Tissue samples in this study were all gained from patients experiencing operative treatments in the First Affiliated Yijishan Hospital of Wannan Medical College (Wuhu, Anhui province, China), from May 2017 to July 2021. Our research group randomly selected 119 both PC and paired normal peritumoral (NP) tissue samples from patients without preoperative chemotherapy or radiotherapy. Clinicopathological data of patients collected in this study are all shown in Table 1 . Operation application on those patients included pancreatectomy as well as choledochojejunostomy and gastroenterostomy, according to the criteria of the National Comprehensive Cancer Network (NCCN 2019) guideline for PC (31). Histopathology from the Department of Pathology was applied to finalize the diagnosis of PC. The samples were gathered from the PC tissues by resection or palliative surgery. A tissue biopsy gun (MG15-22, Tempe, AZ, USA) was also used to assemble PC tissues for unresectable PC patients. Next, these samples were all embedded in paraffin or stored by liquid nitrogen cryotherapy. All protocols were accepted in advance by the ethics committee of our hospital.

The PC cell lines containing PANC-1, BXPC-3, AsPC-1, SW1990, and MIAPaCa-2 cell lines were all acquired from the American Type Culture Collection (ATCC, Rockville, MD, USA), while human pancreatic duct epithelial (HPDE) cells were obtained from China Beijing Be-Na Culture Collection (BNCC, Beijing, China). PC cells were authenticated and identified by DNA fingerprinting in recent 6 months, and these cell lines were passaged for less than 6 months after resuscitation. Next, we cultivated these cell lines in HyClone RPMI-1640 medium matched up with 10% characterized fetal bovine serum and penicillin (100 U/ml)–streptomycin (100 mg/ml) combination in a steady incubator of unflagging 37°C. Hypoxia simulation models were designed as our previous study (11, 12, 19), which contain two protocols: (1) hypoxia induced by fixed concentration (100 µM) of CoCl₂ solution and (2) hypoxia induced in modular incubator chamber. The specific protocols were stated thoroughly (32).

Human short interfering RNAs (RNAi) for HIF-1α (siHIF-1α), ZEB1-AS1 (siZEB1-AS1), ZEB1 (siZEB1), histone deacetylase 1 (HDAC1) (siHDAC1), and negative controls (siNC) were obtained from Ribo Biological Company (Guangzhou, China). The plasmids including HIF-1α, ZEB1-AS1, ZEB1, and HDAC1 and the corresponding NC plasmid were synthesized and purchased from GeneChem Company (Shanghai, China). Based on provided instructions, lipofectamine 2000 (Invitrogen, USA) were utilized for transfections. Lentivirus vector containing ZEB1-AS1-siRNA (LV-siZEB1-AS1) and matched NC (LV-siNC) were gathered from China GeneChem Co. as well. Protein and total RNA were extracted at 48 h posttransfection after BXPC-3/PANC-1 cells were incubated in a solution with polybrene and lentivirus. Total sequences of siRNAs are exhibited on Supplementary Table S1 .

Total RNAs were extracted from samples or cells by means of RNAiso Plus via official protocols (TaKaRa, Dalian, China). All mRNAs and lncRNAs were reverse transcribed through the protocol of the PrimeScript^®RT Master Mix Perfect Real Time (TaKaRa, Dalian, China) and One Step PrimeScript^® miRNA cDNA Synthesis Kit (Perfect Real Time) (TaKaRa, Dalian, China), following qRT-PCR analysis with the SYBR Premix Ex Taq II (TaKaRa, Dalian, China) as the official protocol. Expression levels of lncRNA and mRNA were measured via algorithm of 2^−△△CT. This research applied GAPDH to standardize the level of lncRNA and mRNA expression. All qRT-PCR tests were independently tested three times. Total primers are exhibited in Supplementary Table S2 .

Pancreatic cancer cells were treated with RNA polymerase II inhibitor α-amanitin (50 μM, Sigma-Aldrich, St. Louis,MO, USA) RNAs to block new synthesis. We then collected cells at different time points (0, 6, 12, 18, and 24 h), and total RNA were extracted by using Trizol reagent (TaKaRa, Dalian, China). The levels of RNA were recorded by qRT-PCR and normalized to 18S rRNA, the percentage of the remaining RNA to test RNA degradation were measured relative to time 0.

