Inhibition of lncRNA TCONS_00077866 Ameliorates the High Stearic Acid Diet-Induced Mouse Pancreatic β-Cell Inflammatory Response by Increasing miR-297b-5p to Downregulate SAA3 Expression

Long-term consumption of a high-fat diet increases the circulating concentrations of saturated fatty acids (SFAs), which induce pancreatic β-cell dysfunction. This precedes the onset of type 2 diabetes by some years, and high SFA concentrations may accelerate its development (1). The SFAs found in food mainly comprise palmitic acid (C16:0) and stearic acid (SA) (C18:0), but although the proportions of palmitic acid, whether in lard, palm oil, or soybean oil, are higher than those of SA (2,3), the serum ratios of palmitic and SA achieved after the consumption of these foods differ significantly, probably because of differences in their digestion and metabolism (4,5). We previously showed that the concentrations of SA were significantly higher, under both fasting and postprandial conditions, in the serum of patients with type 2 diabetes or hyperlipidemia and in mice fed a high-fat diet (6–9). Furthermore, SA has a more toxic effect on β-cells than other SFAs (9,10), which suggests that excessive exposure to high concentrations of circulating SA may have a large impact on β-cell dysfunction. However, the molecular mechanisms underlying this SA-induced lipotoxicity remain to be fully elucidated.

The mechanisms that have been proposed to be responsible for SFA-induced β-cell dysfunction include apoptosis, endoplasmic reticulum stress, abnormal stimulus-secretion coupling, aging, and senescence (11–15). In recent years, evidence has accumulated to indicate that activation of a pancreatic β-cell inflammatory response by high concentrations of circulating SFAs is closely linked to β-cell dysfunction (16,17), although the interaction between inflammation and β-cell dysfunction in type 2 diabetes remains under debate (18). In the SFA-stimulated inflammatory process, some proinflammatory factors are produced from β-cells, such as cytokines (e.g., tumor necrosis factor [TNF]-α, interleukin [IL] 6 and IL-1β) and chemokines (e.g., C-X-C motif chemokine ligand 10, IL8), which eventually lead to the failure of insulin secretion and diabetes (19,20). However, the specific contribution of SA to the activation of the β-cell inflammatory response and the potential mechanism of this action have received little attention to date.

Long noncoding RNAs (lncRNAs), which are >200 nucleotides in length, are emerging as new players in multiple cell processes (21,22). They regulate every aspect of gene expression, including chromatin structure modification, transcription, imprinting, splicing, and protein degradation, and also competitively bind miRNAs (23,24). Accumulating evidence suggests that the dysregulation of lncRNAs is implicated in many human metabolic diseases that are caused by chronic consumption of a high-fat diet, such as obesity (25), atherosclerosis (26), and type 2 diabetes (27). Although lncRNAs play a critical role in the development and function of pancreatic β-cells (28,29), no studies have examined the relationship between lncRNAs and SA-induced β-cell inflammation, which prompted us to characterize the involvement of lncRNAs in this pathological process.

To this end, we profiled lncRNA expression in SA-treated mouse pancreatic β-TC6 cells by using high-throughput sequencing and identified the most upregulated lncRNA, TCONS_00077866 (lnc866). We then determined the effect of lnc866 on the SA-induced β-cell inflammatory response and impaired insulin secretion by using a loss-of-function approach in mouse and cell models. Using an lncRNA-miRNA-mRNA coexpression network, we identified the miR-297b-5p/serum amyloid A3 (SAA3) axis as a potential downstream target of lnc866 during the SA-stimulated β-cell inflammatory response. This may provide potential targets for the prevention and treatment of type 2 diabetes.

Stock solutions of SA and palmitic acid (Sigma-Aldrich, St. Louis, MO) supplemented with BSA (3 mmol/L fatty acid: 1.5 mmol/L BSA; molar ratio, 2:1) were prepared, as described previously (30).

Mouse pancreatic β-TC6 cells were obtained from the Shanghai Academy of Chinese Sciences Cell Library. Mouse islets were isolated after ductal injection of collagenase P (cat. no. 11213873001, Roche Molecular Biochemicals, Mannheim, Germany) to digest the pancreas and purified with different concentrations of Ficoll 400 (cat. no. 17-0300-10, Pharmacia). β-TC6 cells and islets were cultured as described previously (30) and treated with 400 μmol/L SA or palmitic acid in the presence of 1.3% BSA for 24 h.

