Roles for the long non-coding RNA Pax6os1/PAX6-AS1 in pancreatic beta cell identity and function

Aim/Hypothesis Long non-coding RNAs (lncRNAs) are emerging as crucial regulators of beta cell development and function. Here, we investigate roles for an antisense lncRNA expressed from the Pax6 locus (annotated as Pax6os1 in mice and PAX6-AS1 in humans) in beta cell identity and functionality. Methods Pax6os1 expression was silenced in MIN6 cells using siRNAs and changes in gene expression were determined by RNA sequencing or qRT-PCR. Mice inactivated for Pax6os1 and human PAX6-AS1-null EndoC-βH1 cells, were generated using CRISPR/Cas9 technology. Human islets were infected with lentiviral vectors bearing a targeted shRNA or PAX6-AS1, which were used to silence or overexpress, respectively, the lncRNA. RNA sequencing or RT-qPCR were used to measure transcriptomic changes and RNA pulldown in mice and human cells followed by mass spectrometry/western blot were performed to explore RNA protein interactions. Results Pax6os1/PAX6-AS1 expression was upregulated at high glucose concentrations in derived beta cell lines as well as in mouse and human islets, and in pancreatic islets isolated from mice fed a high fat diet (n=6, p=0.003) and patients with type 2 diabetes (n=11-5, p<0.01). Silencing or deletion of Pax6os1/PAX6-AS1 in MIN6 or EndoC-βH1cells increased the expression of several β-cell signature genes, including PDX1 and INS. Female, but not male, Pax6os1 null mice fed a high fat diet showed slightly enhanced glucose clearance. ShRNA-mediated silencing of PAX6-AS1 in human islets robustly increased INS mRNA, enhanced glucose-stimulated insulin secretion and calcium dynamics, while overexpression of the lncRNA exerted opposing effects. Pax6os1/AS-1 interacted with histones H3 and H4 in mouse and human cells, indicating a possible role for this lncRNA in histone modifications in both species. Conclusions Increased expression of PAX6-AS1 at high glucose levels may impair beta cell functionality and thus contribute to the development of type 2 diabetes. Thus, targeting PAX6-AS1 may provide a promising strategy to enhance insulin secretion and improve glucose homeostasis in this disease. Research in context What is already known about the subject? Long non-coding RNAs (lncRNAs) are crucial components of the pancreatic islet regulome, whose misexpression may contribute to the development of diabetes. What is the key question? Is the lncRNA Pax6os1/PAX6-AS1 involved in beta cell functionality and type 2 diabetes? What are the new findings? The expression of Pax6os1/PAX6-AS1 is upregulated in mice fed a high fat diet and in pancreatic islets from type 2 diabetes donors. Overexpression of PAX6-AS1 in human pancreatic islets reduces insulin expression, glucose stimulated secretion and intracellular calcium dynamics. Silencing PAX6-AS1 in human pancreatic islets upregulates insulin expression, enhances glucose stimulated insulin secretion and increases intracellular calcium dynamics. How may this impact the clinic in the foreseeable future? Understanding the genetic factors induced by high glucose/obesity involved in beta cell dysfunction is crucial for the development of new therapies to treat T2D.


Introduction
Type 2 Diabetes (T2D) usually develops when β-cells within the pancreatic islet no longer secrete sufficient insulin to overcome insulin resistance and thus lower circulating blood glucose levels. A vicious circle may then ensue leading to further β-cell failure, more severe hyperglycaemia and ultimately disease complications 1 . In a subset of T2D patients, defective insulin secretion is observed despite near-normal insulin sensitivity 2 . In all forms of the disease, changes in β-cell "identity" are now thought to play an important role in functional impairment and the selective loss of glucose responsiveness 3 . A deeper understanding of the mechanisms that influence these changes is therefore likely to facilitate the discovery of new treatments 4 .

