Graphical abstract

   graphic file with name fx1.jpg
   [60]Open in a new tab

Highlights

     * •
       Glucose upregulates Pax6os1/PAX6-AS1 in human and mouse beta cell
       lines and islets
     * •
       Silencing Pax6os1/PAX6-AS1 increases beta cell signature gene
       expression
     * •
       On a high-fat diet, female Pax6os1 null mice show slightly enhanced
       glucose tolerance
     * •
       Pax6os1/PAX6-AS1 bind EIF3D and histones H3 and H4, suggesting a
       role in epigenetic regulation
     __________________________________________________________________

   Cell biology; Cellular physiology; Molecular biology

Introduction

   Type 2 diabetes (T2D) develops when beta cells within pancreatic islets
   no longer secrete sufficient insulin to lower circulating blood glucose
   levels, usually in the presence of insulin resistance.[61]^1 However,
   in a subset of T2D patients, defective insulin secretion is observed
   despite near-normal insulin sensitivity.[62]^2 Therefore, in all forms
   of the disease, changes in beta cell “identity” are thought to play an
   important role in functional impairment and the selective loss of
   glucose responsiveness.[63]^3

   Loss of normal beta cell function is often characterized by decreased
   expression of insulin (INS) and of genes critical for glucose entry and
   metabolism.[64]^3^,[65]^4 These changes may be accompanied by increased
   expression of so-called “disallowed genes,” whose levels are unusually
   low in healthy beta cells compared to other cell types.[66]^5
   Furthermore, in several models of diabetes, the aforementioned changes
   are associated with decreased expression of transcription factors that
   are required to maintain a mature beta cell phenotype, including
   pancreatic duodenum homeobox-1 (PDX1)[67]^6 and MAF BZIP transcription
   factor A (MAFA).[68]^6^,[69]^7 The transcription factor Pax6 regulates
   the expression of several genes involved in insulin processing and
   secretion, while repressing signature genes defining different
   endocrine cell lineages, such as ghrelin (Ghrl).[70]^8^,[71]^9^,[72]^10
   As a result, Pax6 expression is key to maintaining beta cell identity
   and function. Embryonic deletion of Pax6 in the murine pancreas leads
   to a drastic reduction in the number of alpha and beta cells, resulting
   in the death of mutant mice at postnatal days 3–6 due to severe
   hyperglycaemia.[73]^11 Conditional inactivation of Pax6 in adult mice
   leads to impaired beta cell function and glucose intolerance,[74]^12
   demonstrating the continued importance of this gene in mature beta
   cells. Further highlighting the importance of this locus in diabetes,
   pancreatic Pax6 cis-regulatory elements that interact with the Pax6
   promoter and neighboring long non-coding RNAs modulate the activity of
   pancreas-related transcription factors such as Pax4.[75]^13 In humans,
   loss-of-function mutations in PAX6 are associated with aniridia (iris
   hypoplasia) and T2D.[76]^14^,[77]^15

   Long non-coding RNAs (lncRNAs), defined as transcripts >200 nucleotides
   in length that are not translated into proteins, are crucial components
   of the pancreatic islet regulome, whose misexpression may contribute to
   the development of T2D.[78]^16 LncRNAs are expressed in a
   tissue-/cell-specific manner and more than 1,100 have been identified
   in human and murine pancreatic islets.[79]^17^,[80]^18 Furthermore, the
   expression of several of these is modulated by high glucose
   concentrations, suggesting that they may be involved in beta cell
   compensation in response to high insulin demand.[81]^19 Interestingly,
   a number of beta-cell-enriched lncRNAs are mapped to genetic loci in
   the proximity of beta cell signature genes, such as PLUTO-PDX1 or
   Paupar-Pax6, and regulate their expression in cis.[82]^19^,[83]^20

   In the current study, we sought to determine whether a lncRNA expressed
   from the PAX6 locus, previously annotated as Pax6 opposite strand 1
   (Pax6os1) in mice and PAX6 antisense 1 (PAX6-AS1) in humans,[84]^21
   might impact beta cell identity and/or function through the modulation
   of Pax6 expression or by other mechanisms.

Results

Pax6os1/PAX6-AS1 expression is enriched in pancreatic islets and upregulated
by high glucose as well as in T2D

   The lncRNA Pax6os1/PAX6-AS1 is a 1,464/1,656 nucleotide transcript
   mapped to a syntenically conserved region in chromosome 2 in mice and
   chromosome 11 in humans. It is transcribed antisense to the Pax6 gene,
   overlapping with intron 1 in both species. The first intron of
   Pax6os1/PAX6-AS1 also overlaps with Paupar, another lncRNA that is
   mainly expressed in alpha cells, and it is involved in Pax6
   splicing.[85]^17 As opposed to other lncRNAs including Paupar,
   Pax6os1/PAX6-AS1 is not highly conserved at the nucleotide level
   between species, containing four exons in mice and three in humans
   ([86]Figure 1A), whereas important differences are also found in the
   predicted secondary structures ([87]Figure 1B). A strong tendency
   toward moderate correlation between Pax6os1 and PAX6AS1 was found in
   2mers (r = 0.469), whereas 3mers (r = 0.198), 4mers (r = 0.92), and
   5mers (r = 0.03) only showed weak correlations far from being
   statistically significant ([88]Figures S1A–S1C). Curiously, a
   correlation was also found in 2mers between PAX6-AS1 and MALAT1 (r =
   0.61) ([89]Figure S1 related to [90]Figure 1), an lncRNA that has been
   shown to be located both in the nucleus and cytoplasm in beta cells,
   where it is involved in chromatin remodeling and microRNA sponging,
   respectively.[91]^16 Tissue distribution of Pax6os1 in the mouse is
   similar to that described for Pax6,[92]^22 being predominantly
   expressed in pancreatic islets and, to a lesser extent, in the eye and
   brain ([93]Figure 1C). Within the islet, previously published data
   indicate that Pax6os1 is enriched in beta and delta cells,[94]^23
   whereas there is no detectable expression in alpha cells, where Paupar
   is strongly expressed ([95]Figure S2 related to [96]Figure 1).[97]^17

Figure 1.