Genewiz (Suzhou, China) performed the lncRNA high-throughput sequencing with depth ranging from 40× to 70×, by using a 2 × 150 paired-end configuration, and conducted image analysis and base calling using the HiSeq Control Software + Off-Line Basecaller (OLB) + GAPipeline-1.6 (Illumina) on the HiSeq instrument. They processed nine samples from β-TC6 cells (three control, three SA-treated, and three palmitic acid-treated samples). Total RNA of each sample was extracted using TRIzol Reagent (Invitrogen) and an RNeasy Mini Kit (QIAGEN) with rRNA depletion. Differential expression analysis used the DESeq Bioconductor package in R software by first transforming the raw count data to log₂ counts per million reads using the “voom” function (P < 0.05). Transcripts were merged with cuffmerge and filtered by known nonlong intervening noncoding RNA (lincRNA) annotation, nonlincRNA characters, open reading frames (ORFs), and protein-coding potential methods. Known nonlincRNAs include known protein-coding RNAs, miRNAs, transfer RNAs, small nucleolar RNAs, rRNAs, and pseudogenes. Three characteristics were considered: nonlincRNA transcripts must contain more than one exon, their length must be >200 base pairs, and the coverage of each transcript must be more than three. ORFs were predicted by HMMER 3.1b1 software. Coding Potential Calculator was used to predict protein-coding potential. Only lncRNA sequences that do not have the capability to produce small peptides, even from noncanonical translation, were obtained. The lncRNA-miRNA-mRNA coexpression network was constructed using Cytoscape with Pearson correlation coefficients ≥0.850 for lncRNA-mRNA and ≤0.850 for lncRNA-miRNA and miRNA-mRNA (P < 0.05), according to the miRNA and mRNA expression profile published in our previous study (10) and the lncRNA profile in the current study.

Adeno-associated virus serotype 9 (AAV9) expressing green fluorescent protein (GFP) and a short RNA targeting lncRNA-866 (sh866-V) and its negative control (shNC-V) were constructed by Hanbio Biotechnology Co., Ltd. (Shanghai, China). The AAV9 packaging system includes three plasmids, pAAV-RC, pHelper, and shuttle plasmid (carrying sh866). The details of sh866-V and shNC-V are displayed in Supplementary Table 1 and Supplementary Fig. 1.

Male C57BL/6J mice (6 weeks old) were purchased from the Beijing Vital River Laboratory Animal Technology Company (Beijing, China). C57BL/6J mice were randomized to be administered shNC-V or sh866-V via an intraductal pancreatic route (n = 70 per group; body mass 22–25 g). Six weeks after AAV injection, mice that had been administered shNC-V or sh866-V were randomly allocated to normal diet and high-SA diet (HSD) groups (n = 35 per group). The normal diet (cat. no. 1025) and HSD (cat. no. H10060) were obtained from Beijing HFK Bioscience Co., Ltd. (Beijing, China) (Supplementary Table 2). After 10 weeks of consumption of the normal diet or HSD, tissue and blood samples were collected for further analyses. All of the animal procedures were approved by the Harbin Medical University Institutional Animal Care and Use Committee, and the animals were maintained according to the guidelines of the Harbin Medical University Animal Experimental Center.

Intravenous glucose tolerance testing was performed in overnight-fasted mice. Glucose (0.75 g/kg) was administered to mice via the tail vein, and blood samples were collected 0, 1, 5, 10, 20, 30, and 60 min later for the measurement of insulin and glucose concentrations.

Fasting serum glucose, total cholesterol (TC), triacylglycerol (TG), HDL-cholesterol (HDL-C) and LDL-cholesterol (LDL-C) concentrations were measured using an automatic analyzer (Hitachi 7100; Hitachi, Tokyo, Japan) and kits purchased from Biosino Biotechnology Co. (Beijing, China). Serum insulin concentration was measured using a mouse/rat insulin ELISA kit (cat. no. EZRMI-13K, Millipore, Burlington, MA).

Fasting serum nonesterified fatty acids were transformed to fatty acid methylesters, as described previously (6). Heptadecanoic acid (C17:0) was used as an internal standard.