Loss of normal β-cell function is often characterized by decreased expression of insulin (INS)
as well as of genes critical for glucose entry (the glucose transporters GLUT1/SLC2A1, GLUT2/SLC2A2) and metabolism (e.g. Glucokinase, GCK) 5 3 . These changes may be accompanied by increased expression of so-called "disallowed genes", whose levels are unusually low in healthy βcompared to other cell types 6 . In several models of diabetes, the above changes are associated with decreased expression of transcription factors that are required to maintain a mature β-cell phenotype, including pancreatic duodenum homeobox-1 (PDX1) 7 and MAF BZIP Transcription Factor A (MAFA) 7,8 .
Remarkably, embryonic deletion of Pax6 in the murine pancreas leads to a drastic reduction in the number of αand β-cells, resulting in the death of mutant mice at postnatal days 3-6 due to severe hyperglycaemia 12 . Furthermore, conditional inactivation of Pax6 in adult mice leads to impaired β-cell function and glucose intolerance 12,13 , demonstrating the continued importance of this gene in the mature β-cell. Moreover, Pax6 levels are reduced in pancreatic islets from Zucker diabetic fatty (ZDF) 14 rats as well as in pregnant rats fed with high fat diet 14, 15 . In humans, loss-of-function mutations in PAX6 are associated with aniridia (iris hypoplasia) and T2D 16,17 . In addition, decreased PAX6 expression in human pancreatic islets correlates with impaired insulin secretion and increased glycated haemoglobin (HbA1c) 18 . Thus, a better understanding of how PAX6 expression is regulated may provide useful insights into the mechanisms involved in T2D pathogenesis.
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 mis-expression may also contribute to the development of T2D 19 . Importantly, lncRNAs are expressed in a highly cell-type specific manner, making them well placed to be involved in cell lineage specification. More than 1,100 lncRNAs have been identified in both human and murine pancreatic islets. Furthermore, the expression of several of these is modulated by high glucose concentrations, suggesting that they may be involved in β-cell compensation in response to high insulin demand 20 . Interestingly, a number of β-cell-enriched lncRNAs are mapped to genetic loci in the proximity of β-cell signature genes, such as PDX1, and regulate their expression in cis 20,21 .
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 22 , might impact β-cell identity and/or function through the modulation of Pax6 expression or by other mechanisms. We show that Pax6os1 expression is enriched in islets and, more specifically, in β-cells. Furthermore, Pax6os1/PAX6-AS1 expression was upregulated in pancreatic islets from mice challenged with high fat diet (HFD) as well as in human islets from type 2 diabetic donors. Importantly, we also show that silencing or inactivation of Pax6os1 and PAX6-AS1 in mouse and human β-cells, respectively, upregulates several β-cell signature genes and enhances insulin production.

Pax6os1/PAX6-AS1 expression is enriched in pancreatic islets and is upregulated in type 2 diabetes.
The lncRNA Pax6os1/PAX6-AS1 is a 1,464/1,656 nucleotide transcript mapped in 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 α-cells and it is involved in Pax6 splicing 21 . Nevertheless, Pax6os1 is not highly conserved between species at the nucleotide level, containing four exons in mice and three in humans, and predicted secondary structures are different between species ( Figure 1A and Supplemental Figure 1). Assessed across multiple mouse tissues by qRT-PCR analysis, the tissue distribution of Pax6os1 was similar to that of Pax6, being predominantly expressed in pancreatic islets and, to a lesser extent, in the eye and brain ( Figure 1B).
However, despite the upregulation of several β-cell signature genes, β-cell functionality, as determined by glucose-stimulated insulin secretion (GSIS) assays, was not measurably affected by Pax6os1 silencing in MIN6 cells ( Figure 2D-E).

Impact of Pax6os1 deletion on glucose homeostasis in the mouse
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. Analysis of Super-Low Input Carrier-Cap analysis of gene expression (SLIC-CAGE) data (NH, NC, BL, AMS, unpublished) in mouse islets identified independent transcription start sites (TSS) for Pax6 and Pax6os1, located ~1kb apart (Supplemental Figure 2). Thus, the deletion generated only spanned Pax6os1 TSS and its putative promoter, as suggested by the presence of accessible chromatin in this region (ATAC-seq data, unpublished results), and of H3K4me3 23 and H3K27Ac 24 chromatin marks (Supplemental Figure 2). However, whereas Pax6os1 expression was lowered by > 95 % in islets from KO mice, Pax6 mRNA levels were unaffected ( Figure 3A-B).
No statistically significant differences were observed in vivo between wild-type (WT) and Pax6os1 knockout (KO) male mice in weight ( Figure 3C), glucose clearance ( Figure 3E) or insulin secretion ( Figure 3G) under standard (STD) diet at 8-9 weeks of age.