   [98]Figure 1
   [99]Open in a new tab

   Pax6os1 is chiefly expressed in pancreatic islets in the mouse and is
   upregulated by high-fat diet as well as in islets from patients with
   type 2 diabetes

   (A) Schematic representation of the long non-coding RNA identified at
   the Pax6 locus in mice and humans.

   (B) Secondary structure predicted using RNAfold from the VIENA package
   for Pax6os1 and PAX6-AS1, respectively, and drew using FORNA webserver.

   (C) Tissue distribution of Pax6os1 expression. n = 3.

   (D and E) Pax6os1 and Pax6 expression in MIN6 cells and CD1 mouse
   pancreatic islets cultured at different glucose concentrations for 48 h
   (note that the standard glucose concentration for MIN6 cells culture is
   25 mM). MIN6cells: n = 6. CD1 islets: n = 3.

   (F) Pax6os1 and Pax6 expression in pancreatic islets from C57/BL6 mice
   in standard (STD) or high-fat diet (HFD) for 8 weeks. n = 6.

   (G and H) PAX6-AS1 and PAX6 mRNA expression in EndoC-βH1 cells and
   human islets cultured with different glucose concentrations for 48 h
   (note that the standard concentration for EndoC-βH1 culture is 5.5 mM).
   EndoC-βH1: n = 5. Human islets: n = 7.

   (I) PAX6-AS1 and PAX6 expression in human pancreatic islets from
   normoglycemic or diabetic donors. Control: n = 10; Diabetic: n = 5.
   Data are represented as the mean ± SEM. ∗p < 0.05; one-way ANOVA
   repeated measurements.

   To determine whether Pax6os1/PAX6-AS1 expression may be modulated under
   conditions of glucotoxicity, levels of the lncRNA were measured in both
   murine and human cell lines as well as in primary islets maintained at
   different glucose concentrations. Culture for 48 h in the presence of
   high glucose induced Pax6os1 expression in both MIN6 cells (15 and 35
   vs. 5 mM glucose; n = 5, p = 0.02 and 0.03, respectively) and CD1 mouse
   islets (11 vs. 3 mM glucose; n = 3, p < 0.01) ([100]Figures 1D and 1E).
   Furthermore, Pax6os1 expression was increased in pancreatic islets from
   mice fed a high-fat diet (HFD) compared to controls (n = 6–5, p =
   0.003), whereas Pax6 mRNA levels remained unaffected ([101]Figure 1F).
   Likewise, PAX6-AS1 expression was upregulated in the human EndoC-βH1
   cell line (n = 5, p = 0.01) as well as human pancreatic islets (n = 7,
   p = 0.03) cultured at elevated glucose concentrations ([102]Figures 1G
   and 1H) ([103]Table S1 related to [104]Figures 1 and [105]6). More
   importantly, expression of PAX6-AS1 was substantially (4- to 5-fold)
   increased in islets from donors with T2D (Hba1c ≥ 6.5% or fasting
   glucose ≥126 mg/dL) vs. normoglycemic donors (n = 11–5, p < 0.01)
   ([106]Figure 1I). In contrast, PAX6 mRNA levels remained constant
   independently of the glucose concentration or disease status
   ([107]Figures 1G–1I).

Figure 6.

   [108]Figure 6
   [109]Open in a new tab

   PAX6-AS1 knockdown in human islets enhances GSIS, whereas PAX6-AS1
   overexpression exerts opposite effects

   (A) mRNA expression of PAX6-AS1, beta-signature genes, and markers from
   other endocrine cell lineages in islets infected with a scrambled or a
   shRNA-targeting PAX6-AS1. n = 4–5.

   (B and C) GSIS represented as the fold change or % of total insulin
   content in PAX6-AS1-silenced islets. n = 5.

   (D) Total islet insulin content.

   (E) Trace showing calcium response in PAX6-AS1-silenced islets. n = 4.

   (F and G) Area under the curve (AUC) for calcium dynamics in response
   to 17 mM glucose and 20 mM KCl, respectively, in PAX6-AS1-silenced
   islets. n = 4.

   (H) Glucagon secretion in PAX6-AS1-silenced islets compared to
   scrambled. n = 4.

   (I) mRNA expression of PAX6-AS1 and beta cell signature genes in
   control and islets overexpressing PAX6-AS1. n = 6–3.

   (J and K) GSIS represented as the fold change or % of total insulin
   content in PAX6-AS1-overexpressing islets. n = 3.

   (L) Total islet insulin content.

   (M) Trace showing calcium response in PAX6-AS1-overexpressing islets.
   n = 4.

   (N and O) Area under the curve (AUC) for calcium dynamics in response
   to 17 mM glucose and 20 mM KCl, respectively, in
   PAX6-AS1-overexpressing islets. n = 4.

   (P) Glucagon secretion in PAX6-AS1-overexpressing islets compared to
   control. n = 4. Data are represented as mean ± SEM. ∗p < 0.05, paired
   Student’s t test.

