LncRNA Malat1 regulates iPSC-derived β-cell differentiation by targeting the miR-15b-5p/Ihh axis

Background: Differentiation of induced pluripotent stem cell (iPSC)-derived β-like cells is a novel strategy for treatment of type 1 diabetes. Elucidation of the regulatory mechanisms of long noncoding RNAs (lncRNAs) in β-like cells derived from iPSCs is important for understanding the development of the pancreas and pancreatic β-cells and may improve the quality of β-like cells for stem cell therapy. Methods: β-like cells were derived from iPSCs in a three-step protocol. RNA sequencing and bioinformatics analysis were carried out to screen the differentially expressed lncRNAs and identify the putative target genes separately. LncRNA Malat1 was chosen for further research. Series of loss and gain of functions experiments were performed to study the biological function of this lncRNA. Quantitative real-time PCR (qRT-PCR), Western blot analysis and immunouorescence (IF) staining were carried out to separately detect the functions of pancreatic β-cells at the mRNA and protein levels. Cytoplasmic and nuclear RNA fractionation and uorescence in situ hybridization (FISH) were used to determine the subcellar location of lncRNA Malat1 in β-like cells. Flow cytometry and enzyme-linked immunosorbent assays (ELISAs) were performed to examine the differentiation and insulin secretion of β-like cells after stimulation with different glucose concentrations. Structural interactions between lncRNA Malat1 and miR-15b-5p and between miR-15b-5p and Ihh were detected by dual luciferase reporter assays (LRAs). Results: We found that the expression of lncRNA Malat1 declined during differentiation, and overexpression of this lncRNA notably impaired the differentiation and maturation of β-like cells derived from iPSCs in vitro and in vivo. Localized to the cytoplasm, lncRNA Malat1 could function as a competing endogenous RNA (ceRNA) of miR-15b-5p to regulate the expression of Ihh according to bioinformatics prediction, mechanistic analysis and downstream experiments. Conclusion: This study established an unreported regulatory network of lncRNA Malat1 and

Noncoding ribonucleic acids (ncRNAs) are not translated or do not encode proteins through standard mechanisms and are considered nonfunctional RNAs [8]. As research has progressed, ncRNAs such as microRNAs (miRNAs), long noncoding RNAs (lncRNAs) and circular RNAs (circRNAs) have been found to participate in cellular signalling and regulation of various processes [9][10][11]. In recent years, as transcripts have been shown to exceed 200 bp, lncRNAs have been proven to control gene regulation, organ differentiation and both developmental and pathological processes [12][13][14]. Moreover, studies have indicated that lncRNAs participate in pancreatic and islet development, β-cell formation and function and T cell differentiation in vitro [15][16][17][18]. Given the above, identi cation of functional lncRNAs in the β-cell differentiation process and exploration of their mechanisms in detail are conducive to optimizing β-cell differentiation protocols and elucidating islet β-cell development in vivo.
Herein, we established a three-step iPSC-derived β-cell differentiation protocol to identify crucial lncRNAs and clarify their molecular mechanisms. RNA sequencing (RNA-seq) was performed at four time points: the iPSC stage and the early, middle and late stages of β-cell differentiation in vitro. The lncRNA metastasis-associated lung adenocarcinoma transcript 1 (Malat1) was highly expressed in iPSCs and decreased over time during β-cell formation. Functional experiments and RNA-seq indicated that Malat1 overexpression affected pancreatic progenitor differentiation by upregulating Indian hedgehog (Ihh), a Hedgehog pathway ligand with low reported levels in developing and mature pancreatic tissues [19,20]. Furthermore, using bioinformatics analysis and biochemical techniques, we demonstrated that Malat1 could act as a competing endogenous RNA (ceRNA) of miR-15b-5p to regulate Ihh expression. To supplement this nding, we performed an in vivo transplantation experiment, which proved that Malat1 overexpression in β-cells could suppress the insulinogenic ability of the pancreas; this nding was consistent with the results of in vitro differentiation.