Glucose-stimulated insulin secretion (GSIS) involves uptake of the sugar into β cells via
Glut2/Slc2a2 (plus Glut1 and Glut3 in human islets), enhanced intracellular ATP synthesis and the closure of ATP-sensitive K + channels. This, in turn, leads to the opening of voltage-gated Ca 2+ channels, Ca 2+ influx and the fusion of insulin-containing secretory granules with the plasma membrane 25 . In line with the in vivo findings above, GSIS in vitro and glucosestimulated intracellular Ca 2+ dynamics, assessed using the trappable florescent indicator Cal520, were not significantly different between pancreatic islets from knockout and wild-type male mice under STD (Supplemental Figure 3A-D). Under HFD, male Pax6os1 KO mice displayed normal weight gain, glucose tolerance and insulin secretion in vivo (Supplemental Figure 4A,C,E). In contrast, islets derived from Pax6os1 KO mice under HFD showed enhanced GSIS (n=9-5, p=0.04), while no effect was observed in Ca 2+ dynamics in response to glucose (Supplemental Figure 5A-B and E-F).
Tendencies were seen towards increases in the β-cell-enriched genes G6pc2, Znt8/Slc30a8 and Gipr, and for a lowering of Ghrl expression, whilst Pax6 expression was not affected (Supplemental Figure 6A). These findings were reinforced by Gene Set Enrichment Analysis (GSEA) revealing that 'Pancreas Beta Cells' was the second most upregulated gene set in KO mouse islets, as assessed by Normalised Enrichment Score (Supplemental Figure 6B-C), and suggest that Pax6os1 may exert a weak effect to impair β cell identity in the mouse.

PAX6-AS1 knockdown enhances, whilst overexpression impairs, GSIS from human islets
In order to extend our results to fully differentiated human β-cells, we used lentiviral shRNA vectors to silence PAX6-AS1 in pancreatic islets from post-mortem donors (Table 1). Consistent with the results obtained in EndoC-βH1 cells, a reduction in PAX6-AS1 expression of 49 ± 12%, upregulated INS mRNA levels (3.26 ± 1.05 fold change, n=4, p= 0.04), while GHRL was downregulated (0.57 ± 0.05 fold change, p< 0.0001) and a small tendency towards increased GLUT2/SLC2A2 expression was observed ( Figure 5A). However, no significant differences were observed in the expression of other β-cell signature genes such as PAX6 or MAFA ( Figure   5A). More importantly, while total insulin content was not affected ( Figure 5L).

Pax6os1/PAX6-AS1 modulates the expression of its target genes at the transcriptional level
LncRNAs may regulate gene expression through a number of different mechanisms. These include chromatin remodelling, activation/repression of transcription factors in the nucleus as well as modulation of mRNA/protein stability in the cytoplasm 26,27 . Therefore, the subcellular localization of a lncRNA may provide a reliable indicator of its mechanism(s) of action.
Determinations of Pax6os1 subcellular localization in MIN6 cells (Methods) indicated that this lncRNA was located in both the nucleus (~40%) as well as in the cytoplasm (~60%) ( Figure   6A). Consistent with the above subcellular fractionation results, both nuclear and cytoplasmic proteins were identified by mass spectrometry as binding protein partners of Pax6os1.
Since our results from mass spectrometry suggested that Pax6os1 might regulate the expression of target genes through a number of mechanisms, including chromatin remodelling (Histones H1.0, H4, H3.2, H2B) and RNA translation and stability (Eif3d, 3'-5' RNA helicase YTHDC2), we next sought to determine whether PAX6-AS1 affected the expression of its target genes transcriptionally or post-transcriptionally. To this end, we assessed INS mRNA levels (whose expression was more robustly affected by PAX6-AS1 silencing/overexpression in different biological systems) 24 and 32 h after treatment with actinomycin D (5 µg/ml) 28 in control and PAX6-AS1 KO EndoC-βH1 cells. No significant differences were observed in INS mRNA stability between the different genotypes as assessed with this method ( Figure 6D). In contrast, PAX6-AS1 KO cells displayed increased expression of nascent (intronic) INS mRNA, similar to the increase observed in mature INS mRNA levels ( Figure 6E). Therefore, our results indicate that PAX6-AS1 may regulate the transcription of its target genes, rather than mRNA stability or processing.