Pax6os1 silencing upregulates beta cell signature genes in MIN6 cells

   To explore the potential roles of Pax6os1 in beta cell function or
   survival, we first transfected murine MIN6 cells with a small
   interfering RNA (siRNA) targeting the lncRNA. RNA sequencing (RNAseq)
   analysis was then performed, revealing that Pax6os1 silencing
   (“knockdown”; KD) in MIN6 cells upregulated the expression of several
   beta cell signature genes, including Ins1, Slc2a2 (Glut2), and Pax6
   while further downregulating several “disallowed genes” such as Slc16a1
   and Ldha ([110]Figure 2A). In addition, enriched KEGG pathways in
   Pax6os1-silenced MIN6 cells included “insulin secretion,” “maturity
   onset diabetes of the young,” and “type II diabetes mellitus”
   ([111]Figure 2B). RT-qPCR analyses in cells confirmed a 35 ± 5%
   decrease in Pax6os1 expression (p = 0.0005) as well as an increase in
   Pax6 (1.28 ± 0.046-fold change; p = 0.05), Glut2/Slc2a2 (1.52 ±
   0.15-fold change; p = 0.0144), and Mafa (1.72 ± 0.31-fold change; p =
   0.040) mRNA levels ([112]Figure 2C). However, despite the upregulation
   of several beta cell signature genes, glucose-stimulated insulin
   secretion (GSIS) was not affected by Pax6os1 silencing ([113]Figures 2D
   and 2E).

Figure 2.

   [114]Figure 2
   [115]Open in a new tab

   Pax6os1 silencing upregulates beta cell signature genes in MIN6 cells

   (A) Differential expressed genes by Pax6os1 knockdown as determined by
   RNA-seq performed in MIN6 cells 72 h post-transfection with siRNA
   targeting Pax6os1. n = 4.

   (B) KEGG pathway enrichment analysis relative to (A). Significantly
   enriched KEGG pathways (p < 0.05) are presented, and the bar shows the
   fold enrichment of the pathway.

   (C) mRNA levels of beta cell signature genes and markers characteristic
   of other endocrine cell lineages in control and Pax6os1 knockdown
   cells. n = 7.

   (D) Fold change of insulin secreted relative to 3 mM glucose. n = 5.
   (E) Total insulin content. n = 14.

   (F) Pax6os1 subcellular distribution.

   (G) Schematic representation of Pax6os1 pull-down and MS.

   (H) Relationship in abundance ratios above the 1.1 cut between the two
   experimental replicates performed. Top 5 hits are labeled.
   Short/branched chain-specific acyl-CoA dehydrogenase, mitochondrial 1
   (ACADSB), eukaryotic translation initiation factor 3 subunit D (EIF3D),
   inosine-5′-monophosphate dehydrogenase 2 (IMPDH2), histone 1.0 (H1.0),
   and uncharacterized protein Rab8a (Rab8a). Data are represented as the
   mean ± SEM. ∗p < 0.05, Student’s t test.

   We next sought to determine the mechanism by which Pax6os1 may modulate
   the expression of target genes in MIN6 cells. LncRNAs may regulate gene
   expression through a number of different mechanisms, including
   chromatin remodeling, activation/repression of transcription factors in
   the nucleus as well as modulation of mRNA/protein stability in the
   cytoplasm.[116]^16 Therefore, the subcellular localization of an lncRNA
   may provide a guide as to its mechanism(s) of action. Determinations of
   Pax6os1 subcellular localization in MIN6 cells by subcellular
   fractionation indicated that the lncRNA was located in both the nucleus
   (∼40%) as well as the cytoplasm (∼60%) ([117]Figure 2F). Consistent
   with these results, both nuclear and cytoplasmic proteins were
   identified by mass spectrometry as binding protein partners of Pax6os1
   ([118]Table 1). Interestingly, the top five hits included Ras-related
   protein (RAB8A), eukaryotic translation initiation factor 3 subunit D
   (EIF3D), inosine-5′-monophosphate dehydrogenase 2 (Impdh2),
   short/branched chain-specific acyl-CoA dehydrogenase (ACADSB), and
   Histone H1.0 ([119]Figures 2G and 2H). In addition, histones H4, H3.2,
   H2B, H1.1, and H1.4 in the nucleus as well as 3′-5′ RNA helicase YTHDC2
   in the cytoplasm were also identified as Pax6os1-binding partners
   ([120]Table 1).

Table 1.

   Pax6os1 protein binding partners identified by mass spectrometry
   Name % Protein coverage Unique peptides Average ratio (Control vs.
   Pax6os1)
   Uncharacterized protein Rab8a 16.35 1 100
   Eukaryotic translation initiation factor 3 subunit D 1.64 1 52.13
   Histone H4 51.46 6 2.73
   Histone H1.0 23.71 4 4.03
   Inosine-5′-monophosphate dehydrogenase 3.89 2 5.71
   Regulator of G-protein signaling 1 3.40 1 2.29
   Histone H3.2 18.23 4 2.38
   Histone H2B 27.40 5 2.18
   Core histone macro-H2A.1 25.80 8 1.94
   Short-/branched-chain-specific acyl-CoA dehydrogenase, mitochondrial 1
   4.95 2 9.93
   Histone H1.1 32.86 7 1.45
   Pyrroline-5-carboxylate 3.43 1 1.30
   Uncharacterized protein (fragment) Rpn2 2.12 1 1.36
   DEAD (Asp-Glu-Ala-Asp) box polypeptide 23 5.25 4 2.07
   3′-5′ RNA helicase YTHDC2 0.62 1 1.39
   Spectrin beta chain, non-erythrocytic 1 0.76 2 1.98
   Histone H1.4 50.68 5 1.31
   Trifunctional enzyme subunit beta, mitochondrial 8.42 4 1.48
   [121]Open in a new tab

