LncRNA HIF1A-AS1 Promotes Gemcitabine Resistance of Pancreatic Cancer by Enhancing Glycolysis through Modulating the AKT/YB1/HIF1α Pathway

Pancreatic cancer is one of the most lethal diseases in developed countries with a five-year survival rate of less than 6%, and it is expected to become the second leading cause of cancer-related deaths in 2030 (1, 2). Most patients with pancreatic cancer are unaware of physical deterioration until the disease develops to an advanced stage. The low survival rate is mainly attributed to an early-stage local infiltration and distant metastasis at first diagnosis (3). More importantly, poor response to therapies due to development of resistance in pancreatic cancer remains a significant clinical challenge and contributes to overall poor patient prognosis. Gemcitabine (GEM) alone or in combination with other drugs is the current standard for advanced metastatic pancreatic cancer, such as GEM/nab-paclitaxel, GEM/capecitabine, and GEM/erlotinib (4–6). However, many patients with pancreatic cancer show initial sensitivity to GEM treatment, followed by rapid development of drug resistance, which results in drastic reduction in patient survival (7). Researchers have tried to explore the mechanism of GEM resistance from multiple perspectives, including drug transport, drug metabolism, and signaling pathway (8). For example, high human equilibrative nucleoside transporter 1 (hENT1), as a key nucleoside transporter, has been shown a significant role in GEM resistance (9). Deoxycytidine kinase is the main rate-limiting enzyme for intracellular activation and metabolism of GEM, and its inactivation is one of the key mechanisms of GEM resistance (10). Meanwhile, overactivation of the PI3K/AKT pathway has been confirmed to involve in GEM resistance by several selective inhibitors (11). Nevertheless, it is still urgent to clarify the underlying mechanisms contributing to chemoresistance of pancreatic cancer, which may afford promising targets for therapeutic treatment.

The long noncoding RNAs (lncRNA) are a kind of noncoding RNAs with a size of longer than 200 nucleotides and usually losing protein-coding potential (12). Multiple studies showed that lncRNAs have emerged as an essential regulator in almost every aspect of cancer cell capacity for tumor initiation, growth, metastasis, and chemoresistance (13, 14). Especially, study had revealed that lncRNAs were dysregulated in GEM-resistant (GEM-R) pancreatic cancer cell lines, indicating that lncRNAs serve as novel regulators of GEM sensitivity (15). For example, lncRNA growth arrest-specific transcript 5 was indicated to reverse tumor stem cell–mediated GEM resistance by targeting miR-221/SOCS3 in pancreatic cancer (16). LncRNA Homosapiens glutathione S-transferase mu 3, transcript variant 2 (GSTM3TV2) acted as a key competing endogenous RNA to enhance chemoresistance by upregulating L-type amino acid transporter 2 (LAT2) and oxidized low-density lipoprotein receptor 1 in pancreatic cancer (17). Another study reported that lncRNA LINC00346 could promote GEM resistance of pancreatic cancer, which was partly mediated by antagonization of miR-188–3p and induction of BRD4 (18). Although these results reminded that lncRNAs might serve as essential regulator in chemoresistance of pancreatic cancer, the functional roles and mechanisms of lncRNAs in the chemoresistance of pancreatic cancer to GEM treatment remain poorly understood and thus need to be further clarified.

In the present study, we established GEM-R pancreatic cancer cell lines (BxPC3^GEM-R/PANC1^GEM-R) by increasing GEM treatment in medium. LncRNA-seq revealed that HIF1A-AS1, the antisense RNA of HIF1α, was obviously increased in BxPC3^GEM-R cells compared with parental BxPC3 cells. Thereby, we presumed that HIF1A-AS1 might be a crucial onco-lncRNA in GEM resistance of pancreatic cancer cells. Coincidently, our study revealed that HIF1A-AS1 facilitated GEM resistance of pancreatic cancer cells in a glycolysis-dependent manner by regulating the expression of HIF1α. The potential mechanism that HIF1A-AS1 regulated the expression of HIF1α was further investigated. Furthermore, the regulatory mechanism of upregulation of HIF1A-AS1 in GEM-R pancreatic cancer cells was explored.