A locked nucleic acid-modified oligonucleotide probe targeting lnc866 and a fluorescence in situ hybridization (FISH) kit were purchased from Ribobio Co., Ltd., Guangzhou, China. According to the manufacturer’s instructions, β-TC6 cells and pancreatic tissue sections were fixed in 4% formaldehyde at room temperature for 10 min and washed three times. After permeabilization, they were blocked in Blocking:Prehybridization solution (1:99) at 37°C for 30 min. Next, they were incubated with a hybridization mixture containing lnc866 Probe Mix or U6/18S at 37°C overnight. The preparations were subsequently washed and incubated with DAPI staining solution at room temperature for 10 min. The slides were observed with a laser confocal microscope.

Pancreatic tissues were fixed and then frozen in embedding reagent (Tissue-Tek OCT Compound, 4583; Sakura, Torrance, CA), sectioned, and immunostained using anti-insulin (BM1621, Boster) or anti-glucagon (3014S, Cell Signaling Technology) antibodies, as described previously (9).

Smart Silencer for lnc866 (ss-866), AL137118.20, and their negative controls were transfected using a RiboFect CP Transfection Kit (C10511-05). miR-297b-5p mimics, antimiR-297b-5p oligonucleotides (AMO-297b-5p), siRNA-Saa3, siRNA-Rela, siRNA-Il6, and siRNA-Tnfα, and their negative controls were transfected using Lipofectamine 2000 (Invitrogen), according to the manufacturer’s instructions. The target sequences of the lncRNA ss-866 and human AL137118.20 and all siRNAs are shown in Supplementary Tables 3, 4, and 5, respectively. ss-866 and AL137118.20 comprised three siRNAs and three antisense oligonucleotides, respectively. All of these materials were purchased from RiboBio Co., Ltd.

Human embryonic kidney (HEK-293) cells (ATCC, Manassas, VA) were cotransfected with 200 ng pmiR-RB-REPORT dual-luciferase reporter vectors (RiboBio Co., Ltd.) carrying the 3′- untranslated region (UTR) of mouse Saa3 or lnc866 containing wild-type or mutated target sites for miR-297b-5p and miR-297b-5p mimics. We collected the cell lysate 24 h after transfection and measured luciferase activity using a dual luciferase reporter assay kit (Promega) and a GloMax 20/20 Luminometer (Promega). We also constructed luciferase reporter plasmids carrying a human homolog of lnc866 (AL137118.20) or the 3′-UTR of SAA1 encompassing wild-type or mutated binding sites for hsa-miR-297 and investigated the specific interaction between human AL137118.20-miR-297-SAA1.

RNA was extracted from β-TC6 cells, mouse tissues, and human islets using TRIzol reagent (Invitrogen), according to the manufacturer’s protocol. miRNAs were isolated using a mirVana miRNA Isolation Kit (Ambion, Austin, TX). Real-time quantitative (q) PCRs were performed as described previously (12). All primers were synthesized by Sangon Biotech Co., Ltd. (Shanghai, China), and their sequences are listed in Supplementary Table 6. The lncRNA and SAA1 mRNA PCR products were subjected to Sanger sequencing (Sangon Biotech Co., Ltd., Shanghai, China).

Glucose-stimulated insulin secretion (GSIS) was measured for β-TC6 cells, mouse, and human islets (control group), as described previously (9,30), by using a mouse/rat insulin ELISA kit for mouse and a human insulin ELISA kit (cat. no. EZHI-14K, Millipore, Burlington, MA) for human. Insulin was normalized by total protein levels.

Protein samples (∼80 µg) were separated by SDS-PAGE and transferred to polyvinylidene fluoride membranes. The following primary antibodies were used: anti-SAA3 (dilution of 1:50; Abcam, ab231680,); anti–nuclear factor (NF)-κB p65 (dilution of 1:1,000; Cell Signaling Technology, #8242); anti-IL6 (dilution of 1:1,000; Cell Signaling Technology, #12912); anti–TNF-α (dilution of 1:1,000; WanleiBio, WL01581), and anti–β-actin (dilution of 1:800; Santa Cruz Biotechnology, sc-130657). Secondary antibodies used were anti-rabbit alkaline phosphatase-conjugated antibody (dilution of 1:7,500; Promega, S373B) and horse radish peroxidase-conjugated rabbit anti-rat antibody (dilution of 1:5,000; BBI Life Sciences, D110273). Protein detection was performed using stabilized substrate for alkaline phosphatase (Promega, S3841) and the ECL Chemiluminescent Substrate Kit (Biosharp, BL520A), and imaging was performed using the FluorChem R system (ProteinSimple, San Jose, CA).