Pax6os1 deletion does not impact glucose homeostasis in T2D mouse models

   In order to explore the possible consequences of Pax6os1 loss for
   insulin secretion and glucose homeostasis in vivo, we used CRISPR/Cas9
   gene editing to delete exon 1 of Pax6os1 plus the immediate 5′ flanking
   region from the mouse genome in C57BL/6 mice ([122]Figure 3A). Analysis
   of super-low input carrier-cap analysis of gene expression (SLIC-CAGE)
   data (unpublished) in mouse islets identified independent transcription
   start sites (TSS) for Pax6 and Pax6os1, located ∼1 kb apart
   ([123]Figure S3 related to [124]Figure 3). Thus, the deletion generated
   spanned only the Pax6os1 TSS and its putative promoter, as suggested by
   the presence of accessible chromatin in this region (ATAC-seq data) and
   of H3K4me3 ^24 and H3K27Ac[125]^24 chromatin marks ([126]Figure S3
   related to [127]Figure 3). Although Pax6os1 expression was lowered
   by > 95% in islets from knockout (KO) mice, Pax6 mRNA levels were
   unaffected ([128]Figure 3B).

Figure 3.

   [129]Figure 3
   [130]Open in a new tab

   Pax6os1female null mice display mildly improved glucose tolerance and
   normal insulin secretion compared to WT animals under HFD

   (A) Schematic representation of the mutation generated in the Pax6os1
   locus through CRISPR gene editing.

   (B) Pax6os1 and Pax6 expression in islets isolated from wt (+/+),
   Pax6os1 heterozygous (+/−), and Pax6os1 homozygous (−/−) mice.

   (C) Body weights (g) of male Pax6os1 null mice under HFD.

   (D and E) Circulating glucose levels and insulin in plasma after
   receiving an intraperitoneal load of glucose in male Pax6os1 null mice.
   Glucose: wt (n = 11), Pax6os1 +/− (n = 22), Pax6os1 −/− (n = 6);
   insulin: wt (n = 9), Pax6os1 (n = 4).

   (F) As (C) but in female mice.

   (G and H) As (D) and (E), respectively, but in female Pax6os1 null
   mice. Glucose: wt (n = 6), Pax6os1 +/− (n = 18), Pax6os1−/− (n = 8);
   insulin wt (n = 6), Pax6os1 −/− (n = 8).

   (I) Insulin secreted (represented as % of the total) at different
   glucose concentrations and after depolarization with KCl in pancreatic
   islets isolated from male Pax6os1 null mice. n = 10–5.

   (J) Intracellular calcium in pancreatic islets isolated from male
   Pax6os1 null mice. n = 3.H.

   (K and L) As (I) and (J) for female mice. n = 5–9 (I); n = 3 (J). Data
   are represented as the mean ± SEM. ∗p < 0.05; two-way ANOVA repeated
   measurements.

   No statistically significant differences were observed in vivo between
   wild-type (WT) and Pax6os1 KO male mice in weight, glucose clearance,
   insulin secretion under standard (STD) ([131]Figure S4 related to
   [132]Figure 3), or HFD ([133]Figures 3C–3E). Similarly, no significant
   differences were observed between wild-type or Pax6os1 KO female mice
   in STD diet in weight, glucose clearance, or insulin plasma levels
   ([134]Figure S4 related to [135]Figure 3). In contrast, Pax6os1 KO
   female (but not male) mice under HFD displayed a tendency toward
   reduced body weight ([136]Figure 3F) and significantly lower
   circulating glucose at 30 min (p = 0.041) during the IPGTT, with a
   strong trend toward a lower AUC during the experiment (WT: 1212 ± 169
   a.u. vs. Pax6os1 KO: 1030 ± 134 a.u. p = 0.069) ([137]Figure 3G).
   However, no differences were observed in insulin secretion in vivo
   ([138]Figure 3H). Furthermore, there were no significant differences in
   GSIS or intracellular calcium dynamics between islets isolated from
   Pax6os1 KO mice and WT independently of sex in STD ([139]Figure S4
   related to [140]Figure 3) or HFD ([141]Figure 3I–3L).

   In order to explore further the tendency observed in female mice in
   another animal model of T2D, we treated db/db male and female mice with
   antisense oligonucleotides (ASOs) targeting Pax6os1.[142]^25 A
   significant and near-significant downregulation of Pax6os1 in
   pancreatic islets could be observed after 4 weeks of treatment with
   ASOs in female and male mice, respectively ([143]Figure S5 related to
   [144]Figure 3). However, no significant differences were observed in
   body weight or glucose clearance between the different experimental
   groups ([145]Figures S5C–S5H).

