LINC00472 suppressed by ZEB1 regulates the miR‐23a‐3p/FOXO3/BID axis to inhibit the progression of pancreatic cancer

Abstract The tumour‐suppressive role of LINC00472 has been extensively reported in various human cancers such as lung, colon and ovarian cancers, yet its function in pancreatic cancer remains unidentified. Here, the current research aimed to explore the role and regulatory axis mediated by LINC00472 in the progression of pancreatic cancer. RT‐qPCR was adopted to determine LINC00472 expression in the harvested pancreatic cancer tissues and adjacent normal tissues. Loss‐of‐function and gain‐of‐function experiments were performed to examine the effects of LINC00472 on proliferation and apoptosis in vitro and tumorigenesis in vivo. Immunoblotting was performed to detect the expression of several proliferation and apoptosis‐related proteins. Bioinformatic analysis, dual‐luciferase reporter assay and RNA pull‐down were conducted to profile the relationships between LINC00472 and miR‐23a‐3p, between miR‐23a‐3p and FOXO3 and between FOXO3 and BID. The LINC00472 expression was down‐regulated by ZEB1 in the pancreatic cancer cells and tissues. LINC00472 could competitively bind to miR‐23a‐3p to enhance the expression of FOXO3, which consequently could promote the BID expression, thereby suppressing proliferation and promoting the apoptosis of pancreatic cancer cells. Meanwhile, the inhibitory role of LINC00472 in tumorigenesis was validated in vivo, and the LINC00472‐mediated miR‐23a‐3p/FOXO3/BID axis was also demonstrated in the nude mouse tumour formation model. The study substantiated the antitumour activity of LINC00472 in pancreatic cancer and proposed a regulatory axis in which LINC00472 competitively binds to miR‐23a‐3p to enhance the FOXO3 expression and promote BID expression. Consequently, these findings provide theoretical basis for developing potential targets for the treatment of pancreatic cancer.


| INTRODUC TI ON
Pancreatic cancer is regarded as the 17th leading cause of cancerrelated death worldwide as indicated by statistics. Currently, the prognostic outcome for pancreatic cancer patients is certainly poor with an insignificant 5-year survival rate of around 10%. [1][2][3] The high mortality of pancreatic cancer is associated with late diagnosis, as in the early stages, pancreatic cancer tends to be asymptomatic and only becomes symptomatic upon progression into advanced or metastatic stages. 4 As for the risk factors for pancreatic cancer, intrinsic parameters such as elder age, family history and genetic susceptibility are the chief non-modifiable risk factors, while parameters such as smoking, alcohol consumption and obesity have been extensively investigated and verified as the most significant modifiable risk factors. 5 Over decades, a combination approach with surgical resection and chemotherapy or radiation therapy has persisted as the gold standard treatment protocol for pancreatic cancer patients as only a limited number are localized, surgically resectable tumours. 6,7 For future advancements in this field, investigating the development pathways of pancreatic cancer is of great significance for identifying novel diagnostic biomarkers and therapeutic targets to improve the prognosis of patients with pancreatic cancer.
Long intergenic non-protein coding RNAs (lincRNAs) are a set of autonomously transcribed non-coding RNAs with transcript lengths exceeding 200 nucleotides, and this group constitutes over 50% of all long non-coding RNA (lncRNA) transcripts in humans. 8 LincRNA 00472 (LINC00472) has demonstrated functionality as a competing endogenous RNA (ceRNA) to bind to several microRNAs (miRNAs) to suppress the progression of multiple types of cancers, including lung cancer 9 and colorectal cancer. 10 Additionally, miRNA has demonstrated to have vital significance in the development of human cancers by post-transcriptional regulation of tumour suppressors or promoters, 11 and miR-23a-3p could exercise a tumour-promoting effect on pancreatic cancer by targeting the transforming growth factorβ receptor type II (TGFBR2). 12 Moreover, forkhead box O3 (FOXO3) was a tumour suppressor targeted by miRNA in human pancreatic cancer progression, 13 while FOXO3a has been previously demonstrated to regulate BH3-interacting domain death agonist (BID), which is a pro-apoptotic protein belonging to the B-cell lymphoma-2 (Bcl-2) protein family. 14 An existing study elicited BID as an essential regulator of cell apoptosis in pancreatic cancer. 15 As we predicted that FOXO3 was a target of miR-23a-3p and LINC00472 could bind to miR-23a-3p based on the bioinformatic analysis, we speculated the functionality of LINC00472 as a ceRNA of miR-23a-3p to regulate FOXO3 and subsequently influence BID-dependent cancer cell apoptosis. In addition, Zinc finger E-box binding homeobox 1 (ZEB1) has been demonstrated to interact with lncRNAs to regulate their transcription and affect pancreatic cancer migration, invasion and epithelial-mesenchymal transition, 16 suggesting that ZEB1 may be involved in the cell apoptotic pathway initiated by LINC00472. In this study, we constructed both cellular and animal models to systematically investigate the signalling axis ZEB1/ LINC00472/miR-23a-3p/FOXO3 in pancreatic cancer.