The cell lines (BxPC3, PANC1) were purchased from the ATCC. Cells were tested and authenticated for genotypes by DNA fingerprinting within 6 months. Cells were cultured in RPMI-1640 (Gibco) containing 10% FBS (Gibco) 5% CO₂. To induce hypoxia, cells were cultured with 1% O₂ or treated with CoCl₂ (400 μmol/L). All cultures were monitored routinely and found to be free of contamination by Mycoplasma or fungi.

GEM (APExBIO) was used to establish the chemotherapy-resistant pancreatic cancer cell lines (BxPC3^GEM-R/PANC1^GEM-R). Specifically, parental BxPC3/PANC1 cells were exposed to an initial GEM concentration of 0.1 μmol/L in RPMI-1640 plus 10% FBS. Then, the concentration of GEM of the surviving population of cells was sequentially increased to 0.5, 1.0, 2.0, 4.0, and 8.0 μmol/L, and finally to the clinically relevant plasma concentration of 20 μmol/L. A serial cell viability assay was used to assess the level of GEM resistance.

For IC₅₀ assays, cells were cultured at 5,000 per well in 96-well plates with fresh medium. The cells were given corresponding concentrations of drug and cultured for 48 hours. Cell Counting Kit-8 (Sangon Biotech) was used to measure drug sensitivity at 450 nm using a microplate reader (Thermo Fisher Scientific) after incubating at 37°C for additional 2–4 hours.

For glucose uptake assay, cells were treated with 2-NBDG (50 μmol/L; APExBIO) for 1 hour, followed by FACS analysis using BD LSRFortessa X-20 Flow Cytometer (BD Biosciences). Cellular glucose uptake was quantified using fluorescence intensity. For lactate production assay, cells were seeded in 6-well plates and cultured for 24 hours. Then, lactate concentration in the media was determined using Lactic Acid Assay Kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, China). Lactate production was expressed as lactate concentration per 1 × 10⁴ viable cells. Meanwhile, the pH of the media was measured by INESA REX PHSJ 4F pH meter (REX).

Cells were incubated with 100 μg/mL cycloheximide (MedChemExpress) at 37°C for 30 minutes, and then were pelleted and lysed on ice with polysome lysis buffer. The lysate was collected, loaded onto a 10%–50% (w/v) sucrose gradient solution prepared in lysis buffer, and then centrifuged at 4°C for 4 hours at 30,000 rpm using ultra-centrifuge Optima L-70 and SW 41 Ti Rotor (Beckman Coulter). The sample was then fractioned and collected into 12 1-mL fractions, and analyzed by Gradient Master software (BioComp Instruments) equipped with an EM1-Econo UV monitor (Bio-Rad) as well as fraction collector FC203B (Gilson). RNA was purified by TRIzol (TaKaRa) from each fraction and subjected to qRT-PCR analysis.

Biotin-labeled antisense and sense RNA probes for HIF1A-AS1 or HIF1α mRNA were produced using the FISH Tag RNA Multicolor (Invitrogen) and MAXIscript (Ambion) kits. The probes for HIF1A-AS1 and HIF1α mRNA were transcribed in vitro with the T7 RNA polymerase, then were labeled with green and red fluorescence, respectively. Labeled probes were hybridized overnight with cell samples at 55°C in a humid environment. Stained results were observed by using LSM 5 Pascal Laser Scanning Microscope (Carl Zeiss). Probe sequences are given in Supplementary Table S1. If necessary, the HIF1α and nucleolin (Proteintech) protein immunofluorescence was conducted after the FISH assays.

RNA-binding protein immunoprecipitation (RIP) assays were performed using the Magna RIP RNA-binding protein immunoprecipitation kit (Merck Millipore). Briefly, magnetic beads were preincubated with anti-YB1 (Proteintech), anti-AKT (Proteintech), anti–p-YB1 (Abcam), anti-FLAG (Proteintech) or IgG for 30 minutes at room temperature and washed by RIP wash buffer. The coprecipitated RNAs were detected by qPCR.

For MS2-RIP assay, we first established a plasmid expressing HIF1A-AS1 tagged with MS2-binding site hairpin loops (HIF1A-AS1-MS2bs), and a plasmid expressing MS2–GST–NSL fusion protein (GeneChem). And a plasmid only expressing MS2 (MS2) RNA as control. After transfection for 48 hours, cell lysates were harvested then conducted pulldown assays with the anti-GST (Proteintech). Finally, the purified RNAs were detected by qPCR, and the proteins were detected by Western blot.