PAX6-AS1 depletion in EndoC-βH1 cells induces the expression of beta cell
signature genes

   In order to determine the effect of PAX6-AS1 KO in human beta cells, we
   used a tailored CRISPR/Cas9 approach to delete ∼80 bp within the first
   exon of PAX6-AS1 from fetal-human-pancreas-derived EndoC-βH1 cells
   ([146]Figure S6 related to [147]Figure 4). Despite previously mentioned
   differences between PAX6-AS1 and its ortholog in mouse, an RNA-seq
   analysis revealed several genes commonly modulated after silencing the
   lncRNA in MIN6 and EndoC-βH1 cells. Indeed, PAX6-AS1-depleted EndoC-βH1
   cells displayed increased expression of the beta cell signature genes,
   INS, PDX1, and NKX6-1 ([148]Figures 4A and 4B). However, no differences
   were observed in the expression of LDHA and SLC16A1, although the
   disallowed gene SMAD3 was downregulated in PAX6-AS1 KO compared to
   control cells. Further supporting the hypothesis that PAX6-AS1
   depletion favors the expression of beta cell signature genes over genes
   typically expressed in other endocrine cell types, somatostatin (SST)
   expression was robustly reduced in PAX6-AS1 KO cells ([149]Figures 4A
   and 4B). Interestingly, although the calcium channels CACNA2D1 and
   CACNA2D2 were downregulated in PAX6-AS1-depleted cells ([150]Figures 4A
   and 4B), calcium signaling appeared as one of the KEGG pathways
   significantly activated after PAX6-AS1 depletion due to the
   upregulation of RYR2 and FGF5 ([151]Figures 4C and 4D). Other activated
   pathways included protein processing in the endoplasmic reticulum and
   the mitogenic secondary branch of insulin signaling Ras-MAPK
   ([152]Figures 4C and 4D), whereas cAMP signaling and ECM-receptor
   interaction pathways were suppressed ([153]Figures 4C and 4D).
   Increased expression of INS (2.727 ± 0.6649-fold change, p = 0.01),
   PDX1 (0.2625 ± 0.07348-fold change, p = 0.01), and NDUFS6 (0.5690 ±
   0.1227-fold change, p = 0.003) was confirmed by RT-qPCR in
   PAX6-AS1-depleted cells ([154]Figure 4E). Intriguingly, several genes
   that did not appear to be significantly modulated by PAX6-AS1 depletion
   in our RNA-seq data, such as PAX6 (p = 0.009; p-adj = 0.15), showed
   increased mRNA levels in PAX6-AS1 KO compared to control cells when
   measured by RT-qPCR (1.60 ± 0.16-fold change, p = 0.002). However, no
   differences in PAX6 protein levels were observed as measured by western
   blot or immunofluorescence ([155]Figures 4F–4I), indicating that PAX6
   changes are unlikely to underlie the phenotype observed after PAX6-AS1
   deletion. In contrast, PAX6-AS1 KO cells displayed a strong tendency
   toward higher protein insulin levels when measured by western blot
   (0.5438 ± 0.23-fold change, p = 0.057) ([156]Figures 4F and 4G), which
   was significant when measured by immunofluorescence ([157]Figures 4H
   and 4I).

Figure 4.

   [158]Figure 4
   [159]Open in a new tab

   CRISPR/Cas9-mediated PAX6-AS1 deletion in EndoC-βH1 cells increases
   insulin expression and enhances GSIS

   (A) Volcano plot representing genes significantly modulated in PAX6-AS1
   depleted vs. control EndoC-βH1 cells.

   (B) Heatmap representing the log expression for selected genes.

   (C and D) Dotplot and cnetplot depicting the KEGG pathways
   significantly modulated in PAX6-AS1 KO EndoC-βH1 cells.

   (E) mRNA expression of PAX6-AS1, beta cell signature genes, and markers
   from other endocrine cell lineages in PAX6-AS1-deleted EndoC-βH1 cells.
   PAX6-AS1, n = 6; PAX6, n = 6; INS, n = 4; GLUT2/SLC2A2, n = 5; PDX1,
   n = 4; GHRL, n = 4; NEUROG3, n = 3; NEUROD1, n = 5; LDHA, n = 4.

   (F) Western blot showing Pax6 and insulin protein levels.

   (G) Densitometric analysis for (F).

   (H) Representative immunofluorescence images of control and PAX6-AS1 KO
   EndoC-BH1 cells stained for PAX6 and insulin. Scale bar: 100 μm.

   (I) Mean intensity for PAX6 and insulin staining.

   (J and K) Proliferation in PAX6-AS1-deleted EndoC-βH1 cells assessed by
   EdU staining: representative images (J) and quantification (K, n = 5).
   At least 1,000 cells were counted per experiment using ImageJ software.
   Scale bar: 100 μm.

   (L) Raw traces of mixed sodium and calcium currents elicited by a
   100 ms depolarization from −70 to 0 mV in PAX6-AS1 KO and control
   cells.

   (M–O) Quantification of the current density (see [160]STAR Methods) at
   the peak, which is mainly composed of the rapidly inactivating
   voltage-gated sodium and calcium currents (M), at 5 ms when only the
   calcium component remains (the sodium component is inactivated) (N),
   and for the sustained component (O). n = 15.

   (P) Cell size for control and PAX6-AS1 KO EndoC-βH1 cells. n = 15 cells
   in both cell types.

   (Q) Cumulative exocytosis was determined upon 10 depolarizations
   (pulses) from −70 to 0 mV (top panel) using membrane capacitance
   measurement.

   (R) Quantification of the cumulative exocytosis at each pulse. n =
   14–13.

   (S and T) Fold change of glucose-induced insulin secretion (S) and
   insulin secreted as percentage of total content (T). n = 8.

   (U) Insulin secretion induced by 17 mM glucose and with addition of
   35 mM KCl and 25 μM forskolin (FRSK).

   (V) Determination of total insulin content per well. n = 8.

   (W and X) Representative images showing calcein (green) and propidium
   iodide (red) staining and quantification of the percentage of
   propidium-iodide-positive cells. n = 3. At least 1,000 cells were
   counted per experiment using ImageJ software. Scale bar, 100 μm.