| Microarray-based gene expression profiling
Pancreatic cancer-related gene expression data sets GSE46234 and GSE24279 were retrieved from the Gene Expression Omnibus and normalization on each data set. 17 Non-specific filtering of the expression profile data was conducted by the linear models, while the screening of the differentially expressed lncRNAs and miRNA was conducted by a combination of the empirical Bayes methods in the Limma installation package and the traditional t test. 18 The starBase database (http://starb ase.sysu.edu.cn/) was adopted to predict the miRNA bound by lncRNA and the target gene of miRNA. 19 (Table S1). An investigation was conducted for the expression of LINC00472 in the pancreatic tissues and adjacent normal tissues by reverse transcriptionquantitative polymerase chain reaction (RT-qPCR). Subsequently, the median LINC00472 expression in the pancreatic cancer tissues was used as cut-off to distinguish between the cancer tissues exhibiting high and low expression of LINC00472.

| Follow-up
The participants were followed up by telephone or on appointment basis. The follow-up deadline was February 2019. The overall survival (OS) of the patients was defined as the time starting from randomization until the death of patient caused by any reason. The 5-year OS of patients was calculated. As of the follow-up deadline, a total of five patients were during the follow-up period among the initial 70 patients, resulting in a follow-up rate of 89.55%.

| Cell culture and treatment
The normal immortalized human pancreatic epithelial cell line hTERT-HPNE (ATCC ® CRL-4023) and four human pancreatic  Table 1. For cell transfection, the cells were grown in 6-well plates until they attained 60% confluence, after which the cells were transfected with miRNA mimic or shRNA, or infected with the lentiviral vectors using the Lipofectamine 3000 kit (Invitrogen, Carlsbad, CA, USA).

| Dual-luciferase reporter assay
The potential binding sites of ZEB1 protein in LINC00472 promoter re-
After transfection, the pancreatic cancer cells were resuspended and seeded into 96-well plates at a density of 2 × 10 3 cells with

| Western blot analysis
The cells were harvested by trypsin detachment and lysed with the enhanced radioimmunoprecipitation assay lysis buffer (BOSTER,

| RNA pull-down assay
The cells were transfected with biotinylated WT miR-23a-3p and

| Chromatin immunoprecipitation (ChIP)
The ChIP kit (Millipore, Billerica, MA, USA) was used for this assay.
Upon attaining 70%-80% cell confluence, the cells were fixed in 1% formaldehyde at room temperature for 10 minutes to produce DNAprotein cross-linking. The cells were then subjected to ultrasonic treatment to produce chromatin fragments of appropriate size at 120 w, 2-s on and 5-s off each time, for a total of 15 cycles. Subsequently, the cells were subject to centrifugation at 28341.3 g at 4°C after which the supernatant was collected and divided into three tubes, which were incubated with the positive control antibody to RNA polymerase II, NC antibody to normal human IgG and rabbit anti-FOXO3 (at a dilution ratio 1:100, ab12162, Abcam, Cambridge, UK) overnight at 4°C. Protein Agarose/Sepharose was added for precipitation of the endogenous DNA-protein complex. After centrifugation and removal of the non-specific complex, the DNA-protein complex was subject to overnight incubation at 65°C to terminate cross-linking. Phenol/ chloroform was then added to extract and purify the isolated DNA fragments. Subsequently, RT-qPCR was conducted to determine the expression pattern of BID promoter. 20