A proven technique of MTT assay was employed to measure the proliferation ability of PC cells. Cells were cultivated in 96-well plate (2,000/well) containing eight vices well corresponding with each sample. After cells were stably planted at 37°C, the proliferation of cells was observed and recorded from the next 1 to 5 days. For detecting the ability of proliferation of PC cells, we cultured cells by adding 20 μl MTT (5 mg/ml) to each well for 4 h. After that, we replaced the mixture with 150 μl DMSO (Sigma, USA). Following complete dissolution of the crystal in each well, ELISA reader was adopted to record the absorbance at 570 nm.

In this study, a wound-healing assay was used to evaluate the migration capacity. We seeded PC cells into 12-well plates, and cells rise to 90% confluence manually. The cell monolayer was scratched by the tip of a micropipette and gowned in a serum-free solution. After incubation for 24 h, cell migration images were measured at 0, 24, and 48 h based on the width of the wound scratches. To explore the invasion ability, Transwell migration assays (Corning-Costar, NY, USA, pore size 8 μm) mixed with Matrigel (Sigma, USA) were applied to determine the cell invasion capacity. We placed 5 × 10⁴ PC cells, coated with 250 μl serum-free medium in the upper chamber, while the lower chamber with 700 μl solution blended with 30% FBS. Following 2 days of fixation and staining, we collected the lower chamber and further calculated the number of cells from nine fields via a CX33 biological microscope (Olympus, Beijing, China).

Western blot analysis has been done through repeated experiments previously (19). Protein extracts from PC cells and the concentration of protein were quantified using a BCA assay. Next, the appropriate denatured protein (30 µg per well) was separated in polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membranes (PVDF). For this study, antibodies were recorded as follows: Rabbit anti-ZEB1 (1:1,000, catalog#3396, Abcam, Cambridge, MA, USA), HIF-1α (1:1,000, catalog#279654, Abcam), HDAC-1 (1:1,000, catalog#10197-1-AP, Proteintech, Rosemont, IL, USA), PAN-AC (1:1,000, catalog#A2391, ABclonal, Woburn, MA, USA), GAPDH (1:1,000, catalog#10494-1-AP, Proteintech). On the other hand, coimmunoprecipitation was adopted to evaluate the combining capacity on protein level. Lysates from PC cells were blended with control mouse/rabbit IgG or primary antibodies at steady-state level of 4°C throughout the night. We mingled lysates with Protein A/G PLUS-Agarose (Santa Cruz Co., Beijing, China) at 4°C for 1.5 h. Next, we separated agarose and then acquired even volumes of each sample using a lysis buffer. Western blot analysis was used for further assessments. All Western blot reactions were independently tested in triplicate.

In this study, chromatin immunoprecipitation (ChIP) assay was conducted by adopting the EZ-ChIPTM kit (Millipore, Billerica, MA, USA), containing anti-HIF-1α and anti-RNA polymerase II antibodies (catalog#ab264350, Abcam) in accordance with manufacturer’s specification. We employed matching IgG as controls. Later, we amplified bound DNA by PCR techniques and obtained results through electrophoretic separation on a 2% agarose gel. All ChIP assays were independently tested in triplicate.

For single-molecule RNA fluorescence in situ hybridization (FISH), we purchased The FISH Tag™ RNA Multicolor Kit (Invitrogen, USA) and MAXIscript^® Kit (ThermoFisher, Waltham, MA, USA). The course of experiments of RNA FISH was applied according to provided instructions (33, 34). In this study, nucleus of PC cells was stained with DAPI and red fluorescent probe was synthesized to identify ZEB1-AS1. Hybridization with probes was performed all night at 55°C. A laser scanning microscope (Carl Zeiss, Jena, Germany) was used for recording the stained results of FISH.

How HIF-1α modulates the binding activity of a potential hypoxia response element (HRE) on the ZEB1-AS1 promoter during normoxia/hypoxia medium was evaluated by luciferase reporter assays. We transfected wild-type (WT) HRE sequence and mutant (MUT) HRE sequence into PANC-1 cells. Next, these cells were transfected with siNC and siHIF-1α which expressed pGL3-based construct involving the HRE of ZEB1-AS1. Dual-Luciferase Report Assay kit (Promega, Madison, WI, USA) was used for luciferase activity evaluation based on the manufacturer’s instruction (11). The intensity of the luciferase activity was standardized to Renilla luciferase. All researches were independently tested three times.