DISCUSSION
Identifying the genetic networks that regulate β-cell differentiation and function is essential to understand the pathogenesis of type 2 diabetes and hence help efforts to find novel therapies for this disease.
In recent years, several studies have shown the importance of long non-coding RNAs in the maintenance of β-cell identity 20,28 . In the present study, we show that Pax6os1/PAX6-AS1, a lncRNA transcribed from the Pax6 locus and previously identified in the murine retina 22 , is chiefly expressed in pancreatic islets. Importantly, our results show that the expression of this lncRNA is upregulated in an animal model of type 2 diabetes (high fat diet) as well as in pancreatic islets from patients with this disease, suggesting that increased expression of PAX6-AS1 may contribute to the pathogenesis of T2D.
Although Pax6os1 silencing increases the expression of several β-cell signature genes in MIN6 cells, indicating a role for the lncRNA in β-cell identity, we did not observe differences in functional assays in vitro. Similar to our findings in MIN6 cells, PAX6-AS1 silencing in human EndoC-βH1 cells, upregulated the expression of insulin and other β-cell signature genes without affecting glucose stimulated insulin secretion. In contrast, PAX6-AS1 silencing in human pancreatic islets not only increased insulin mRNA levels but also enhanced GSIS.
Consistent with these results, PAX6-AS1 overexpression in human islets, reduced INS expression and impaired insulin secretion in response to glucose. Importantly, these differences between the impacts of deletion observed in immortalised versus fully differentiated cells (from the same species) might also suggest a different role for Pax6os1/PAX6AS1 at different stages of development/differentiation. Alternatively, the different phenotypes observed in EndoC-βH1 cells and human islets may reflect the different strategies used to reduce PAX6-AS1 levels, since the shRNA is likely to be more effective on the cytosolic lncRNA, while CRISPR deletion will affect equally the expression of PAX6-AS1 in both the cytoplasm and the nucleus.
Strikingly, and despite the clear-cut effects of Pax6os1/PAX6AS1 deletion or silencing in mouse and human β cells in vitro, mice in which Pax6os1 was deleted in utero displayed little evidence of a glycaemic phenotype, or defective insulin secretion. Nevertheless, tendencies were observed to increase the expression of β-cell-enriched genes (Glut, G6pc2, Slc30a8, Gipr) and to further lower the expression of β-cell disallowed genes (Acot7), suggestive of a weakly reinforced β-cell identity 24 .
Dissecting the functions of different transcripts within complex loci such as PAX6/Pax6 is inherently difficult due to the close proximity of the transcriptional start sites. In an effort to mitigate these problems, we have used complementary techniques (shRNA and CRISPR) here to lower the levels of Pax6os1/AS1. It is important to note that the DNA fragment deleted from the mouse genome (720 bp) by CRISPR/Cas9 to generate the Pax6os1 knockout mouse, might interfere with a regulatory region of Pax6, affecting the expression of the transcription factor directly. Such a mechanism might provide an explanation for the differences observed in Pax6 levels between the different systems, i.e. the loss of an inhibitory action of Pax6os1 on the expression of target genes, potentially including Pax6. Indeed, assessment of the open chromatin state (by ATACSeq), and regulatory histone marks, indicated that the deletion of Pax6os1 intron 1 and the proximal promoter region might conceivably exert an effect on Pax6 expression in cis. Nevertheless, we note that since Pax6os1 deletion exerted no effect on Pax6 mRNA levels in mouse islets any negative action on Pax6 transcription due to loss of a cisacting regulatory region would need, for this scenario to be true, to be exactly balanced by the loss of a positive action of Pax6os1 in trans.
A further possible explanation for our findings is that early developmental compensation occurs in vivo after pax6os1 deletion in the mouse, through presently-undefined mechanisms which result in the normalisation of the expression of Ins2 and other genes essential for normal function. Of note, differences in gene expression were also observed between EndoC-βH1 and islets, indicating that PAX6-AS1 may exert different effects depending on the differentiated state of the cell. Of note, the subcellular localizations of other, previously characterized lncRNAs, such as MALAT1, have been shown to be modulated according to the mitotic state of the cell 29 . Therefore, it is possible that the subcellular localization of Pax6os1 and its function are modulated depending on cellular maturity. This could explain the differences observed between immortalised and fully differentiated pancreatic cells.
According to the results obtained by mass spectrometry, Pax6os1 may interact with both nuclear and cytosolic proteins, affecting gene expression by interacting with histones or playing a role in mRNA capping and synthesis. This finding suggests that Pax6os1 could play a dual role, in a manner similar to that observed for other lncRNAs, including PYCARD-AS1 30 .
However, it is important to note that our results also suggest that PAX6-AS1 acts mainly by affecting the expression of its target genes at the transcriptional level.
Finally, our data also support the view that there may be important species differences in the importance of Pax6os1/PAX6AS1. Thus, and despite several similarities observed between the mouse and human models, PAX6-AS1 depletion in human β-cells generally exerted larger effects than in mouse cellsincluding regulation of INS gene expression. Consistent with these results, predicted secondary structures differed between Pax6os1 and PAX6-AS1 (Supplemental Figure 1).
LncRNAs have emerged in recent years as promising therapeutic targets. Indeed, different approaches, including antisense oligonucleotides and small molecule inhibitors are being used to target lncRNAS in several diseases, such as cancer 31 . In the present study, we have shown that increased expression of PAX6-AS1 may drive β-cell dysfunction during the development of T2D, while PAX6-AS1 silencing enhances insulin secretion. Therefore, targeting PAX6-AS1 could be a promising approach to promote β-cell differentiation and improve glucose homeostasis in patients with T2D, although a more extensive study will be required to further decipher the mechanism of action of this lncRNA.
In conclusion, we describe important roles and potential downstream mechanisms of action for a previously uncharacterised lncRNA in pancreatic β-cell function. We demonstrate roles in the mature mouse and human β-cell, and in the control of insulin and other critical genes expression.