| Immunohistochemistry
The paraffin-embedded tumour tissue sections were taken for immu-

| TUNEL assay
Tumour tissues were fixed with 4% paraformaldehyde, dehydrated by ascending degrees of alcohol, cleared using xylene, embedded in paraffin and cut into 5µm sections. The tissue sections were deparaffinized and proceeded in strict accordance with the conventional histological protocols. To terminate any peroxidase activity, the tissue sections were immersed in 3% hydrogen peroxide solution in ethanol for 15 minutes. After a rinse, the tissue sections were incubated with proteinase K for 20 minute at room temperature. After a rinse, the tissue sections were incubated with the reaction solution provided with the TUNEL staining kit. After another rinse, the tissue sections were incubated with diaminobenzidine for 15 minutes at room temperature and then stained using haematoxylin following another rinse. Lastly, the cells with brown nucleus were regarded as TUNEL-positive cells.

| Statistical analysis
All experimental data were analysed using SPSS 21.0 software (IBM Corp. Armonk, NY, USA). Measurement data were summarized by mean ± standard deviation from at least three independent experiments. Comparison between two groups was performed by the unpaired t test. Comparison among multiple groups was performed by one-way analysis of variance (ANOVA) with Tukey's post hoc test. Comparison among groups at different time-points was performed using two-way ANOVA or repeated-measures ANOVA with Bonferroni's post hoc test. Pearson's correlation analysis was adopted to analyse the correlation between ZEB1 and LINC00472, miR-23a-3p and LINC00472, and FOXO3 and BID. The Kaplan-Meier method was adopted to analyse the survival data, and the logrank test was conducted for comparison between different survival curves. In all statistical references, a value of P < 0.05 was indicative of a statistically significant difference.

| LINC00472 expression was inhibited by zeb1 in pancreatic cancer
Existing literature has demonstrated the functionality of LINC00472 as a tumour suppressor in various type of cancers including lung cancer, 9 colorectal cancer 10 and ovarian epithelial cancer 21 ; however, its role in pancreatic cancer remains unidentified. We initially analysed the GSE46234 data set by GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r/) and identified that LINC00472 was poorly expressed in pancreatic cancer ( Figure 1A), which was further tested and verified by an array of experiments in the clinical patient tissues.
The expression pattern of LINC00472 in the pancreatic cancer tissues and adjacent normal tissues was determined by RT-qPCR.
The results showed that the expression pattern of LINC00472 in pancreatic cancer tissues was reduced compared with the adjacent normal tissues ( Figure 1B). Using an analysis of the relationship between LINC00472 expression and the clinicopathological characteristics of the patients, we identified explicit correlations between the LINC00472 expression pattern and the patient's TNM stage and lymph node metastasis but not with gender and age (Table 3), and we further observed an inferior overall survival of patients with low expression ( Figure 1C). . J, Luciferase activity was detected dual-luciferase reporter assay. K, Luciferase activity was detected. Data between the cancer tissues and adjacent normal tissues were compared by the paired t test. Data between the other two groups were compared by the unpaired t test. Data among multiple groups were compared by one-way ANOVA with Tukey's post hoc test. The Kaplan-Meier method was adopted to calculate the overall survival of patients, and the survival difference of patients was analysed by log-rank analysis. * P < 0.05   The expression pattern of LINC00472 was down-regulated in pancreatic cancer, but the relative mechanism has not been determined yet. As an epithelial-mesenchymal transition (EMT) marker, ZEB1 could evidently promote the development of pancreatic cancer 16,22 and it functioned as a transcription factor to suppress the expression pattern of its corresponding downstream target genes. 23 In order to determine whether ZEB1 can regulate the expression pattern of LINC00472 in the progression of pancreatic cancer, we further examined the expression pattern of ZEB1 in the pancreatic cancer tissues and adjacent normal tissues by RT-qPCR.
The results displayed that, compared with the adjacent normal tissues, the expression pattern of ZEB1 was increased in the pancreatic cancer tissues ( Figure 1D), while a negative correlation was depicted between the ZEB1 expression pattern and the expression pattern of LINC00472 ( Figure 1E). Concurrently, RT-qPCR was