Cell culture and differentiation
We obtained mouse green uorescent protein (GFP)-iPSCs from Innovative Cellular Therapeutics, Ltd. (Shanghai, China) and cultured them in feeders under mouse ESC (mESC) culture conditions. These mouse iPSCs (miPSCs) were induced to differentiate into β-like cells using a three-step protocol [2]. In brief, we obtained embryoid bodies (EBs) from the GFP + iPSCs in step 1, differentiated them into multilineage progenitor cells (MPCs) in step 2 and induced the MPCs to differentiate into β-like cells using β-cell-selective differentiation medium in step 3.
RNA sequencing and data analysis were performed as previously described [21].
LncRNA-mRNA network and Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses We constructed a lncRNA-mRNA network to ensure interactions between lncRNAs and mRNAs based on the normalized signal intensity of speci c expression in both types of RNA. After calculating Pearson's correlation coe cient (PCC), we built the network from signi cantly correlated pairs. Moreover, GO and KEGG analyses were applied to analyse the main functions of the differentially expressed genes (DEGs).
Then, we used the false discovery rate (FDR) to correct the P-value, as described in detail in our previous study.
RNA extraction and quantitative reverse-transcription polymerase chain reaction (qRT-PCR) analysis Total RNA was isolated using a RNeasy Mini Kit (Qiagen, Düsseldorf, Germany). As described in our previous study [21,22], rst-strand complementary DNA (cDNA) of lncRNAs, mRNAs and miRNAs was synthesized, and their relative expression levels were calculated with the 2 −ΔΔCt method. As an internal normalized control, U6 was used for miRNAs and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used for lncRNAs and mRNAs. The qRT-PCR primer sequences for lncRNAs, mRNAs, miRNAs, U6 and GAPDH were designed and synthesized by RiboBio (Guangzhou, China). We performed all experiments independently and in triplicate.

Flow cytometry (FCM)
After all cells were harvested as single cells, we carried out FCM as described in detail in our previous study [21]. Cells were analysed using a BD LSRFortessa X-20 Flow Cytometer (BD Biosciences, Franklin Lakes, NJ, USA), and the results were analysed using FlowJo software (FlowJo LLC [BD Biosciences], Ashland, OR, USA). Primary antibodies included Alexa Fluor 647 mouse anti-natural killer 6 homeobox 1 (NKX6-1) and Alexa Fluor 647 mouse immunoglobulin G 1k (IgG1k) isotype control (BD Biosciences).

Glucose-stimulated insulin secretion
We transferred β-like cells to 24-well plates and cultured them for 12 h. After preincubation in Krebs-Ringer bicarbonate (KRB) buffer without glucose for 2 h, the cells were stimulated with KRB buffer containing various concentrations of glucose (0, 5, 15 and 30 mM) for 2.5 h. Then, we collected the supernatants. We performed an enzyme-linked immunosorbent assay (ELISA) to assess the insulin content using an ultrasensitive mouse insulin assay kit (Mercodia AB, Uppsala, Sweden) per the manufacturer's instructions.

Western blot (WB)
Western blotting was performed as described previously [21]. Primary and HRP-conjugated secondary antibodies (all from Abcam) were anti-Ihh antibody, anti-β-actin antibody and goat anti-rabbit HRP antibody.
Cytoplasmic and nuclear RNA fractionation Cytoplasmic and nuclear RNAs of mouse MPCs were partitioned using a PARIS Kit (Life Technologies) per the manufacturer's instructions. Then, we analysed cytoplasmic and nuclear RNAs via qRT-PCR as described above.

Fluorescence in situ hybridization (FISH)
This assay was performed with a FISH Kit (RiboBio, Guangzhou China) per the manufacturer's instructions. LncRNA Malat1-Cy3 FISH probes were designed and synthesized by RiboBio. We used mouse U6 and mouse 18S FISH probes as nuclear and cytoplasmic controls, respectively. Images were acquired using the Leica TCS SP8 confocal imaging system.