Human islets preparations were obtained from 13 normolipidemic and 9 hyperlipidemic patients who had undergone partial pancreatectomy for different digestive system diseases (pancreatic tumors, biliary obstruction, and common bile duct or duodenal tumors) at the Department of Hepatobiliary and Pancreatic Surgery, Tumor Hospital of Harbin Medical University. All patients consented to the use of their tissues for the current study, which was approved by the Harbin Medical University Tumor Hospital Ethics Committee (No. 2018-104-R). All study protocols were performed according to the guidelines of the Declaration of Helsinki.

Islets were isolated from healthy and soft pancreatic tissues following a previously described procedure (31,32). A digestive enzyme solution, consisting of Collagenase P 1.5–2.0 mg/mL, dissolved in balanced salt solution (NaCl, 4; KCl, 0.4; Na₂HPO₄·12H₂O, 0.1343; KH₂PO₄, 0.06; HEPES, 2.383; and CaCl₂, 0.167 [all in g/L]; pH 7.8) was injected into pancreatic tissues to distend them homogeneously. Next, the tissue was cut into small pieces on ice. After digestion at 37°C, the tissue was shaken until it had the appearance of fine sand and then put into Hanks’ solution supplemented with FBS (4°C) to stop digestion. After filtration, washing, and centrifugation, the pellet was suspended with Ficoll 400 at a density of 1.100 g/mL, and then covered with sequential layers of Ficoll 400 at densities of 1.080, 1.060, and 1.040 g/mL, followed by centrifugation at 2,200g for 20 min. We harvested the islets from the 1.040–1.060 g/mL and 1.060–1.080 interphases for further processing.

All data were analyzed using SPSS 21.0 software (IBM, Armonk, NY) and are reported as the mean ± SD. Two-tailed Student t tests were used to compare two groups, and one-way ANOVA, followed by the Student–Newman-Keuls test, was used to compare multiple groups. P < 0.05 was considered to represent statistical significance.

The data set generated and analyzed in this study is available from the corresponding authors upon reasonable request. The original omics data have been deposited with Gene Expression Omnibus (GSE168825).

In total, 4,293 known and 11,296 novel, distinct lncRNAs were detected using high-throughput sequencing technology. The differential lncRNA expression profiles of control, SA-treated, and palmitic acid-treated β-TC6 cells were distinguishable in a heat map generated by hierarchical clustering (Fig. 1A). We identified 189 lncRNAs that were differentially expressed in SA-treated compared with control cells (101 upregulated and 88 downregulated). For the palmitic acid-treated group, expression was higher for 53 and lower for 28 lncRNAs compared with control cells (Fig. 1B and Supplementary Tables 7–9). The Venn diagram presented in Fig. 1C shows that five lncRNAs were specifically differentially expressed in SA-treated β-TC6 cells compared with the control and palmitic acid-treated groups. These lncRNAs were TCONS_00077866 (lnc866), TCONS_00089573 (lnc573), TCONS_00230830 (lnc830), TCONS_00230836 (lnc836), and TCONS_00252600 (lnc600), of which the expression of lnc866, a previously unknown intergenic lncRNA, was the most upregulated (Table 1). Real-time qPCR showed that the expression of lnc866 and lnc836 was upregulated, whereas lnc600 expression was downregulated in the SA-treated β-TC6 cells and HSD mouse islets, which was consistent with the RNA sequencing results. Of these, lnc866 also exhibited the largest increases in expression (a 2.23 log₂-fold difference in SA-treated β-TC6 cells and a 1.69 log₂-fold difference in HSD mouse islets) (Fig. 1D). These original omics data have been deposited with Gene Expression Omnibus (GSE168825).

A mouse model of high circulating SA concentration was successfully established with a HSD, as evidenced by a significantly higher concentration of circulating SA in HSD-fed versus control mice (Supplementary Table 10).

lncRNAs show high tissue and species specificity; therefore, we first measured lnc866 expression in various tissues of mice that are involved in metabolic disorders induced by a high-fat diet (islets, liver, skeletal muscle, brown adipose tissue, and peritesticular and perirenal adipose depots) and compared the sequences of the qPCR products and full-length lnc866 (Supplementary Table 11) using DNA sequencing. Matching sequences (similarity >96%) were identified in islets, perirenal adipose tissue, and skeletal muscle, but not in liver, peritesticular adipose tissue, or brown adipose tissue (Supplementary Fig. 2). The expression of lnc866 was highest in islets and lowest in perirenal adipose tissue (Fig. 1E). Next, we used National Center for Biotechnology Information-Basic Local Alignment Search Tool (BLAST) to search for homologous human sequences of lnc866. AL137118.20 (143638–144449, 146533–146662, and 145118–145375) exhibited the best sequence-match with lnc866 (total score, 263; identity, 65.32%). The expression of AL137118.20 in human islets from patients with hyperlipidemia was also significantly increased (Fig. 1F). The sequence alignment of AL137118.20 and lnc866 is presented in Supplementary Fig. 3, and the characteristics of the hyperlipidemia patients are presented in Supplementary Table 12.