| LINC00472 inhibited tumorigenicity of pancreatic cancer cells in vivo
The stable infected cell line PANC-1 overexpressing LINC00472 was

| LINC00472 competitively bound to miR-23a-3p to up-regulate foxo3 expression
Accumulating evidence has elicited that LINC00472 can function as a ceRNA to regulate the occurrence and development of various diseases. 9,10 In the current study, in order to further examine the regulatory mechanism of LINC00472 in pancreatic cancer, the starBase website (http://starb ase.sysu.edu.cn/) was used to predict miRNAs that may bind to LINC00472. The obtained miRNAs were intersected with the up-regulated miRNAs in the GSE24279 data set, the results of which identified that miR-23a-3p could independently bind to LINC00472 (Figure 4A), and moreover, miR-23a-3p was highly expressed in pancreatic cancer ( Figure 4B). The specific binding sites between eLINC00472 and miR-23a-3p are shown in Figure 4C. The results of dual-luciferase reporter assay indicated that miR-23a-3p mimic inhibited the luciferase activity of LINC00472-WT without altering that of LINC00472-MUT ( Figure 4D). Meanwhile, a high expression pattern of miR-23a-3p was reported in pancreatic cancer. 12 Therefore, we speculated that LINC00472 may regulate the devel- The results showed that miR-23a-3p was up-regulated in the pancreatic cancer tissues ( Figure 4E), while the miR-23a-3p expression pattern was negatively correlated with the LINC00472 expression pattern ( Figure 4F). Besides, the results of RNA pull-down experiments demonstrated that the biotin-labelled miR-23a-3p probe can disintegrate LINC00472, thus verifying that LINC00472 could bind to miR-23a-3p ( Figure 4G).
Then, we predicted the potential target genes of miR-23a-3p through starBase and identified that miR-23a-3p could bind to FOXO3 3'UTR ( Figure 4H). A positive correlation was identified between LINC00472 and FOXO3 in pancreatic cancer ( Figure 4I). F I G U R E 4 LINC00472 promotes the expression of FOXO3 through competitive binding to miR-23a-3p. A, Venn diagram of miRNAs binding to LINC00472 predicted by the starBase website and up-regulated miRNAs in the GSE24279 data set. B, miR-23a-3p expression pattern in the normal and tumour samples in the GSE24279 data set. C, The binding sites between LINC00472 and miR-23a-3p predicted by the starBase website. D, The binding between LINC00472 and miR-23a-3p verified by dual-luciferase reporter assay. E, RT-qPCR determination of the expression pattern of miR-23a-3p in 70 pancreatic cancer tissues and adjacent normal tissues. F, Correlation analysis of LINC00472 expression pattern and miR-23a-3p expression pattern in the pancreatic cancer tissues. G, RNA pull-down assay was used to verify the binding between LINC00472 and miR-23a-3p. H, The binding site between miR-23a-3p and FOXO3 3'UTR predicted by starBase website. I, Correlation analysis of LINC00472 expression and FOXO3 expression analysed by the starBase website. J, RNA pull-down assay was used to verify the binding between miR-23a-3p and FOXO3. K, Dual-luciferase reporter gene assay was used to confirm the binding between miR-23a-3p and FOXO3 3'UTR. L, After overexpressing LINC00472 in the pancreatic cancer cell PANC-1, RNA pull-down was used to detect changes in miR-23a-3p binding to LINC00472 or FOXO3. M, After silencing LINC00472 in pancreatic cancer cell BXPC3, RNA pull-down was used to detect changes in miR-23a-3p binding to LINC00472 or FOXO3. N, The expression patterns of LINC00472, miR-23a-3p and FOXO3 were determined by RT-qPCR. O, Western blots of LINC00472, miR-23a-3p and FOXO3. Data between two groups were compared by the unpaired t test. Data among multiple groups were compared by one-way ANOVA with Tukey's post hoc test. Pearson's correlation analysis was used to analyse the correlation between LINC00472 and miR-23a-3p. * P < 0. ( Figure 4K). The results of RNA pull-down assay revealed that the binding between miR-23a-3p and LINC00472 was enhanced after LINC00472 overexpression, while the binding between miR-23a-3p and FOXO3 was attenuated ( Figure 4L). Conversely, such alterations can be neutralized by silencing LINC00472 ( Figure 4M). Finally, an increased expression pattern of FOXO3 was identified after overexpressing LINC00472 in the PANC-1 cells, while the expression pattern of FOXO3 was reduced after overexpressing miR-23a-3p. However, simultaneous overexpression of miR-23a-3p could annul the regulatory effects on FOXO3 expression pattern induced by overexpressing LINC00472, while LINC00472 could neutralize the weakening effects on FOXO3 expression caused by overexpressing miR-23a-3p separately ( Figure 4N-O). To summarize, the preceding results indicated that LINC00472 increased the FOXO3 expression by competitively binding to miR-23a-3p.