Human islets
Human islets were cultured in RPMI-1640 (11879-020) supplemented with 5.5 mM glucose, 10% FBS, 1% penicillin/streptomycin and 0.25 μg/ml fungizone. The characteristics of the donors and isolation centres used for this study are outlined in Supplemental Table 1.

Small interfering RNA Pax6os1 in MIN6
MIN6 cells were transfected with a pool of three small interfering RNAs (siRNAs) targeting Pax6os1 or three control siRNAs (table 2) using Lipofectamine™ RNAiMAX, according to the manufacturer's protocol.

Generation of Pax6os1-null mice.
A Pax6os1 null mouse line was generated using a CRISPR-Cas9 strategy. Guides targeting Pax6os1 for Cas9-mediated CRISPR disruption were identified using online design software at http://crispr.mit.edu. The two targeting sequences were designed to delete ~720 bp including exon 1 of Pax6os1 and upstream region ( Table 2). DNA oligos were synthesized by Sigma and the overlapping PCR products were cloned into a pX330-U6-Chimeric_BB-CBh-hSpCas9 plasmid (Addgene). Guide RNAs were validated in vitro in MIN6 cells and the deletion was confirmed via extraction of genomic DNA from an unselected pool of cells and PCR amplification using primers spanning a 930 bp region, encompassing the targeted region.
Following confirmation that the CRISPR-Cas9 had successfully produced the desired deletion, pronuclear injection of the two chosen gRNAs was performed by the MRC transgenics unit, Imperial College London. F0 compound homozygous males (carrying different mutations in the Pax6os1 gene) were crossed with WT females. F1 heterozygous mice were sequenced to determine which mutation they were harbouring. Heterozygous mice positive for the same mutation were then crossed to generate wild-type, heterozygous and homozygous littermates.