Immunohistochemistry was conducted as a previously described method. Paraffin-embedded PC tissue sections from patients were collected, dried, dewaxed, and rehydrated. Sections were incubated with the primary antibody to ZEB1 (1:1,000, catalog#3396, Abcam) or primary anti-HIF-1α antibody (1:1,000, catalog#279654, Abcam) and a horseradish peroxidase-conjugated secondary antibody (1:200, rabbit anti-goat). Immunohistochemical staining samples of HIF-1α and ZEB1 were estimated under CX33 biological microscope (Olympus, Beijing, China). Intensity score was calculated for each sample from 0 to 3 (units of intensity) as previously described (12), and the average intensity score was further examined for each PC patient.

For animal experiment, we implanted PC cells, which are steadily transfected with LV-siZEB1-AS1#1, LV-siZEB1-AS1#2, or LV-siNC, subcutaneously in 4-week-old nude BALB/c mice based on 2 × 10⁶ cells per mouse (Beijing HFK Bio-Technology Co., Beijing, China). Five mice comprised each group and reared for 27 days. Tumor volumes (0.5 × L (length) × W² (width)) were calculated every 3 days. Tumor weights were weighed after adjacent sacrifice of mice. Solid tumor tissues containing lungs and liver of the mice were excised, and we further stained the samples with H&E. The expression of ZEB1 and ZEB1-AS1 were calculated from five sections of metastasis tissue in each mice. Our in vivo experiments were approved by the Animal Research Committee of the First Affiliated Yijishan Hospital of Wannan Medical College.

IBM SPSS Statistics (SPSS v21.0) was applied for all analyses. Data analyses are all based on means ± SD. Specifically, group differences were evaluated by t-test. Paired t-test was employed to compare ZEB1-AS1 in paired PC tissues. The correlation between ZEB1-AS1 and ZEB1 expression was measured by Pearson’s correlation. We used Chi-square test to analyze the relation between ZEB1-AS1 and clinical characteristics. Kaplan-Meier approach was performed to describe the patients’ survivals. Further research of receiver operating characteristic curve analysis (ROC) was applied to evaluate the ability of biomarks to forecast the mortability risk of PC and area under the curve (AUC) was used to measure predictive capability (35). All assays were independently tested in triplicate. Significant differences were significant at ^* p < 0.05 and ^** p < 0.01 as marked significance.

To determine the underlying dysregulated lncRNAs that induce the tumorigenesis of PC, we analyzed a lncRNA microarray. Compared with paired noncancerous peritumoral (NP) tissues, hierarchical clustering data demonstrated that lncRNA-ZEB1-AS1 (ZEB1-AS1) was one of the significantly upregulated lncRNAs in PC samples ( Figure 1A ). Compared with the normal HPDE cell line, the expression of ZEB1-AS1 in five PC cell lines (BXPC-3, PANC-1, SW1990, MIAPaCa-2, and AsPC-1) was significantly higher ( Figure 1B ). Moreover, FISH analysis was performed to determine whether ZEB1-AS1 is located in the nucleus or cytoplasm of BXPC-3/PANC-1 cells. It was indicated that ZEB1-AS1 was mainly located in the nucleus ( Figure 1C ). Although ZEB1-AS1 was a well-known oncogene in many tumors, there has been little research on PC. Hence, it was selected for further investigation in consideration of the genomic location on ZEB1 and ZEB1-AS1 ( Figure 1D ). The full-length ZEB1-AS1 sequence was determined, and the secondary structure of ZEB1-AS1 was determined with minimum free energy (MFE) by searching the NONCODE online database ( Supplementary Figures S1A, B ). Online tools, including the Coding Potential Assessment Tool (CPAT) and Coding Potential Calculator (CPC), predicted that ZEB1-AS1 is a potential noncoding RNA ( Supplementary Figures S1C, D ).