LncRNA Malat1 lentiviral transduction
GeneChem (Shanghai, China) produced lentiviruses overexpressing lncRNA Malat1 and determined the lentivirus titre. On day 4 of step 3, we seeded iPSC-derived β-like cells in six-well culture dishes and infected them with lentivirus-cytomegalovirus (CMV)-lncRNA Malat1 at a multiplicity of infection (MOI) of 20 or lentivirus-CMV-negative control at an MOI of 40. Seventy-two hours after the cells reached 40-50% con uence, we examined the lncRNA Malat1 expression via qRT-PCR.
In vivo implantation of alginate hydrogel Alginate hydrogel was constructed as described in our previous study [23]. C57BL/6J mice were obtained from the animal centre of Nantong University. All animal experiments were carried out according to the Institutional Animal Care guidelines and were approved by the Animal Ethics Committee of the Medical School of Nantong University. We induced diabetes in mice by intraperitoneal injection of streptozotocin (STZ), as reported previously [2]. At day 21, all groups of β-like cells were digested and resuspended in alginate hydrogel. Next, we injected alginate hydrogel containing β-like cells (1×10 6 cells/1 mL hydrogel) into the mouse renal capsules (n = 6). All operations were performed under sterile conditions, and the mice were sanitized using iodophor. All mice were carefully monitored and well cared for after the procedure.

Statistical analysis
We analysed all between-group results using Student's t-test and GraphPad Prism software version 8.0 (GraphPad Software, Inc., San Diego, CA, USA). All error bars represent the mean ± standard error of the mean (SEM). P < 0.05 was considered statistically signi cant.

Results
LncRNA Malat1 was downregulated during pancreatic β-cell differentiation We obtained pancreatic β-cells from mouse GFP + iPSCs using a three-step protocol (Fig. 1a). First, iPSCs were separated into single cells and then cultured into EBs in a low-adsorption dish. Next, we induced EBs to develop into multilineage progenitors (MPs) for further differentiation. Finally, after 21 days of induction, β-cells differentiated to display a tennis racket morphology and formed numerous clusters (Fig.  1b). After 34 days of differentiation, the nal cell cytoplasm expressed two distinctive markers of pancreatic β-cells, insulin and C-peptide protein (Fig. 1c). Moreover, compared with miPSCs, β-like cells expressed insulin and mature beta-cell markers, such as Pdx1, Nkx6-1, Mafa, ISL, Ngn3, GCG, SST and Glut2, at the mRNA level (Fig. 1d). As in our previous study, this three-step protocol was effective and stable.
To select potentially important lncRNAs that might regulate iPSC-driven β-cell differentiation, we performed RNA-seq to identify differentially expressed lncRNAs at four sequential time points (Fig. 2a). We harvested iPSCs, as well as β-cells, in the early (day 4 of step 3), middle (day 14 of step 3) and late (day 21 of step 3) stages and used poly(A) to analyse mRNAs and lncRNAs. LncRNAs with a >2-fold difference in expression (FDR < 0.05) were selected (Fig. 2b). Of these, more than 100 lncRNAs were downregulated, exhibiting four coincident expression patterns in the series test of cluster (STC) analysis results: pro les 1, 2, 4 and 5 (P < 0.05) (Fig. 2c). After prescreening, we selected pro le 2, which contained 33 transcripts with stable expression across all three stages, for further analysis. Coexpression networks of these lncRNAs and of hundreds of correlated mRNAs were determined; most of them had positive correlations (PCC > 0.99) (Fig. 2e). GO analyses of the mRNAs revealed that pro le 2 lncRNAs might participate in transcriptional regulation, cell proliferation and apoptosis, which are all crucial to stem cell differentiation (Fig. 2d). Interestingly, lncRNA Malat1 was identi ed and was downregulated at all four time points, as con rmed by qRT-PCR (Fig. 3a). As mentioned above, bioinformatics analysis indicated that Malat1 might play an important role in iPSC-derived β-cell differentiation.
Aberrant overexpression of Malat1 disturbed pancreatic β-cell differentiation LncRNA Malat1 was highly expressed in iPSCs but was gradually downregulated during step 3 differentiation. Therefore, we speculated that the relatively low expression of Malat1 contributed to the transformation of the stem cell state to a differentiation state. To verify this hypothesis, we used clustered regularly interspaced short palindromic repeats (CRISPR)-associated protein 9 (CAS9)-Malat1 synergistic activation mediator (SAM) to transfect MPCs and maintain a stable high expression level of Malat1 early in the differentiation process (Fig. 3b). After 72 h of transfection, NKX6-1 + pancreatic endocrine progenitor cell formation underwent a substantial decrease, from 21.4±0.35% to 11.2±0.56% (n=3) (Fig. 3e). qRT-PCR analysis revealed that Malat1 overexpression suppressed the transcription of some key regulators of endocrine progenitors, including Pdx1, Nkx6-1, Ngn3, Gata4/6, NeuroD1, and paired box protein 6 (Pax6) (Fig. 3f). Interestingly, the Hedgehog family member Ihh showed noteworthy upregulation compared with the abovementioned transcription factors (TFs). In addition, IF staining of Pdx1, Nkx6-1, Ngn3 and NeuroD1 indicated that the Malat1-transfected cells showed a defect in pancreatic endocrine-lineage formation (Fig. 3c, d). The abovementioned ndings demonstrated that abnormal upregulation of Malat1 could affect Nkx6-1 + pancreatic endocrine-progenitor differentiation and that a time-ordered decline in Malat1 expression might be necessary for β-cell transformation.

