). In processing, RNA transcripts were identified at each scanning plane with Gaussian musk fitting algorithm and projected to the final reconstructed images. The processed images were subjected to counting with home-built MATLAB programs. Cell boundaries were determined and segmented based on the auto fluorescence background, in conjunction with DAPI staining. The source code for smFISH analysis is available upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
DAPI (4′,6-diamidino-2-phenylindole) (1:1000)	Thermo Fisher	Cat#D1306; RRID: AB_2629482
Rabbit monoclonal anti-Glucagon (1:500)	Santa Cruz	Cat#sc-514592; RRID: AB_2629431
Guinea pig anti-Insulin (1:1000)	DAKO	Cat#A0564; RRID: AB_2617169
Rat monoclonal anti-Somatostatin (1:500)	Abcam	Cat#AB30788; RRID:AB_778010
Rabbit polyclonal anti-Ki67 (1:500)	EMD Millipore	Cat#AB9260; RRID: AB_2142366
Alexa Fluor 488 anti-rabbit (1:500)	Jackson ImmunoResearch	Cat#711-545-152; RRID: AB_2313584
Cy3 anti-guinea pig (1:500)	Jackson ImmunoResearch	Cat#706-165-148; RRID: AB_2340460
Alexa Fluor 647 anti-rabbit (1:500)	Jackson ImmunoResearch	Cat#711-605-152; RRID: AB_2492288
Alexa Fluor 647 anti-rat (1:500)	Jackson ImmunoResarch	Cat#712-605-153; RRID: AB_2340694
Purified rabbit anti-Pax6 for ChIP	Biolegend	Cat#901302; RRID: AB_2749901
Rabbit IgG for ChIP	Millipore Sigma	Cat#I5006; RRID: AB_1163659
Bacterial and Virus Strains
One Shot TOP10 Chemically Competent E. coli	Invitrogen	Cat#3879S
SW106 cells	Warming et al., 2005	N/A
Biological Samples
Mouse embryonic stem cells (mESCs)	129SV background	N/A
Human cadaveric islets (non-diabetic and T2D donors)	Integrated Islet Distribution Program	https://iidp.coh.org
Chemicals, Peptides, and Recombinant Proteins
Proteinase K	Promega	Cat#MC5005
Q5 HiFi DNA polymerase	NEB	Cat#M0491S
Gibson assembly kit	NEB	Cat#E2611S
Terrific Broth (TB)	Difco	Cat#243820
LB Agar	Difco	Cat#240110
RNase H	NEB	Cat#M0297S
DpnI	NEB	Cat#R0176S
EcoRI-HF	NEB	Cat#R3101S
HindIII-HF	NEB	Cat#R3104S
Go Taq DNA Polymerase	Promega	Cat#M3001
dNTPs nucleotide mix	Roche	Cat#11581295001
Sodium borohydride	Millipore Sigma	Cat#452882
Formamide	Roche	Cat#11814320001
Sucrose	Millipore Sigma	Cat#S0389
DMEM media	Thermo Fisher	Cat#11995
Penicillin-Streptomycin	Thermo Fisher	Cat#15140163
HEPES buffer	Gibco	Cat#15630-080
Non-essential amino acids solutions	Sigma	Cat#M7145
Bovine Serum Albumin	Thermo Fisher	Cat#15260037
Sodium bicarbonate	Thermo Fisher	Cat#MT25035CI
Antibiotic-Antimycotic	Thermo Fisher	Cat#15240062
M199 media	Invitrogen	Cat#11150067
Fetal Bovine Serum	Gemini Bio Products	Cat#100106
Precision plus protein kaleidoscope protein standard	Bio-rad	Cat#1610375
Lipofectamine 2000 transfection Reagent	Thermo Fisher	Cat#11668-019
D-Glucose	Millipore Sigma	Cat#G8270
L-Arginine	Millipore Sigma	Cat#A5006
Insulin (NovoLog)	Novo Nordisk	U-100
iQ Sybr Green Supermix	Bio-rad	Cat#1708880