To further verify the microarray results, ZEB1-AS1 levels were measured in 119 paired PC and NP tissues using real-time PCR analysis. The ZEB1-AS1 transcript levels in the PC tissues were significantly higher than those in paired noncancerous tissues ( Figure 1E ). After blocking RNA synthesis through the RNA polymerase II inhibitor, α-amanitin, in PANC-1 cells, the levels of ZEB1-AS1 were evaluated. The results indicate that α-amanitin treatment decreased ZEB1-AS1 level significantly after 24 h, instead of the level of 18S mRNA. The results verified that ZEB1-AS1 synthesis was regulated by RNA polymerase II rather than RNA polymerase I ( Figure 1F ).

Because ZEB1-AS1 is upregulated in PC tissues, short interference siRNAs for ZEB1-AS1 (siZEB1-AS1) were utilized to suppress the levels of ZEB1-AS1 in PC cells. Two siZEB1-AS1 siRNAs targeting ZEB1-AS1 were designed, and they significantly knocked down endogenous ZEB1-AS1 expression and were used for further experiments ( Figure 2A ). To elucidate potential impacts of ZEB1-AS1, we transfected siZEB1-AS1 and a ZEB1-AS1 overexpression (ZEB1-AS1-OE) vector into BXPC-3 and PANC-1 cells. Knockdown of ZEB1-AS1 suppressed the proliferation and migration of BXPC-3 and PANC-1 cell lines ( Figures 2B, C ). In addition, Matrigel-coated Transwells confirmed that ZEB1-AS1 depletion caused a significant inhibition of the invasion ability of BXPC-3 and PANC-1 cells ( Figure 2D ). Furthermore, transfection of ZEB1-AS1-OE into BXPC-3 and PANC-1 cells resulted in overexpression of ZEB1-AS1 compared with the vector control ( Supplementary Figure S2A ). Moreover, overexpression of ZEB1-AS1 markedly induced the proliferation, migration, and invasion capacity of these cells ( Supplementary Figures S2B–D ). Taken together, these results demonstrated that ZEB1-AS1 promotes the proliferation capacity and invasion ability of PC cells.

Many studies have established that lncRNAs modulate the levels of neighboring genes by cis regulation. As ZEB1-AS1 is near the ZEB1 gene, which is a tumorigenic driver in several cancers, we hypothesized that ZEB1-AS1 exerts a cis-acting role on ZEB1 expression. The expression of ZEB1 increased with prolonged culture in a hypoxia microenvironment from 0 to 48 h ( Figure 3A ). However, the expression quantity of ZEB1 under normoxia conditions from 0 to 48 h of culture did not show a significant difference ( Supplementary Figure S3A ). In BXPC-3 and PANC-1 cells, ZEB1-AS1 knockdown significantly decreased ZEB1 mRNA and protein expression ( Figure 3B ). It is inspiring to note that overexpression of ZEB1-AS1 showed a similar trend ( Figure 3C ).

Hypoxia medium (hypoxia microenvironment undergoing 48 h) was used for further experiments. In this study, hypoxia medium increases the mRNA and protein levels of ZEB1-AS1-induced ZEB1 expression ( Figure 3D ), indicating that ZEB1 is a potential target of ZEB1-AS1 under hypoxia conditions. Similarly, the invasion and migration capacities of BXPC-3 and PANC-1 cells were upregulated in ZEB1-AS1-overexpressing cells under hypoxia conditions, and these effects were weakened by si-ZEB1 ( Figures 3E, F ). Together, these results demonstrated that ZEB1 is critical for ZEB1-AS1 to exert its oncogenic and malignant functions in PC cells under hypoxia conditions.

Next, we investigated the potential mechanism by which hypoxia conditions control ZEB1-AS1 expression in PC cells. Recent studies have investigated whether variable expression of lncRNAs is attributed to hypoxia conditions in a variety of cancer cells. Because our previous research indicated that hypoxia facilitates lncRNA-NUTF2P3 enrichment (11), we investigated whether transcriptionally upregulated ZEB1-AS1 is closely related to the hypoxia environment of PC. Genomic structure analysis indicated one predictable HRE in the promoter region of ZEB1-AS1, suggesting that ZEB1-AS1 is a hypoxia-responsive lncRNA ( Figure 4A ). The levels of ZEB1-AS1 also statistically increased under hypoxia medium but are insignificantly different under normoxia environment in the same time course (from 0 to 48 h) of the experiment ( Supplementary Figures S3B, C ).