Successful delivery of sh866-V or shNC-V into mouse pancreas was confirmed by a significant fluorescent signal from GFP expressed from the viral vectors (Supplementary Fig. 4). The injection of sh866-V into HSD mice significantly ameliorated their high fasting glucose and insulin concentrations, but the HSD-induced increase in body mass, and the serum concentrations of TG, TC, HDL-C, and LDL-C, were not significantly altered by the knockdown of lnc866 in mouse islets (Supplementary Table 133). Real-time qPCR (Fig. 2A) and FISH (Fig. 2B) confirmed that sh866-V significantly reduced lnc866 expression in the islets and pancreas of mice that were or were not consuming the HSD. FISH of mouse pancreas also showed that lnc866 was principally present in the cytoplasm (Fig. 2B). Furthermore, the HSD-induced impairment in glucose tolerance was ameliorated by sh866-V injection (Fig. 2C), and most strikingly, the first-phase insulin secretion was increased and the second-phase insulin release was significantly reduced by sh866-V injection (Fig. 2D). This indicates that the impairment in GSIS was substantially ameliorated by sh866-V. Moreover, the number of α-cells in the sh866-V group fed the HSD (sh866-V + HSD) was significantly lower than in the HSD group, and the normal cell distribution pattern was significantly restored by the sh866-V administration (Fig. 2E). Additionally, as shown in Fig. 2F, the injection of sh866-V ameliorated the HSD-induced increases in NF-κB p65, IL6, and TNF-α expression. In vitro, transfection with ss-866 significantly reduced lnc866 expression in β-TC6 cells (Fig. 3A and B), which rescued the SA-induced impairment in GSIS (Fig. 3C). However, no significant effect was observed on GSIS when ss-866 was transfected into control cells (Fig. 3C). Furthermore, the SA-induced increases in expression of NF-κB p65, IL6, and TNF-α were also ameliorated (Fig. 3D). Additionally, in human islets, inhibition of AL137118.20 (homologous human sequences of lnc866) also remarkably restored SA-impaired GSIS (Supplementary Fig. 5). The inhibitory efficiency of Smart Silencer for AL137118.20 is shown in Supplementary Fig. 6.

We performed a bioinformatic analysis of lncRNA, miRNA, and mRNA expression profiles and constructed a coexpression network for lnc866, which comprised lnc866, miR-297b-5p, and 31 mRNA nodes (Supplementary Fig. 7 and Supplementary Tables 14–17). To confirm an interaction between lnc866 and miR-297b-5p, we first performed sequence alignment analysis and found three binding sites of miR-297b-5p in lnc866 (Fig. 4A). Then, we measured the expression of miR-297b-5p after the inhibition of lnc866 expression in vivo and in vitro. In mouse islets, the administration of sh866-V increased miR-297b-5p expression in the presence or absence of HSD feeding (Fig. 4B). Consistent with this, miR-297b-5p expression in β-TC6 cells was also increased by the transfection of ss-866, in the presence or absence of SA (Fig. 4C and Supplementary Fig. 8). Moreover, cotransfection of β-TC6 cells with ss-866 and AMO-297b-5p blocked the protective effect of lnc866 against the SA-induced reduction in miR-297b-5p expression (Fig. 4C), the impairment of GSIS (Fig. 4D), and the inflammatory response (Fig. 4E) compared with the administration of SA and ss-866. Furthermore, the luciferase activity of the wild-type lnc866 luciferase reporter vector was significantly reduced by the miR-297b-5p mimic, with the binding site 1 in lnc866 showing the most significant change, compared with the other two sites. However, this effect was not observed with the luciferase reporter vector carrying a mutant miR-297b-5p binding site (Fig. 4F). Additionally, we also predicted one potential binding motif for hsa-miR-297 in the homologous human sequence of lnc866 (AL137118.20:143638–144449) (Fig. 4G). Transfection of hsa-miR-297 mimic significantly suppressed luciferase activity of a vector carrying a fragment of wild-type AL137118.20 (Fig. 4H). Meanwhile, knockdown of AL137118.20 prevented the SA decrease in the level of hsamiR-297 in human islets (Supplementary Fig. 9).