Real Time PCR Mastermix for Taqman	Eurogentec	Cat#RTQP2X0315+
iTaq Universal SYBR Green One-Step Kit	Bio-rad	Cat#1725150
Collagenase P	Millipore Sigma	Cat#11213857001
Donkey Serum	Sigma	Cat#D9663
Dynabeads MyOne Streptavidin C1	Invitrogen	Cat#65001
Paraformaldehyde EM Grade	Polysciences, Inc.	Cat#00380
Trizol LS	Invitrogen	Cat#10296010
Critical Commercial Assays
RNeasy Plus Mini Kit	Qiagen	Cat#74134
RNeasy Plus Micro Kit	Qiagen	Cat#74034
TruSeq Stranded Total RNA (with Ribo-Zero)	Illumina	Cat#RS-122-2201
Ribo-Zero rRNA removal kit	Epicentre	Cat#MRZH11124
Pierce BCA protein assay kit	Thermo Fisher	Cat#23225
SuperScript III First-Strand Synthesis System	Invitrogen	Cat#18080051
Protein and RNA Isolation System (PARIS) Kit	Ambion	Cat#AM1921
Glucagon ELISA	Mercodia	Cat#10-1281-01
Insulin ELISA	Mercodia	Cat#10-1113-01
ChIP-IT High Sensitivity kit	Active Motif	Cat#53040
Deposited Data
Raw and processed e15.5 embryonic mouse pancreas RNA-sequencing data	This study	GEO: GSE122033
Raw and processed 12-week-old adult mouse islets RNA-sequencing data	This study	GEO: GSE122033
Raw and processed 6-week-old Paupar WT and KO RNA-sequencing data	This study	GEO: GSE121884
Raw and processed RNA-sequencing data from MIN6 cells following control KD or Paupar KD	This study	GEO: GSE132072
Raw and processed Pax6 ChIP-sequencing data from alpha-TC-1 cells following control KD or Paupar KD	This study	GEO: GSE132069
RNA-seq data from alpha-TC-1 and beta-TC-6 cells	Lawlor et al., 2017	GEO: GSE99954
RNA-seq data from whole mouse eyes	Mustafi et al., 2016	GEO: GSE38359
RNA-seq data from N2A neuroblastoma cells	Han et al., 2014	GEO: GSE45119
RNA-seq FACS purified mouse alpha and beta cells	DiGruccio et al., 2015	GEO: GSE80673
ENCODE RNA-sequencing data	ENCODE Project Consortium. 2012	GEO: PRJNA66167
RNA-seq FACS purified human alpha and beta cells	Blodgett et al., 2015	GEO: GSE67543
RNA-seq human fetal pancreas	Blodgett et al., 2015	GEO: GSE67543
RNA-seq human islets	Fadista et al., 2014	GEO: GSE50398
RNA-seq FACS purified e15.5 Ngn3+ pancreas cells	Churchill et al., 2017	GEO: GSE80444
Experimental Models: Cell Lines
alpha TC1 clone 6 (αTC) cells	American Type Culture Collection	Cat#:CRL-2934
Mouse Insulinoma (MIN6) cells	Miyazaki et al., 1990	N/A
Experimental Models: Organisms/Strains
Paupartm(H2BGFP)Suss mice	This study	N/A
Glucagon-Venus mice	Reimann et al., 2008	N/A
C57BL/6J mice	Jackson Laboratories	Cat#000664
FLPe transgenic mice	Jackson Laboratories	Cat#003946
db/db mice	Jackson Laboratories	Cat#000697
Oligonucleotides
Genotyping primers	This study	Table S5
smFISH probes	This study (IDT)	Table S5
Antisense Oligonucleotide (ASO) sequences	This study (IDT)	Table S5
qRT-PCR primers	This study	Table S5
CHART capture oligonucleotides (COs)	This study (IDT)	Table S5
MyoD1 negative control ChIP-qPCR primers	Keller et al 2007	Table S5
MafB ChIP-qPCR primers	Menéndez-Gutiérrez et al., 2015	Table S5