   (Y) MTT assay. n = 5. Data are represented as the mean ± SEM.
   ∗p < 0.05; Student’s t test or two-way ANOVA.

   In spite of the activation of the mitogenic Ras-MAPK pathway, no
   differences were observed in the proliferation between control and
   PAX6-AS1 KO EndoC-βH1 cells as determined by EdU staining
   ([161]Figures 4J and 4K). Similarly, no significant differences were
   found in calcium currents between different cell types, which displayed
   similar cell size ([162]Figures 4L–4P). A strong tendency toward
   reduced exocytosis consistent with cAMP pathway suppression could also
   be observed in PAX6-AS1 KO cells ([163]Figures 4Q and 4R). In contrast,
   PAX6-AS1 depletion slightly enhanced GSIS in EndoC-βH1 cells as
   determined by an increased fold change in insulin secretion between
   0.5 mM and 17 mM glucose (Control: 1.584 ± 0.13, PAX6-AS1 KO: 1.894 ±
   0.19; p = 0.041) ([164]Figures 4S and 4T) but not when cells were
   directly depolarized with KCl ([165]Figure 4U).

   Remarkably, the expression of SVB2 ([166]Figures 4A and 4D), which is
   involved specifically in the exocytosis of
   GABA-containing synaptic-like microvesicles but not in insulin
   release,[167]^26^,[168]^27 was reduced in PAX6-AS1 KO cells, whereas
   other genes involved in acidification and vesicular trafficking such as
   ATPV0A2[169]^28^,[170]^29 were upregulated ([171]Figures 4A and 4B).
   The improvement in GSIS was also accompanied by a strong tendency
   toward increased insulin content ([172]Figure 4V) with no variations in
   cell number as suggested by the lack of significant differences in
   proliferation ([173]Figures 4J and 4K) or cell death ([174]Figures 4W
   and 4X). Furthermore, PAX6-AS1-depleted cells displayed increased
   mitochondrial activity as indicated by MTT assay ([175]Figure 4Y) and
   the upregulation of NDUFS6 and IMMP1L ([176]Figures 4B and 4E).

PAX6-AS1 directly interacts with histones and EIF3D

   In order to identify the molecular mechanisms of action of PAX6-AS1, we
   next sought to determine its subcellular localization in EndoC-βH1
   cells. Remarkably, PAX6-AS1 displayed the same expression pattern than
   Pax6os1, being located in the nucleus and the cytoplasm at similar
   proportions (∼40% and ∼60%, respectively) ([177]Figure 5A). Next, we
   sought to validate in the human cell line the binding partners
   previously identified by mass spectrometry in MIN6 cells. To this end,
   we performed an RNA antisense pull-down (RAP), using biotinylated DNA
   probes antisense to our lncRNA followed by western blot analysis
   ([178]Figure 5B).[179]^30 Successful RNA pull-down was confirmed
   by RT-qPCR in the RNA elution fraction, obtaining a ∼40% PAX6-AS1
   enrichment vs. input using four specific probes targeting our lncRNA
   ([180]Table S4 related to [181]Figure 5) ([182]Figure 5C). In contrast,
   only 1.5% PAX6-AS1 enrichment vs. input was observed in the control
   group hybridized with probes targeting luciferase. Both groups showed
   <0.002% β-ACTIN enrichment vs. input, confirming the specificity of our
   probes ([183]Figure 5C). A direct interaction between PAX6-AS1 and
   EIF3D, H3, as well as H4 was confirmed by western blot
   ([184]Figure 5D), whereas direct binding of the lncRNA to other
   partners previously identified such as H1 could not be confirmed (data
   not shown). These results suggest that PAX6-AS1 may regulate protein
   translation[185]^31 as well as transcription of target genes. In line
   with these results, PAX6-AS1 seemed to regulate INS expression at the
   transcriptional level as determined by increased levels of INS nascent
   mRNA and the lack of significant differences in mRNA stability between
   control (INS half-life: 11.29 ± 4.21 h) and PAX6-AS1-depleted cells
   (INS half-life: 9.04 ± 1.7 h) (p = 0.46) after treatment with
   actinomycin D (5 μg/mL) ([186]Figures 5E and 5F).

Figure 5.

   [187]Figure 5
   [188]Open in a new tab

   PAX6-AS1 directly interacts to EIF3D, H3, and H4

   (A) Subcellular localization of PAX6-AS1 in EndoC-βH1 cells. n = 3.

   (B) Schematic representation of the RNA antisense pull-down. (C)
   PAX-AS1 mRNA enrichment vs. input in cells hybridized with probes
   targeting the lncRNA.

   (D) Western blots showing EIF3D, H3, and H4 in cells hybridized with
   probes against our lncRNA or luciferase.

   (E) Mature and nascent insulin mRNA expression as determined by qPCR in
   control and PAX6-AS1 depleted cells. n = 5.

   (F) Determination of insulin mRNA stability as determined by
   actinomycin D treatment in control and PAX6-AS1 KO cells. n = 6–3. Data
   are represented as the mean ± SEM. ∗p < 0.05, Student’s t test.