Subcellular location of lncRNA Malat1 in MPCs and target gene prediction
To determine how Malat1 regulated β-cell differentiation in vitro, we performed mRNA-seq to identify the target genes of Malat1. We constructed a Malat1-targeting small interfering RNA (siRNA) for transfection of MPCs. Hierarchical clustering data revealed that 296 genes were downregulated and 364 genes were upregulated within 72 h after Malat1-siRNA transfection (Fig. 4a). Intriguingly, as shown in the volcano plot, the expression level of Ihh decreased, while those of Gata6, motor neuron and pancreas homeobox 1 (Mnx1) and integrin subunit alpha 1 (Igta1) increased, which was consistent with the results of Malat1 overexpression (Fig. 4b). As reported, loss of Ihh can promote pancreatic endocrine lineage development and lead to the malformation known as annular pancreas in humans. To validate the veracity of the abovementioned DEGs, we performed qRT-PCR, and the results correlated well with the mRNA-seq data (Fig. 4d). KEGG analysis of the DEGs revealed that Malat1 in uenced several important pathways in cell differentiation and survival, such as the phosphatidylinositol-4,5-bisphosphate 3-kinase (PI3K)-protein kinase B (Akt) and Ras signalling pathways, as well as cell adhesion molecules (CAMs) (Fig. 4c). The mRNA-seq results suggested that silencing Malat1 might negatively regulate Ihh, which would in turn increase the expression of some key TFs and promote endocrine progenitor differentiation.
If Ihh is the target gene of lncRNA Malat1, what is the detailed molecular mechanism of this regulatory relationship? To address this issue, we rst determined the subcellular location of Malat1. FISH and cell fractionation analysis both indicated that lncRNA Malat1 was mostly located in the cytoplasm (Fig. 5a,  5b), indicating it may regulate Ihh expression via a ceRNA network. In addition, bioinformatics prediction analysis showed that lncRNA Malat1 had a miRNA-15b-5p targeting site and that Ihh might be a downstream target of miR-15b-5p.
MiR-15b-5p regulated β-cell differentiation by targeting Ihh MiR-15b-5p has been reported as an important regulator in lung cancer, breast cancer, Alzheimer's disease and other pathological and physiological processes [24][25][26][27] but not in β-cell differentiation. Therefore, we wondered whether miR-15b-5p could regulate β-cell formation and whether Ihh was its downstream target. To validate the relationship between miR-15b-5p and Ihh, we used Western blotting to detect the protein level of Ihh. The results showed that Ihh was obviously negatively regulated by overexpressed miR-15b-5p (Fig. 6a). Next, in accordance with the putative binding sites between miR-15b-5p and Ihh, we constructed wild-type (wt) and mutant (mut) vectors of the 3′-UTR of Ihh (Fig. 6b). LRA results revealed that only the wt 3′-UTR of Ihh showed a signi cant response to miR-15b-5p, displaying a low level of translation after miR-15b-5p overexpression (Fig. 6c). Therefore, our results showed that miR-15b-5p may speci cally target the 3′-UTR of Ihh and regulate its translation.