Gcg ChIP-qPCR primers	Schaffer et al., 2013	Table S5
Pax6 TaqMan AOD	Thermo Fisher	Mm00443081_m1
Glucagon TaqMan AOD	Thermo Fisher	Mm00801712_m1
MafB TaqMan AOD	Thermo Fisher	Mm00627481_s1
Arx TaqMan AOD	Thermo Fisher	Mm00545903_m1
NeuroD1 TaqMan AOD	Thermo Fisher	Mm01946604_s1
Foxa2 TaqMan AOD	Thermo Fisher	Mm01976556_s1
Irx1 TaqMan AOD	Thermo Fisher	Mm01352526_m1
Irx2 TaqMan AOD	Thermo Fisher	Mm01340315_m1
Recombinant DNA
Paupar BAC clone	BAC-PAC resources	RP23-465J7
H2B:GFP	Kanda et al., 1998	Addgene Cat#11680
pL451	Nam and Benezra, 2009	Addgene Cat#22687
pMCS-DTA	Generous gift from Kosuke Yusa, Osaka University, Japan	N/A
Software and Algorithms
DESeq2	Love et al., 2014	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
GraphPad Prism 7	GraphPad Software	https://www.graphpad.com/scientific-software/prism
ImageJ	NIH	https://imagej.nih.gov/ij
Photoshop CC 2018	Adobe	N/A
Illustrator CC 2018	Adobe	N/A
R Software Package 3.3.1	The R Foundation	https://www.r-project.org
HISAT2 (v2.1.0)	Kim et al., 2013	https://ccb.jhu.edu/software/hisat2/index.shtml
Samtools (v1.4)	Li et al., 2009	http://samtools.sourceforge.net
HTSeq (v0.10.0)	Anders et al., 2015	https://htseq.readthedocs.io/en/master/install.html
Genomic Regions Enrichment of Annotations Tool (GREAT) (v3.0.0)	McLean et al., 2010	http://great.stanford.edu/public/html
TopHat2 (v2.1.1)	Kim et al., 2013	http://ccb.jhu.edu/software/tophat/index.shtml
Bowtie2 (v2.2.8)	Langmead and Salzberg, 2012	http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
Bedtools (v2.17.0)	Quinlan and Hall, 2010	http://bedtools.readthedocs.io/en/latest
Heatmapper	Babicki et al., 2016	http://heatmapper.ca
Coding-Potential Assessment Tool (CPAT) (v1.2.4)	Wang et al., 2013	http://rna-cpat.sourceforge.net
PhyloCSF	Lin, Jungreis, and Kellis, 2011	https://github.com/mlin/PhyloCSF/wiki
UCSC Genome Browser	Kuhn, Haussler, and Kent, 2013	https://genome.ucsc.edu
Coding Potential Calculator (CPC)	Kong et al., 2007	http://cpc.cbi.pku.edu.cn
Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) (v10.0)	Szklarczyk et al., 2017	https://string-db.org
RBPmap (v1.1)	Paz et al., 2014	http://rbpmap.technion.ac.il
Replicate Multivariate Analysis of Transcript Splicing (rMATS) (v4.0.2)	Shen et al., 2014	http://rnaseq-mats.sourceforge.net
Contaminant Repository for Affinity Purification (CRAPome)	Mellacheruvu et al., 2013	https://www.crapome.org/
IGV	The Broad Institute	http://software.broadinstitute.org/software/igv
ApE	M. Wayne Davis	http://jorgensen.biology.utah.edu/wayned/ape
Gene Ontology	Ashburner et al., 2000	http://geneontology.org
MACS2 (v2.1.2)	Zhang et al., 2008	https://github.com/taoliu/MACS
DiffBind (v2.12.0)	Ross-Innes et al., 2012	https://bioconductor.org/packages/release/bioc/html/DiffBind.html
ChIPSeeker (v1.20.0)	Yu, Wang, and He, 2015	https://www.bioconductor.org/packages/release/bioc/html/ChIPseeker.html
Other