After treatment with hypoxia medium or chemically inducing hypoxia with CoCl₂ (100 µM) for 48 h, the level of ZEB1-AS1, in accordance with HIF-1α expression, was significantly increased in BXPC-3 and PANC-1 cells ( Figure 4B ). Moreover, HIF-1α knockdown suppressed hypoxia-induced ZEB1-AS1 upregulation in PC ( Figures 4C, D ). ChIP assays showed that the HRE region in the ZEB1-AS1 promoter modulated HIF-1α binding to the ZEB1-AS1 promoter area, which was further increased under hypoxia conditions ( Figure 4E ). Pol-II was also shown to bind to the HRE of the ZEB1-AS1 promoter region. Similarly, this binding was enhanced during hypoxia medium, indicating a protranscriptional ability of this binding ( Figure 4F ). Furthermore, RNA-FISH assay demonstrated that hypoxia-induced ZEB1-AS1 overexpression was suppressed through HIF-1α knockdown ( Figure 4G ). To further verify the functional binding of HIF-1α to the ZEB1-AS1 promoter region, we performed a luciferase reporter assay in PANC-1 cells, which were transfected with luciferase reporter vectors containing the WT promoter of ZEB1-AS1 or MUT promoter of ZEB1-AS1. Hypoxia medium remarkably increased the luciferase activity in cells transfected with WT but not MUT ZEB1-AS1. However, knockdown of HIF-1α decreased the hypoxia-induced luciferase activity of PC cells transfected with WT ZEB1-AS1 ( Figure 4H ). Thus, these findings suggested that ZEB1-AS1 is transcriptionally regulated by HIF-1α in PC cells in hypoxia medium.

The relationship between ZEB1-AS1 and HIF-1α was further explored to determine the mechanism by which ZEB1-AS1 affects the interaction between HIF-1α and the ZEB1 promoter. The results showed that the protein level of HIF-1α was suppressed by the knockdown of ZEB1-AS1, indicating that HIF-1α may be a direct target of ZEB1-AS1 in BXPC-3 and PANC-1 cells. However, downregulation of ZEB1-AS1 did not affect the mRNA expression of HIF-1α, and a similar experimental result was repeated by RNA agarose gel electrophoresis ( Figure 5A ). These findings indicated that ZEB1-AS1 influenced the level of HIF-1α at the posttranscriptional level and not at the transcriptional level. Furthermore, we investigated how ZEB1-AS1 regulates the protein stability of HIF-1α. The HIF-1α protein level was measured following cycloheximide (CHX) treatment, which disrupts protein synthesis. During CHX treatment, the stability of HIF-1α was inhibited via downregulation of ZEB1-AS1 in a time-dependent manner ( Figure 5B ). Next, MG132 (proteasome inhibitor) was used to rescue the decrease in HIF-1α protein levels in ZEB1-AS1 knockdown PC cells ( Figure 5C ). We further verified that ZEB1-AS1 knockdown significantly induced HIF-1α acetylation and decreased the ZEB1 and HIF-1α binding under hypoxia conditions ( Figure 5D ). Moreover, we demonstrated that ZEB1 overexpression rescued the ZEB1-AS1 knockdown-induced downregulation of HIF-1α and decreased HIF-1α and ZEB1 binding, and it reduced the acetylation level of HIF-1α ( Figure 5E ).

HDAC1 interacts with HIF-1α and decreases the protein level of HIF-1α. Research has shown that the transcriptional repressor, ZEB1, recruits HDAC1-containing corepressor complexes (CRC) to decrease transcription of the cadherin 1 (CDH1) gene and downregulate E-cadherin in PC. Thus, we hypothesized that HDAC1 participates in ZEB1-AS1-promoted destabilization of HIF-1α in PC cells. Coimmunoprecipitation (CoIP) demonstrated that HIF-1α and HDAC1 bind to each other and that this binding is intensified by hypoxia conditions ( Supplementary Figure S4A ). This result indicated that HDAC1, HIF-1α, and ZEB1 may form a ZEB1-AS1-induced complex. To investigate whether HDAC1 is involved in the regulation of HIF-1α protein expression by ZEB1, we used trichostatin A (TSA), a specific inhibitor of HDACs, to assess the stability of HIF-1α protein. The results showed that TSA aggravated the inhibitory effect of ZEB1 knockdown on HIF-1α protein levels under hypoxia conditions in PC ( Supplementary Figure S4B ). Consistently, TSA downregulated the upregulation of HIF-1α at the protein level, which was promoted by ZEB1 overexpression ( Supplementary Figure S4C ). In addition, overexpression of HDAC1 rescued the suppression of HIF-1α expression by ZEB1 knockdown. However, HDAC1 knockdown inhibited HIF-1α expression, which was upregulated by ZEB1 overexpression ( Supplementary Figures S5A, B ). To evaluate the effect of HDAC1 on the interaction of HIF-1α with ZEB1, we measured the acetylated levels of HIF-1α via a CoIP assay. The results revealed that ZEB1 repressed TSA-induced HIF-1α acetylation ( Supplementary Figure S5C ). In contrast, TSA-induced si-ZEB1 accelerated HIF-1α protein acetylation ( Supplementary Figure S5D ). Together, these results suggested that ZEB1-AS1 promotes an interaction among HDAC1, HIF-1α, and ZEB1, which inhibits acetylation of HIF-1α by promoting the deacetylase capacity of HDAC1, further resulting in HIF-1α stabilization.