miR-297b-5p expression was significantly lower in SA-treated β-TC6 cells, but this effect was substantially ameliorated by the application of miR-297b-5p mimics (Fig. 5A). Overexpression of miR-297b-5p also ameliorated the SA-induced reduction in GSIS (Fig. 5B) and the inflammatory response (Fig. 5C). However, the transfection of miR-297b-5p mimics alone did not have a significant protective effect on β-cell function. On the contrary, the transfection of AMO-297b-5p alone into β-TC6 cells significantly reduced miR-297b-5p expression (Supplementary Fig. 10A) but had slightly negative effect on β-cell function in response to SA (an impairment in insulin secretion) (Supplementary Fig. 10B) and increased expression of the proinflammatory genes NF-κB p65, IL6, and TNF-α (Supplementary Fig. 10C). Similarly, overexpression of miR-297b-5p reversed SA-induced impaired GSIS and upregulation of NF-κB p65, IL6, and TNF-α protein levels in mouse islets (Supplementary Fig. 11A and B).

On the basis of the lncRNA-miRNA-mRNA coexpression network and the prediction of binding sites using RNAhybrid, TargetScan, miRanda, and miRTarBase, we hypothesized that SAA3 may be the downstream target of miR-297b-5p (Fig. 5D) that mediates its effects on SA-induced β-cell inflammation. As shown in Fig. 5E, levels of SAA3 mRNA and protein were high in SA-treated β-TC6 cells and were reduced by the overexpression of miR-297b-5p. A consistent pattern of Saa3 mRNA expression was observed in mouse islets exposed to SA (Supplementary Fig. 11B). Similarly, transfection of AMO-297b-5p alone increased SAA3 mRNA and protein levels (Fig. 5F). The ability of miR-297b-5p to bind Saa3 was evaluated using a luciferase activity assay. miR-297b-5p reduced the luciferase activity originating from a wild-type vector that encoded the 3′-UTR of Saa3 (Fig. 5G). To confirm that the miR-297b-5p/SAA3 axis is downstream of lnc866, we measured SAA3 expression after the inhibition of lnc866. Figure 5H shows that SAA3 expression at the mRNA and protein levels was significantly higher in the islets of HSD-fed mice and that this expression was reduced by the knockdown of lnc866. Similar patterns of SAA3 expression of were measured in SA-treated β-TC6 cells (Fig. 5I). Additionally, in islets from hyperlipidemic patients, hsa-miR-297 expression was also significantly downregulated, and the level of SAA1 (human homolog of mouse Saa3) expression was increased (Supplementary Fig. 12), consistent with the findings in mice. Meanwhile, sequence alignment (Fig. 5J) and luciferase activity assay results (Fig. 5K) indicated that the 3′-UTR of SAA1 contained binding sites for hsamiR-297, indicating that SAA1 is also a direct target of hsa-miR-297 in human.

In β-TC6 cells, the transfection of siRNA-Saa3 significantly ameliorated the SA-induced increase in SAA3 mRNA and protein expression (Fig. 6A) and ameliorated the impairment in GSIS (Fig. 6B). Furthermore, the SA-induced increase in NF-κB p65, IL6, and TNF-α expression was also largely prevented by pretreatment with siRNA-Saa3 (Fig. 6C). However, in the absence of SA, the transfection of siRNA-Saa3 did not affect β-cell function. In mouse islets, silencing Saa3 produced a protective effect on β-cell function and an inhibitory action on NF-κB p65, IL6, and TNF-α expression in the presence of SA (Supplementary Fig. 13A and B).

Transfections of siRNA-Rela, siRNA-Il6, and siRNA-Tnfα clearly ameliorated the SA-induced increase in NF-κB p65, IL6, and TNF-α expression, respectively (Fig. 7A–C). Meanwhile, knockdown of NF-κB p65 and IL6 significantly reduced the SA-induced GSIS impairment in mouse islets (Fig. 7D and E). However, these reversal effects were relatively slight after Tnfα suppression (Fig. 7F).