Immunohistochemistry

Tissues were fixed in 4% PFA in PBS overnight, washed in cold PBS, incubated in 30% sucrose, and cryopreserved. Immunofluorescence was performed on 7 mM sections. See Key Resources Table for details on primary and secondary antibodies used. DAPI was applied for 15 min following a 2-hour secondary antibody incubation.

Morphometric analysis

For all morphometric analysis, the entire pancreas was sectioned, and at least six evenly distributed sections were analyzed. To determine percentage hormone positive area over islet area, sections were stained with insulin, glucagon, and DAPI and five islets per section were imaged at 20X on a Zeiss Confocal LSM 710 microscope. Hormone positive area was measured using FIJI with a standardized signal threshold. Islet area was measured manually using FIJI based on morphology and guided by staining. Islets were arranged by size, and the total islet number was used to calculate the percentage represented in each size group. To determine the percentage of proliferating α cells, sections were stained with glucagon, Ki67, and DAPI, and five islets per section were imaged for quantification. Data presented is from n=3 mice per genotype unless otherwise indicated. Scorer was blinded for genotypes during quantification.

Flow Cytometry

Isolated islets from Glucagon-Venus mice were digested in 0.25% trypsin for 7 minutes at 37°C. The enzyme was deactivated by addition of M199 media supplemented with 10% FBS. The resulting single cell suspension was washed in cold 1X dPBS and filtered through a 35 μm filter prior to FACS. DAPI was added to the cell suspension as a viability marker. A BD Aria II Cell Sorter was used to separate populations of Venus-positive or -negative cells. Cells were sorted into Trizol LS and RNA was extracted according to the manufacturer’s protocol.

Assessment of islet function

Glucose tolerance test

Mice were fasted overnight (~16 hours), followed by an IP injection of 2 mg D-glucose per gram mouse weight. Tail vein blood samples were collected at 0, 15, 30, 60, 90, and 120 minutes after the injection and glucose concentration was determined using the Accu-Chek Compact Plus Blood Glucose Meter (Roche).

Insulin tolerance test

Mice were fasted for 5 hours starting at 9 am. Insulin (Humalog in PBS) was injected IP at 0.75 units per kg mouse weight. Tail vein blood samples were collected at 0, 15, 30, 60, 90, and 120 minutes after the injection and glucose concentration was determined using the Accu-Chek Compact Plus Blood Glucose Meter (Roche). To measure plasma glucagon during an insulin tolerance test, blood was collected into heparinized tubes immediately before, and 30 minutes after, an insulin injection. Plasma was separated from whole blood by centrifugation and glucagon concentration was measured by ELISA (Mercodia).

Islet glucagon secretion and content

After isolation, duplicate samples of 20 islets per mouse were cultured overnight at 37°C in resting medium (RPMI 1640 with 10% FBS, 1% P/S, and 5.6 mM D-glucose). The next day, islets were transferred into modified Krebs buffer with 0.1% BSA and 20 mM D-glucose for 30 minutes to equilibrate. Islets were then transferred to Krebs buffer with 20 mM glucose for 30 minutes, followed by 2 mM glucose for 30 minutes. Lastly, islets were transferred to either Krebs buffer with 20 mM glucose or 10 mM L-Arginine for 30 minutes. Supernatant was collected following each treatment. Islets were then sonicated in 50 μl of lysis buffer (TE with 0.1% SDS) and lysate was used to determine DNA concentration and measure glucagon content. Glucagon concentration in supernatant and islet lysate was measured by glucagon ELISA (Mercodia).

HOMA-IR determination

Cohorts of mice were fasted overnight for ~16 hours. After the fast, tail vein blood was collected and blood glucose concentration was determined using the Accu-Chek Compact Plus Blood Glucose Meter (Roche). To determine insulin plasma concentration, fasted blood was collected into heparinized tubes. Plasma was separated from whole blood by centrifugation and insulin concentration was measured by ELISA (Mercodia). HOMA-IR was determined by the formula: HOMA-IR = fasting plasma insulin (μU/L) * fasting blood glucose (mmol/L) / 22.5.

RNA sequencing (RNA-seq)

For RNA-seq experiments done on islets from Paupar WT and KO mice, as well as MIN6 cells following control or Paupar KD, total RNA was extracted and converted into cDNA libraries (TruSeq RNA sample preparation kit, Illumina) according to manufacturer’s instructions. Sequencing was performed to a depth of 30 million, single-end 100 nt reads in three biological replicates per condition. Reads were aligned to the mouse genome (mm10) using HISAT2 v2.1.0 (Kim et al., 2013). Aligned reads were assigned to genes using annotations from Ensembl (Mus_musculus.GRCm38.73.gtf) and HTseq-count v0.10.0 (Anders et al., 2015). Differential expression across cohorts was assessed using DESeq2 v1.18 (Love, Huber, and Anders, 2014). Inclusion of samples required RNA integrity (RIN) values > 8.0 as determined with Agilent Bioanalyzer 2100. Complete RNA-seq data are available through GEO accession numbers: GSE122033 (embryonic pancreas and islet lncRNAs), GSE121884 (Paupar WT and KO islets), and GSE132072 (MIN6 cells following control KD or Paupar KD).