As miR-15b-5p has not been previously reported to be involved in β-cell differentiation and function, we aimed to elucidated its role in β-cell differentiation in vitro. Using Ihh as a negative control (NC), we downregulated miR-15b-5p expression at the MPC stage using antagomir and cultured the cells until day 21. The qRT-PCR results demonstrated that although this was not true for Ihh, the absence of miR-15b-5p led to the downregulation of genes such as INS, solute carrier family 2 member 2 (Glut2), insulin gene enhancer protein (ISL), Pdx1, Nkx6-1, GCG and somatostatin (SST), indicating abnormal β-cell maturation (Fig. 6g). Moreover, regardless of the glucose concentration, miR-15b-5p de ciency impaired insulin production and secretion of β-cells (Fig. 6f). In addition, IF staining proved that aberrant low expression of miR-15b-5p weakened the protein levels of INS, Nkx6-1 and MafA in β-cells (Fig. 6d, e). Therefore, stable expression of miR-15b-5p was necessary for β-cell differentiation in vitro, and its absence was detrimental to cell function and maturation.
Malat1 regulated β-cell differentiation via the miR-15b-5p-Ihh axis Using bioinformatics to compare sequences, we found that Malat1 contained binding sites for miR-15b-5p (5023-5037), suggesting a possible interaction between these two molecules. Then, we constructed a luciferase vector of Malat1 (Luc-Malat1-wt) and a mutated form (Luc-Malat1-mut) (Fig. 7a). The results indicated that miR-15b-5p could effectively suppress the luciferase activity of Malat1 but not of mutated Malat1, indicating that the predicted binding sites of Malat1 is likely to bind to miR-15b-5p speci cally (Fig. 7b). However, more research will be needed to clarify whether Ihh is direct target gene of Malat1 during the differentiation.
In addition to structural interactions between lncRNA Malat1 and Ihh, Malat1 regulation of Ihh expression at the protein level needs further validation. As shown by the WB results, Malat1 de ciency led to downregulation of Ihh (Fig. 7c). Moreover, Malat1 overexpression upregulated the expression of Ihh (Fig.  7d). In addition, lncRNA Malat1 neutralized the inhibitory effect of miR-15b-5p on Ihh expression (Fig. 7e).
Rescue experiments indicated that downregulation of Ihh expression reversed the negative impact of Malat1 overexpression on β-cell differentiation (Fig. 7g), indicating that Ihh might be a functional target of Malat1. Taken together, these ndings con rmed that Malat1 could regulate Ihh expression by competitively binding miR-15b-5p, thereby in uencing iPSC-derived β-cell differentiation.