The long-term consumption of a diet rich in SFAs is a major risk factor for β-cell dysfunction, which precedes the development of type 2 diabetes. In the current study, we demonstrated that lnc866 is robustly upregulated in the islets of mice fed an HSD and in SA-treated β-TC6 cells, resulting in a significant impairment in pancreatic β-cell function. Furthermore, consistent with a causative role, the knockdown of lnc866, both in vivo and in vitro, restored the SA-induced impairment in β-cell insulin secretion by inhibiting SA-stimulated upregulation of NF-κB p65, IL6, and TNF-α expressions, via relieving the repression of miR-297b-5p and thereby suppressing SAA3 expression (Fig. 8). These findings provide a fuller understanding of the role of lncRNAs in SA-induced β-cell injury.

Although accumulating evidence indicates that noncoding RNAs are critical mediators of SFA-induced lipotoxicity in β-cells, most studies to date have mainly focused on miRNAs (10,33). To evaluate the importance of lncRNAs in SA-induced β-cell dysfunction, we first compared lncRNA expression profiles in SA- and palmitic acid-treated β-TC6 cells and control cells, using RNA sequencing analysis, and found that lnc866, a novel intergenic lncRNA, may be specifically enriched in islets. To further explore the role of lnc866 in SA-induced β-cell dysfunction, we injected an AAV9 vector carrying shRNA lnc866 intraductally into the pancreas of mice to knock down endogenous islet lnc866 expression. This significantly ameliorated the SA-induced impairment in β-cell insulin secretion and the upregulation of inflammatory factors. Similar results were obtained in a cell model following the administration of ss-866, which was a highly effective inhibitor of lnc866 expression, both in the cytoplasm and nucleus. These data indicate that the mechanism of SA-induced β-cell dysfunction and inflammation involves lnc866.

Owing to their ability to form various secondary and tertiary structures, lncRNAs can interact with DNA, RNA, and protein, which enables them to regulate gene expression at multiple levels, such as the transcriptional and posttranscriptional levels (34). A number of lncRNAs have been documented to regulate target genes by absorbing miRNAs through a “sponge-like” activity (35). Using an lncRNA-miRNA-mRNA coexpression network, we predicted that the miR-297b-5p/SAA3 axis is a target signaling pathway of lnc866 in SA-induced β-cell inflammation. To test this hypothesis, we first focused on miR-297b-5p. The coexpression network data showed that miR-297b-5p was the most highly negatively correlated with lnc866 (Pearson correlation coefficient = −0.964, P = 0.002). Furthermore, sequence alignment analysis and luciferase reporter activity assays showed that there may be a direct interaction between lnc866 and miR-297b-5p. Meanwhile, in both the islets of HSD-fed mice and SA-treated β-TC6 cells, the silencing of lnc866 significantly ameliorated the SA-induced reduction in miR-297b-5p expression. These data indicate that the effect of lnc866 in SA-induced β-cell dysfunction and inflammation may be attributed to a change in miR-297b-5p expression.

Most previous studies of miR-297 are related to its role as a tumor suppressor (36), but recently, miR-297 was reported to play a protective role in the inflammatory response to sepsis and lung injury (37,38). However, the role of miR-297 in pancreatic β-cell inflammation has not yet been explored. Here, we demonstrated that miR-297b-5p expression in β-TC6 cells and mouse islets was markedly reduced by SA treatment. Furthermore, overexpression of miR-297b-5p significantly reduced the SA-stimulated expression of inflammatory factors and ameliorated the reduction in insulin secretion. These results, together with those of our previous study (30), indicate that miR-297b-5p influences the effects of SA on β-cell function through at least two different processes: apoptosis and the inflammatory response.

Another key finding of the current study, based on bioinformatic analysis, is that SAA3 may be the miR-297b-5p target that mediates its effects on SA-induced β-cell inflammation. This is the first report of a role for SAA3 in islets and β-cells. SAA3, an important member of the SAA family, is closely associated with chronic inflammatory conditions, such as obesity (39), atherosclerosis (40), and hyperlipidemia (41). Its expression increases in a variety of tissues that are subject to injury, infection, or chronic inflammation (42). Here, we found that in the islets of both HSD-fed mice and SA-treated β-TC6 cells, SAA3 expression was high at both the mRNA and protein levels. Furthermore, the knockdown of SAA3 significantly ameliorated the proinflammatory effect of SA on β-cells and the SA-induced impairment of GSIS. SAA3 induces the production of many cytokines that promote the inflammatory response (43,44), which explains why inhibition of SAA3 ameliorates SA-induced β-cell inflammation. To determine the effect of miR-297b-5p on SAA3 expression, we overexpressed or inhibited miR-297b-5p expression in β-TC6 cells in the absence or presence of SA, respectively, and found that SAA3 mRNA and protein expression was significantly negatively related to miR-297b-5p expression. Meanwhile, a luciferase reporter assay confirmed that SAA3 may be a direct target of miR-297b-5p. Moreover, silencing lnc866 expression also significantly suppressed the SA-induced increase in SAA3 expression in both mouse islets and β-TC6 cells. Thus, inhibition of lnc866 ameliorates the SA-induced mouse pancreatic β-cell inflammatory response, probably via the miR-297b-5p/SAA3 axis.