RNA interference

The day before transfection, 500,000 cells were plated in each well of a 12-well tissue culture plate. Cells were transfected in triplicate with 40 nM Paupar or control antisense oligos (ASOs) (IDT) using Lipofectamine 2000, per the manufacturer’s instructions (Thermo Fisher). Cells were harvested for RNA 48 hours post transfection. ASO sequences are listed in Table S5.

Capture Hybridization Analysis of RNA Targets (CHART) enrichment and analysis

CHART enrichment experiments were performed as previously described (Simon et al., 2011). Briefly, CHART extract was prepared from approximately 5 × 10⁷ αTC cells per CHART reaction and hybridized with 120 pmol biotinylated capture oligonucleotides (COs) (Table S5) overnight with rotation at room temperature. Complexes were captured using MyOne Streptavidin C1 beads (Invitrogen). Bound material was stringently washed and eluted using RNase H (New England Biolabs) for 15 minutes at room temperature. For RNA analysis, CHART-enriched material was incubated with XLR buffer (final concentrations: 2 mg/mL Proteinase K (Ambion), 33.3 mM Tris pH 7.2, 0.33% SDS, 16.7 mM EDTA) at 55°C for 1 hour and 65°C for 1 hour to reverse cross-links. RNA was purified using Trizol LS according to the manufacturer’s instructions. qPCR analysis utilized iTaq Universal SYBR Green One-Step Kit to quantify Paupar enrichment compared to control genes. For analysis of CHART enriched proteins, eluate was precipitated using trichloroacetic acid and resuspended in SDS (4.25%), Tris pH 8.8 (529 mM), EDTA (64 mM) and β-mercaptoethanol (1.37 M). Samples were incubated at 98°C for 30 minutes and 65°C for 2 hours to reverse cros s-links. Samples (including 1%, 2.5%, and 5% input) were resolved by SDS-PAGE and transferred to nitrocellulose membranes. Membranes were incubated with antisera to detect PAX6 (Biolegend) at 1 μg/mL and bands visualized with SuperSignal West Pico PLUS Chemiluminescent Substrate.

CHART-MS

Proteins in CHART eluate were precipitated overnight at −20 °C with 4 volumes of cold acetone. The precipitated sample was centrifuged for 20 min at 16,000 × g at 4 °C and the supernatant discarded. The remaining pellet was washed with 1 mL cold acetone, centrifuged for 5 min at 16,000 × g at 4 °C, and the supernatant discarded. The remaining pellet was dried under vacuum. The protein pellet was digested to peptides using the TFE digest protocol (Wang et al., 2005) and peptide samples were analyzed by mass spectrometry. For LC-MS/MS analysis, 5 μl of 0.1 μg/μl of peptides per sample were analyzed by reverse phase separation (C18) using a Waters nanoEquityTM UPLC system interfaced with a QExactive Plus Orbitrap mass spectrometer (Fort et al., 2018). LC-MS/MS datasets were converted to peak lists (DTA files) using the DeconMSn software (Mayampurath et al., 2008) and searched with MS-GF+ software (Kim, Gupta, and Pevzner, 2008; Kim and Pevzner, 2014) against Uniprot/SwissProt mus musculus database. The identified spectra were filtered based on their MSGF+ QValue score, which represents the false discovery rate of peptide-spectrum-matches (PSMs). PSMs with QValue less than 0.01 were retained. Proteins were quantified using the spectral counting quantification approach (Wang et al., 2008). Downstream analyses identified Paupar binding proteins that were enriched > 2-fold over controls, excluding proteins found in > 75% of control experiments listed in the Contaminant Repository for Affinity Purification (CRAPome; Mellacheruvu et al., 2013). We used Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) (Szklarczyk et al., 2017) to identify known and predicted protein-protein interaction networks among our Paupar-interacting proteins.

ChIP-qPCR

The ChIP-IT High Sensitivity kit (Active Motif) was used according to the kit instructions. Briefly, 48 hours post ASO transfection (after ensuring efficient Paupar KD), αTC cells were crosslinked, lysed, and sonicated to obtain DNA fragments with an average length of 200–1200 bp. Supernatant containing DNA-protein complexes was used for immunoprecipitations reactions consisting of 30 μg chromatin and 4 μg rabbit anti-PAX6 antibody (Biolegend) or rabbit IgG control antibody (included in kit). Immunoprecipitated chromatin was collected using protein G agarose beads, then washed, eluted, and purified according to kit instructions. ChIP-qPCR were performed using the iQ Sybr Green Supermix (Biorad) and primers against negative control (MyoD1), MafB, Glucagon promoter regions. Data were normalized to the input signal and IgG values.