PAX6-AS1 knockdown enhances, whereas overexpression impairs, GSIS from human
islets

   To extend our results to fully differentiated human beta cells, we used
   lentiviral shRNA vectors to silence PAX6-AS1 in pancreatic islets from
   postmortem donors ([189]Table S1, related to [190]Figures 1 and
   [191]6). Transduced islets displayed a reduction in PAX6-AS1 expression
   of 49 ± 12%, increased INS mRNA levels (2.727 ± 0.6649-fold change, n =
   5, p = 0.04), and reduced GHRL expression (0.57 ± 0.05-fold change,
   p < 0.0001) ([192]Figure 6A). Importantly, the upregulation in INS mRNA
   levels was accompanied by enhanced GSIS (Scrambled: 3.44 ± 0.74-fold
   change vs. PAX6-AS1 shRNA: 6.69 ± 1.78-fold change, n = 5, p = 0.03),
   although total insulin content was not affected ([193]Figures 6B–6D).
   PAX6-AS1-silenced islets also showed increased intracellular Ca^2+
   dynamics in response to 17 mM glucose as assessed by the AUC for mean
   fluorescence (scrambled: 13.09 ± 0.16 a.u. vs. PAX6-AS1 shRNA: 13.69 ±
   0.10 a.u., p = 0.049, paired t test) ([194]Figures 6E and 6F). In
   contrast, Ca^2+ responses to plasma membrane depolarization with KCl,
   added to open voltage-gated Ca^2+ channels directly, were not
   significantly affected by PAX6-AS1 silencing ([195]Figure 6G). No
   additional effect was observed on glucagon secretion elicited by 1 mM
   glucose, suggesting that PAX6-AS1 downregulation does not affect alpha
   cells ([196]Figure 6H).

   Demonstrating a deleterious effect on human beta cell function,
   PAX6-AS1 overexpression in human islets led to a strong reduction in
   INS (0.29 ± 0.08-fold change, p < 0.0001) and PDX1 (0.54 ± 0.14-fold
   change, p = 0.008) expression, whereas GHR was not affected
   ([197]Figure 6I). The decrease in the expression of beta cell signature
   genes was accompanied by impaired GSIS (Control: 3.32 ± 0.5-fold change
   vs. PAX6-AS1 overexpression: 1.89 ± 0.41-fold change, p = 0.02) but
   unaltered total insulin content ([198]Figures 6J–6L). Correspondingly,
   PAX6-AS1-overexpressing islets displayed a significant reduction in
   intracellular Ca^2+ dynamics in response to 17 mM glucose (Control:
   16.05 ± 1.03 a.u. PAX6-AS1 overexpression: 15.01 ± 1.085, p = 0.01,
   paired t test) ([199]Figures 6M and 6N), whereas there were no
   significant differences in the response to depolarization with KCl
   ([200]Figure 6O). In line with silencing experiments, overexpressing
   PAX6-AS1 in islets did not significantly affect alpha cell
   functionality. Nevertheless, a tendency toward decreased glucagon
   secretion could be observed, suggesting that although this lncRNA is
   not normally expressed in alpha cells, forcing its expression can be
   detrimental in all pancreatic endocrine cells ([201]Figure 6P).

Discussion

   We show that Pax6os1/PAX6-AS1, a lncRNA transcribed from the Pax6 locus
   and previously identified in the murine retina,[202]^18 is chiefly
   expressed in beta cells within the pancreatic islet. This distribution
   differs from that of Paupar, also expressed from the Pax6 locus, which
   is largely confined to alpha cells.[203]^17 We demonstrate that Pax6os1
   is upregulated at high glucose concentrations, in an animal model of
   T2D (HFD) and in pancreatic islets from patients with this disease.
   Thus, it is tempting to speculate that increased expression of PAX6-AS1
   contributes to the pathogenesis of T2D. Further detailed studies will
   be needed to determine how, at the molecular level, glucose or other
   factors that contribute to this disease, such as fatty acids, affect
   Pax6os1/PAX6AS1 expression.

   Supporting this hypothesis, Pax6os1 silencing increased the expression
   of several beta cell signature genes in murine MIN6 cells, while
   decreasing mRNA levels of disallowed genes, suggesting a role for the
   lncRNA in beta cell identity. However, no significant differences were
   found in GSIS in vitro. Additional experiments, including an
   exploration of chromatin accessibility and transcription factor binding
   to relevant genomic sites, ultrastructural, proteomic, or other studies
   may nevertheless be useful in the future to more fully explore the
   impact of Pax6os1 on beta cell identity. Mice in which Pax6os1 was
   deleted in utero displayed only a modest phenotype, which was only
   evident in females maintained on an HFD diet.

   Dissecting the functions of different transcripts within complex loci
   such as Pax6/Pax6os1 is inherently challenging due to the close
   proximity of the transcriptional start sites. In this regard, it was
   conceivable that the DNA fragment deleted from the mouse genome
   (720 bp) by CRISPR/Cas9 to generate the Pax6os1 KO mouse might
   interfere with a regulatory region of Pax6, directly affecting the
   expression of the transcription factor. Indeed, assessment of the open
   chromatin state (by ATACSeq), and regulatory histone marks, indicated
   that the deletion of Pax6os1 exon 1 and the proximal promoter region
   might potentially exert an effect on Pax6 expression in cis.
   Furthermore, it was also possible that early developmental compensation
   occurred in vivo after Pax6os1 deletion in the mouse, minimizing the
   effects on beta cell function. However, db/db mice with Pax6os1
   silenced by antisense oligonucleotides at 6 weeks of age showed no
   phenotype in glycemic control, arguing against an effect mediated by
   direct alterations in Pax6 levels by our CRISPR technique or an early
   developmental compensation. Furthermore, the effects of Pax6os1
   inactivation were only observed in female mice, whereas Pax6 changes
   are supposed to affect both sexes equally. The absence of phenotype in
   db/db female mice might be explained due to the fact that treatment was
   started when mice were already hyperglycemic at fasting.