| LINC00472 silencing enhanced the proliferation and impeded apoptosis of pancreatic cancer cells by down-regulating FOXO3
To investigate whether LINC00472 could mediate FOXO3 to influence pancreatic cancer cell functions, LINC00472 was silenced in the pancreatic cancer cell BXPC3 overexpressing FOXO3, after which RT-qPCR was conducted to determine the expression patterns of LINC00472, miR-23a-3p and FOXO3, while Western blot analysis was conducted to determine the protein expression pattern of FOXO3. The expression pattern of FOXO3 was decreased, while the expression pattern of miR-23a-3p was increased upon individual LINC00472 silencing. The expression pattern of FOXO3 reduced by LINC00472 silencing could be annulled by oe-FOXO3 transfection ( Figure 5A

| FOXO3 transcriptionally activated BID expression
As a transcription factor, FOXO3 can initiate the expression patterns of its target genes. 24 The the JASPAR website predicted the  Subsequently, we predicted FOXO3 as a downstream target of miR-23a-3p in pancreatic cancer through the starBase database.
Their binding relationship was ascertained by RNA pull-down, while the binding site was validated by dual-luciferase reporter gene assay. Moreover, miR-23a-3p has been reported to function as an oncogenic regulator of pancreatic cancer. 12 Existing literature has reported that FOXO3 is poorly expressed in pancreatic cancer, thus signifying its vital significance in tumorigenesis and cancer development, 36 which was consistent with our finding.
Essentially, miR-23a-3p could evidently target the downstream target gene so as to regulate cancer progression. 37  we suggested that LINC00472 may up-regulate the expression of BID through FOXO3 to induce the apoptosis of pancreatic cancer.
Such regulatory axis was verified in both in vitro and in vivo settings, and our data demonstrated that LINC00472 silencing had fundamentally suppressed BID expression through miR-23a-3pmediated inhibition of FOXO3 to promote proliferation and inhibit the apoptosis of pancreatic cancer cells.
To conclude, our data presented that LINC00472, a lncRNA down-regulated by ZEB1 in pancreatic cancer, could function as tumour suppressor in pancreatic cancer, thus establishing the significance of LINC00472 as a therapeutic target against pancreatic cancer. Essentially, the identification of the regulatory axis that LINC00472 competitively binds to miR-23a-3p to induce FOXO3 and activate BID contributes to a comprehensive understanding of the underlying mechanisms involved in tumorigenesis and progression ( Figure 9).

ACK N OWLED G EM ENTS
The authors thank the reviewers for their helpful comments.

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data sets generated during the current study are available.