ChIP-sequencing (ChIP-seq)

ChIP-seq libraries were prepared according to the instructions of the KAPA Hyper Prep kit for Illumina. Sequencing was performed using the Illumina HiSeq 2500 system according to the manufacturer’s instructions. Reads were mapped to the reference genome (mm10) using Bowtie2 (Langmead and Salzberg, 2012) with the ‘--sensitive-local’ setting. Peaks were called using MACS2 v2.1.2 (Zhang et al., 2008) with reads extended by 200 bp while bypassing the standard model using a qvalue cutoff of 0.01. Differential binding between control KD and Paupar KD samples was assessed using DiffBind v2.12.0 (Ross-Innes et al., 2012) with default settings. Annotation and visualization of genes nearby PAX6 binding peaks was done using ChIPseeker v.1.20.0 (Yu, Wang, and He, 2015). Complete ChIP-seq data is available through GEO under the accession number GSE132069.

QUANTIFICATION AND STATISTICAL ANALYSIS

Graphs were generated in GraphPad Prism 7 and statistical analyses were performed using Graphpad Prism 7. Statistical parameters including the value of n, statistical test used, and significance (p-value) are reported in the figures and their legends. For studies involving mouse tissues, replicates refer to samples derived from different mice, unless otherwise specified. For studies involving cell culture, replicates refer to transfections performed on different days. No exclusion criteria were employed for either in vivo or in vitro experiments. Unpaired two-tailed t tests were used to assess significance when comparing qRT-PCR expression values of 2-3 genes between two conditions. Linear regression analysis was used to determine if two variables were significantly correlated to each other. The Fisher’s exact test was used to examine the significance of the association (contingency) between two types of classifications. For differential expression of global measurements (RNA-seq), the DESeq2 software (Love et al., 2014) generated adjusted p-values using the Benjamini-Hochberg procedure to correct for multiple-hypothesis testing.

Analysis of publically available RNAseq datasets from GEO was done using the following computational pipeline. Reads were aligned to the mouse genome (mm10) using HISAT2 v2.1.0 (Kim et al., 2013). Aligned reads were assigned to genes using annotations from Ensembl (Mus.musculus.GRCm38.73.gtf) and HTseq-count v0.10.0 (Anders et al., 2015). Differential expression across cohorts was assessed using DESeq2 v1.18 (Love, Huber, and Anders, 2014).

DATA AND CODE AVAILABILITY

The accession numbers for the sequencing data reported in this paper are: GSE122033 (embryonic pancreas and islet lncRNAs), GSE121884 (Paupar WT and KO islets), GSE132072 (MIN6 cells following control KD or Paupar KD), and GSE132069 (PAX6 ChIP-seq following control or Paupar KD in αTCs). All datasets can be accessed using the SuperSeries accession number GSE132318.

Supplementary Material

3

Table S1. Related to Figure 1, Information of pancreatic lncRNAs

4

Table S5. Oligonucleotides Used in This Study, Related to STAR Methods

HIGHLIGHTS.

• The conserved lncRNA Paupar is enriched in postnatal pancreatic islet α cells

• Paupar interacts with SR proteins to promote the alternative splicing of Pax6

• Loss of Paupar alters PAX6 binding specificity to dysregulate essential α cell genes

• Paupar KO mice develop fewer islet α cells and have defective glucagon secretion

CONTEXT AND SIGNIFICANCE.

Understanding and treating diabetes is limited by insufficient knowledge about the regulatory mechanisms that orchestrate pancreatic islet function. Long noncoding RNAs (lncRNAs) are a relatively unexplored class of molecules that have been implicated in many biological processes and are often misregulated in diseased states. In this study, researchers at the University of Colorado and their collaborators characterize a pancreatic lncRNA, Paupar, finding it to be enriched in alpha cells, the class of pancreatic cells that produce the blood-sugar regulating hormone glucagon. Paupar enables cell-specific functions of the transcription factor PAX6 by bestowing preference for a PAX6 variant that activates alpha cell genes. Accordingly, mice lacking Paupar have fewer alpha cells and secrete insufficient levels of glucagon, leading to impaired glucose homeostasis. These results identify the lncRNA Paupar as a potential therapeutic target for the treatment of hyperglycemia and diabetes.