   In line with our results in MIN6 cells and in female mice, deletion of
   PAX6-AS1 in the human beta cell line, EndoC-βH1, upregulated the
   expression of several beta cell signature genes. Nevertheless,
   important differences were found between mouse and human cell lines.
   For instance, insulin was stronger regulated by PAX6AS1 in humans,
   which could partially explain the different phenotype observed between
   the two species in GSIS after deletion of the lncRNA. Actually, protein
   insulin levels were only altered in the human cell line. Therefore, the
   lack of differences in insulin protein levels could underlie, at least
   to some extent, the absence of a major phenotype in mice in vivo. In
   addition, PAX6 in humans was not robustly modulated, finding only
   significant differences in the RT-PCR but not in the RNA-seq or protein
   levels. These discrepancies between diverse methods could be partially
   due to differences in sample size and statistical power. However, it
   could also indicate that PAX6 and other genes may be only slightly
   modulated by PAX6AS1 depending on the cell-cycle state. Despite
   tendencies toward decreased calcium currents and decreased exocytosis,
   GSIS was slightly enhanced by PAX6-AS1 deletion in EndoC-βH1 cells.
   This improvement in insulin release was accompanied by an apparent
   increase in mitochondrial activity and gene expression, which was not
   observed in mice, representing another important difference between
   species that could affect GSIS. Enhanced GSIS was also observed after
   lentivirus-mediated PAX6-AS1 knockdown in human islets, and this was
   accompanied by increased calcium dynamics in response to glucose. In
   contrast, islets overexpressing the lncRNA displayed impaired GSIS and
   cytosolic calcium dynamics. We note that PAX6-AS1 is expressed in delta
   cells, albeit at lower levels than in beta cells, and that SST
   expression was strongly reduced in EndoC-βH1 cells. Therefore, impaired
   somatostatin secretion from delta cells after PAX6-AS1 knockdown in
   islets may contribute to the direct effects of inactivating the lncRNA
   in the beta cell, further enhancing GSIS.

   RNA pull-down experiments revealed that, among other partners,
   Pax6os1/PAX6-AS1 binds the histones H3 and H4. Furthermore,
   Pax6os1/PAX6AS1 was shown to also bind EIF3D, which is a driver of
   noncanonical-cap-dependent translation of specific mRNAs. These results
   suggest that Pax6os1/PAX6AS1 regulates the expression of target genes
   by modulating transcription through epigenetic modulations as well as
   translation of specific mRNAs. Interestingly, in humans, EIF3D has been
   shown to be especially important for the translation of proteins
   involved in metabolism and metabolic stress caused by glucose
   deprivation.[204]^31 Therefore, it is tempting to speculate that some
   of the metabolic alterations observed in human cells and islets after
   PAX6AS1 silencing or overexpression are mediated by the interaction
   between PAX6AS1 and EIF3D.

   Our data suggest that Pax6os1/AS1 affects beta cell signature genes in
   both mice and humans and that they share several binding partners.
   Nevertheless, important species differences were also found, with more
   marked effects observed in humans than in mice. Interestingly, the
   effects of Pax6os1 deletion were sex-dependent in mice. Whether such
   differences also pertain in humans could not readily be explored here
   given the relatively small number of islet samples available and the
   use a cell line (EndoC-βH1) from only one sex (female).[205]^32

   LncRNAs have emerged in recent years as promising therapeutic targets
   in several diseases.[206]^33^,[207]^34 We show here that PAX6-AS1
   silencing in islets enhances insulin secretion, whereas increased
   expression of PAX6-AS1—as observed in T2D—may contribute to beta cell
   dysfunction and impaired GSIS. Although levels achieved in
   overexpression experiments were beyond the (patho-) physiological
   range, even in overexpressing cells, PAX6-AS1 mRNA levels were still
   relatively low compared to most mRNAs, and thus a non-specific toxic
   effect seems unlikely. Targeting PAX6-AS1 might therefore provide a
   novel approach to maintain beta cell functionality in T2D. Experiments
   that explore the impact of modulating PAX6-AS1 expression in islets
   from T2D patients may provide useful insights into the possible
   clinical value of this approach.

Limitations of the study

   One of the limitations of this study is the high level of
   overexpression of PAX6-AS1 achieved in islets, which do not represent
   physiological levels of the lncRNA. However, although some of the
   observed alterations in mRNA levels may be caused by the extraordinary
   high levels of PAX6-AS1, it is important to note that other significant
   changes such as impairment in GSIS and reduced cytoskeletal calcium
   dynamics are also observed in the opposite direction, with only a 40%
   reduction in the expression of this lncRNA. Another important caveat of
   this study is the mild phenotype observed in mice, which points to
   important species differences, but strongly limits the possibilities of
   studying the role of this lncRNA in a physiological setting in vivo.

Resources availability

Lead contact

   Further information and requests for resources should be directed to
   and will be fulfilled by the lead contact, Professor Guy A. Rutter
   (g.rutter@imperial.ac.uk).

Materials availability

   Pax6os1 KO mice are conserved as frozen embryos at Imperial College
   facilities and available upon request.

Data and code availability

     * •
       RNA sequencing data have been deposited at ENA (MIN6) or SRA
       (EndoC-βH1) and are publicly available as of the date of
       publication. Proteomic data have been deposited at PRIDE. Data
       accession numbers are described in the [208]key resources table.
     * •
       No original code is reported in this manuscript.
     * •
       All other data reported in this paper and any additional
       information required to reanalyze it will be shared by the
       [209]lead contact upon request.

Acknowledgments