Aberrant overexpression of Malat1 in uenced the glycaemic control of β-cells in vivo
To further investigate the in uence of high expression of Malat1, we transplanted β-like cells transfected with lentivirus into the renal capsules of diabetic mice using alginate hydrogel (Fig. 8a, 8b) in accordance with a protocol reported in a previous study [23]. Next, we harvested approximately 1×10 6 β-cells from the Malat1 overexpression and NC groups and transplanted them into STZ-induced diabetic mice (n = 6). Twenty-four days post-transplantation, the fasting blood glucose of the mice with transplanted NC cells gradually returned to baseline (Fig. 8d). In contrast, differentiated cells with high expression of Malat1 could not radically reverse hyperglycaemia.
Furthermore, 40 days after transplantation, we harvested mouse kidneys with transplanted cells and stained them using IF. The results indicated that the cells overexpressing Malat1 showed worse maturation in vivo after transplantation than normal β-cells, with lower protein levels of C-peptide Mafa and Nkx6.1, three key markers of pancreatic endocrine cells (Fig. 8c). Therefore, the abovementioned experiments showed that the gradual downregulation of lncRNA Malat1 was important to β-cell differentiation, while aberrant overexpression affected cell formation in vitro and redifferentiation in vivo.
Emerging evidence shows that other molecules, mainly ncRNAs, also play vital regulatory roles in the development of pancreatic and islet cells [17,44,45]. NcRNAs such as miRNAs, small nucleolar RNAs (snoRNAs) and lncRNAs, which are not translated or do not encode a protein, were initially thought to lack biological functions [46,47]. Moreover, cell therapy is considered as a potential therapeutic alternative for diabetes [48][49][50]. Islet or pancreatic cell transplantation can restore glucose homeostasis by replenishing β-cells [51,52]. We adopted a three-step differentiation protocol imitating pancreatic development in vivo to identify differentiation-associated lncRNAs during iPSC-induced β-like cell differentiation in this study [2]. Our data indicated that among hundreds of differentially expressed lncRNAs we discovered using this protocol, lncRNA Malat1 might play an important role in β-like cell differentiation (Fig. 2). This molecule could reduce the Nkx6-1 + cell count and regulate the expression of some key TFs, including Pdx1, Nkx6-1, Ngn3, Gata4, Gata6, NeuroD1 and Pax6, in the early stage of pancreatic development (Fig.  3). PDX1-expressing epithelial progenitors are essential to the development of the pancreas and intestines, giving rise to pancreatic endocrine, exocrine and ductal cells [53,54]. NGN3 is expressed in all endocrine progenitors, starting with a series of TFs whose expression regulates endocrine cell differentiation, including Nkx2.2, NeuroD1, NKX6-1, PAX4, PAX6 and ISL [55][56][57][58].
Our studies revealed that Malat1 likely functioned as a negative regulator in the early period of β-cell differentiation (Fig. 3). Then, bioinformatics prediction indicated that lncRNA Malat1 acted as a molecular sponge for miR-15b-5p via its 3′-UTR binding site and regulated the expression of Ihh, a direct downstream target (Fig. 4). Ihh is localized and expressed in the gut endoderm, which is important for epithelial differentiation [59]. Studies have shown that deletion of Ihh leads to the formation of annular pancreas, suggesting that loss of Ihh from the gut endoderm around the pancreatic bud causes excessive growth of the pancreas [60]. As expected, miR-15b-5p knockdown reduced the quantity and quality of βcells after 21 days of culture. Additionally, these polyhormonal insulin-expressing cells in the experimental group could not release enough insulin to respond to high-concentration glucose (Fig. 6). Malat1 functioned as a ceRNA to block miR-15b-5p and in uenced the expression of Ihh at the protein level in vitro (Fig. 7). This nding indicated that this cytoplasmic lncRNA (Fig. 5) might have a regulatory relationship with the Hedgehog signalling pathway, which is widely reported to negatively regulate PDX1 during early pancreatic development. However, the concrete regulatory mechanism between these molecules needs more research. Our in vivo experiments showed that lncRNA Malat1 knockdown was indispensable for the differentiation of insulin-producing cells, although aberrant overexpression affected cell formation in vitro and differentiation in vivo.

Conclusion
We found that lncRNA Malat1, a broadly studied lncRNA, played a vital role in the differentiation of iPSCderived β-like cells, both in vivo and in vitro. Malat1 regulated the expression of a series of TFs indispensable for the formation and development of the pancreas at the early stage and might affect the expression of the Hedgehog signalling pathway during differentiation. These ndings indicated that Malat1 acted as a molecular sponge to regulate the expression and function of Ihh by directly and competitively binding to miR-15b-5p via its 3′-UTR, thereby mediating the differentiation of iPSC-derived β-like cells. The current study therefore provides new evidence that ncRNAs, including lncRNAs and miRNAs, can serve as epigenetic targets for improving mature β-like cell differentiation e ciency and cell functions.