Transduction of pancreatic islets
Human islets were incubated with 1.0 mL of warm (37 ºC) 0.5 X trypsin-EDTA (250 mg/L trypsin; 0.48 mM EDTA) for 3 min. in a cell culture incubator (37 ºC, 5% CO2). Trypsin activity was subsequently inhibited by adding 1.0 mL of RPMI complete medium and islets were centrifuged at 50 x g for 2 min. Afterwards, the supernatant was removed and islets were resuspended in serum free RPMI. Lentiviruses were added at multiplicity of infection (MOI) 20, assuming that a single islet has 1000 cells. Pancreatic islets were incubated over-night in a cell culture incubator (37 ºC, 5% CO2) and the medium was changed to complete RPMI 32  The Imperial BRC Genomics Facility performed sequencing as 75 bp paired end reads on a HiSeq4000 according to Illumina specifications. An average of 37.7 million reads per sample were mapped to mouse genome (GRCm38) using HiSat2 and quantified using featureCounts and Ensembl annotations (v92). Differential expression analysis was performed with DESeq2 using an adjusted p-value threshold of <0.1 [33][34][35] . Gene Set Enrichment Analysis (GSEA) was performed using the fgsea package with MSigDB gene sets provided by the msigdbr package.
The next day, cells were washed with PBS 1X three times for 5 min. each and incubated with matching secondary antibodies conjugated with fluorophores diluted in PBS + 0.2% Tween for 1 h at room temperature (Table 3).

Click-iT EdU (5-ethynyl-2'-deoxyuridine) Proliferation Assay
Cells were fixed with 100% methanol at -20ºC and were labelled for Edu using the Click-iT EdU Alexa Fluor 488 HCS Assay according to manufacturer's instructions and co-stained with insulin. Cells were imaged using a Nikon spinning disk microscope at x20 magnification and counted using ImageJ software. At least 1000 cells were counted per experiment.

Cell viability
Control and PAX6-AS1 null cells were cultured for 30 min. in 1 ml of PBS with Calcein-AM (1 µl) (Molecular Probes, Eugene, OR) and propidium iodide (1 µl) (SigmaAldrich, St Louis, MO, USA). At least 1000 cells were counted for each experiment. Cells were imaged using a Nikon spinning disk microscope at x20 magnification and counted using ImageJ software. At least 1000 cells were counted per experiment.
Complementary DNA (cDNA) was synthesized using random primers (Roche) and the High-

RNA pulldown assay
Pax6os1 or Slc16a1 (control) were cloned using Sequence-and Ligation-Independent Cloning (SLIC) into a ptRNA-S1 plasmid that harbours a T7 promoter, tRNA-S1 (encoding the streptavidin aptamer) and a bovine growth hormone (BGH) polyadenylation site. Primers for

Mass Spectrometry
The pooled SILAC samples were run into an SDS-PAGE gel and each gel lane cut into three slices. Each slice was subjected to in-gel tryptic digestion using a DigestPro automated digestion unit (Intavis Ltd.) and the resulting peptides were fractionated using an Ultimate 3000 nano-LC system in line with an Orbitrap Fusion Tribrid mass spectrometer (Thermo Scientific).
All spectra were acquired using an Orbitrap Fusion Tribrid mass spectrometer controlled by Xcalibur 3.0 software (Thermo Scientific) and operated in data-dependent acquisition mode. FTMS1 spectra were collected at a resolution of 120 000 over a scan range (m/z) of 350-1550, with an automatic gain control (AGC) target of 300 000 and a max injection time of 100ms.
The Data Dependent mode was set to Cycle Time with 3s between master scans. Precursors were filtered according to charge state (to include charge states 2-6) and with monoisotopic peak determination set to Peptide. Previously interrogated precursors were excluded using a dynamic window (40s +/-10ppm). The MS2 precursors were isolated with a quadrupole mass filter set to a width of 1.4m/z. ITMS2 spectra were collected with an AGC target of 20 000, max injection time of 40ms and CID collision energy of 35%.

Secondary structure prediction
Secondary structure prediction was performed using RNAfold from the Viena package and visualized using VienaRNA webservices as well as R-chie webserver 37,38 .

Statistical analysis
For comparisons between two groups, statistical significance was calculated using non-paired two-tailed Student's t-tests. For comparisons between more than two groups, one-way ANOVA or two-way ANOVA tests were performed. For metabolic tests, repeated measurements two-way ANOVA tests were performed. All the statistical analyses were performed using Graph Pad Prism 8.0. In all cases, a p-value < 0.05 was considered statistically significant. Error bars represent the standard error of the mean (SEM). Fluorescence intensity and images analyses were performed using ImageJ software.          P I 3 K -A k t s i g n a l i n g p a t h wa y E r b B s i g n a l i n g P a t h wa y P a t h wa y s i n c a n c e r P r o s t a t e c a n c e r HI F -1 s i g n a l i n g p a t h wa y T y p e I I d i a b e t e s me l l i t u s