To further evaluate the role of ZEB1-AS1, PANC-1 cells stably transfected with a lentiviral vector including a NC sequence (LV-siNC) or ZEB1-AS1-siRNA sequences (LV-siBZEB1-AS1#1 or LV-siBZEB1-AS1#2) were implanted into 4-week-old nude mice. Compared with the LV-siNC group, the tumor size of the LV-siZEB1-AS1 groups was smaller. Simultaneously, the visible number of liver or lung metastases were lower in the LV-siZEB1-AS1 groups ( Figure 6A ), and the tumor growth rate and weight of the LV-siZEB1-AS1 groups were suppressed ( Figures 6B, C ). qRT-PCR analysis showed that the transcription levels of ZEB1-AS1 and ZEB1 in the LV-siZEB1-AS1 group were significantly lower than those of the LV-siNC group ( Figures 6D, E ). Next, we found that both the number of mice with lung or liver metastases and the number of distinct metastatic nodes in the LV-siZEB1-AS1 groups were significantly reduced compared with the control mice ( Figures 6F, G ). We further verified the above results via H&E staining and further detected the RNA expression quantity of ZEB1 and ZEB1-AS1 in lung or liver metastases ( Supplementary Figure S6 ). Similarly, LV-SiZEB1-AS1 suppressed the proliferation and metastasis ability of BXPC-3 cells in vivo ( Supplementary Figures S7 and S8 ). Overall, these results confirmed the tumorigenic mechanism of ZEB1-AS1 in PC cells.

The clinicopathological results showed that the overexpression level of ZEB1-AS1 was positively associated with histological grade, lymphatic invasion, and distant metastasis in PC patients ( Table 1 ). Furthermore, Kaplan-Meier analysis revealed that PC patients with high expression of ZEB1 experienced a shorter overall survival (OS) time than those patients with low ZEB1 expression ( Figure 7A ). The Kaplan-Meier study also demonstrated a correlation between higher ZEB1-AS1 mRNA expression and shorter OS ( Figure 7B ). In addition, we found that ZEB1-AS1 and ZEB1 expression suggested a positive relationship in 119 patients with PC, which was assessed via Pearson’s correlation analysis or Chi-square test ( Figures 7C, D ). Moreover, the survival analysis indicated that low expression levels of both ZEB1 and ZEB1-AS1 were most beneficial to the OS of patients followed by the high expression level of either ZEB1 or ZEB1-AS1, while overexpression levels of both ZEB1 and ZEB1-AS1 had the worst effect ( Figure 7E ). We further performed a ROC curve analysis of the predictive value for OS. The results confirmed that the combination of ZEB1 and ZEB1-AS1 showed a preferable additive value ( Figure 7F ). Immunohistochemical (IHC) assays demonstrated that the protein expression of both ZEB1 and HIF-1α positively correlated with the expression levels of ZEB1-AS1 ( Figures 7G, H ). TCGA data from ChIPBase V2.0 (36) revealed that HIF-1α expression had a positive relationship with ZEB1 levels in PC ( Figure 7I ). Therefore, these results suggested that HIF-1α/ZEB1-AS1/ZEB1/HDAC1 signaling is involved in the oncogenesis and metastasis of PC ( Figure 7J ).