Although the present work provides evidence that lnc866/miR-297b-5p/SAA3 signaling is involved in SA-induced β-cell dysfunction and inflammatory stress, the causal relationship between these two processes remains controversial to date (16,45,46). To verify whether the proinflammatory factors induced by SA contribute to β-cell dysfunction, we used NF-κB p65, IL6, and TNF-α siRNAs in SA-treated mice islets and found that artificial inhibition of NF-κB p65 and IL6 caused a significant recovery of SA-impaired insulin secretion. This means that SA-induced production of proinflammatory factors can trigger β-cell destruction and accelerate the development of type 2 diabetes.

Whether the noncoding RNAs regulatory mechanisms also occur in humans is a key point for prevention and treatment of human diseases. We found that human AL137118.20 exhibits the best sequence match with lnc866 and that its level is significantly increased in human islets from patients with hyperlipidemia. Silencing AL137118.20 in human islets also restored SA-impaired GSIS. After sequence alignment analysis and luciferase reporter assays, hsa-miR-297 was identified as a direct target for AL137118.20. Meanwhile, hsa-miR-297 was overexpressed when AL137118.20 was inhibited in the absence or presence of SA.

Next, we identified whether the direct interaction between miR-297b-5p and SAA3 also operates in humans. Although human SAA3 has been characterized as a pseudogene, human SAA1 contains 68–71% amino acid homology with SAA3 (47) (Supplementary Fig. 14). SAA1 was also significantly overexpressed, and hsa-miR-297 was downregulated in islets from hyperlipidemic patients, consistent with the results in mice. Binding site prediction and luciferase activity assays indicated that SAA1 is a direct downstream target of hsa-miR-297 in humans. This suggests that the mouse lnc866-miR-297b-5p-SAA3 mechanism can possibly be extrapolated to human islets. Rigorous further studies are required to clarify this issue. Additionally, assessment of GSIS in islets of mice fed the HSD or the use of AAV in vitro would be more informative for confirming the islet response to glucose. Finally, although we have shown that silencing lnc866 significantly reduces SAA3 expression, probably because of a relief of its suppression of miR-297b-5p, a search for other potential direct or indirect mediators of the positive effect of lnc866 on SAA3 expression should be conducted.

Taken together, our findings from in vivo and in vitro experiments indicate that lnc866 knockdown is able to rescue the SA-induced inflammatory response in β-cells and that the miR-297b-5p/SAA3 axis is probably the downstream target of lnc866. Thus, we have provided novel insight into the molecular mechanisms by which lncRNAs influence SA-induced β-cell dysfunction. Inhibition of the inflammatory response, by the inhibition of lnc866, may represent a novel strategy for the prevention or treatment of type 2 diabetes.

Acknowledgments. The authors thank Jeremy Allen, PhD, from Liwen Bianji, Edanz Group China (www.liwenbianji.cn/ac), for editing the English text of a draft of this manuscript.

Funding. This work was supported by the Excellent Youth Foundation of Heilongjiang Province of China (YQ2020H033 to H.L.), the National Key R&D Program of China (2017YFC1307401 to C.S.), the University Nursing Program for Young Scholars with Creative Talents in Heilongjiang Province (UNPYSCT-201705 to H.L.), and the National Natural Science Foundation of China (81773424 to C.S.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.L. and R.G. wrote the manuscript. H.L., R.G., and X.C. reviewed and edited the manuscript. H.L. and C.S. conceived and designed the experiments. R.G. and Yu.Z. analyzed the data. R.G., Yu.Z., S.S., Q.Z., Y.Y., Yo.Z., S.L., and D.S. performed the experiments. H.Sh. and H.Su. calculated the Pearson correlation coefficients and searched for the human homologous sequence of mouse lnc866 and SAA3. H.L. and C.S. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.