Total RNA was extracted from the cells using TRIzol (TaKaRa) and was amplified by qRT-PCR using primers containing a T7 promoter sequence-specific for HIF1A-AS1, and truncates or antisense of HIF1A-AS1, as well as HIF1α mRNA or antisense of HIF1α mRNA. Lysates of cells were prepared using standard lysis buffers in Pierce Magnetic RNA-Protein Pull-Down Kit (Thermo Fisher Scientific). Finally, the retrieved proteins were detected by Western blot. Primer sequences of HIF1A-AS1 and HIF1α mRNA for transcription in vitro are given in Supplementary Table S1.

Briefly, wild-type (WT) HRE3 sequence of HIF1A-AS1 promoter or mutant HRE3 sequence, and full-length 5′-UTR sequence of HIF1α were constructed into pGL3-based vectors and then transfected into GEM-R pancreatic cancer cells. The luciferase activities were measured using a luciferase assay kit (Promega) and fluorescence microplate reader. Firefly luciferase activities were normalized by Renilla luciferase activities.

Male BALB/c nude mice (ages 4 weeks; n = 5/group) were obtained from Vital River Laboratory Animal Technology (Beijing, China). Parental BxPC3 or BxPC3^GEM-R cells (2 × 10⁶ cells/mouse) stably transfected with lentivirus containing different plasmids in 100 μL RPMI-1640 were subcutaneously implanted in the right flank. Sample size of animal experiment was chosen according to similar conventions for well-designed experiments, and statistical methods were not used. Every 6 days, the volume of the tumor was assessed on the basis of: V (mm³) = 0.5 × L (length) × W² (width). Drug administration was adopted when the tumors reached about 100 mm³ in size, at which point, mice were randomized for treatment with DMSO (intraperitoneally), GEM (50 mg/kg/every 2 days, intraperitoneally) or 2-DG (2-Deoxy-D-glucose; 500 mg/kg/daily, intraperitoneally; Beyotime). Animals were sacrificed at 30 days after cell inoculation, and tumors were isolated and weighed. Then the tumor xenografts were embedded in paraffin, followed by staining with hematoxylin and eosin and IHC. No blinding approach was used during this study. The Animal Research Committee of the Academic Medical Center at Huazhong University of Science and Technology (Wuhan, China) approved all aspects of this study. Care and handling of the animals were following the guidelines for Institutional and Animal Care and Use Committees.

The clinical samples were obtained from Pancreatic Disease Institute of Union Hospital (Wuhan, China). 69 patients undergoing neoadjuvant chemotherapy with GEM followed by surgical treatment according to the National Comprehensive Cancer Network guidelines were selected. Among the 69 patients with pancreatic cancer mentioned above, 24 were sensitive to GEM treatment, whereas the other 45 were resistant. The tissues were obtained from surgical resection of patients or biopsy of the palliative surgery patients. Part of the excised tissue specimens was fixed in 10% buffered formalin solution and then embedded in paraffin, and part of the samples was immediately frozen at liquid nitrogen after surgical resection. Diagnosis of pancreatic cancer in these individuals was based on original histopathology findings by two pathologists. All samples were collected with the written informed consent of the patients. The study was approved by the local Research Ethics Committee at the Academic Medical Center of Huazhong University of Science and Technology. All procedures of the study were performed according to the Declaration of Helsinki.

All results were presented as means ± SD by GraphPad Prism 8.0 (GraphPad Software). Student t tests were used to analyze the difference between two groups. The one-way ANOVA test was used for multiple comparisons. Spearman's correlation analysis and the χ² test were used to analyze the correlation between different two genes. Log-rank test was used to assess survival difference. All statistical results were two-sided and P < 0.05 was considered to indicate statistical significance.

Cell viability assay, Colony formation assay, Western blot, RNA isolation, Reverse transcription, Quantitative real-time PCR (qRT-PCR), Transfection, Protein stability assay, RNA stability assay, RNA sequencing (RNA-seq) analysis, IHC, Coimmunoprecipitation (co-IP), and chromatin immunoprecipitation (ChIP) assays are given in Supplementary Methods.

To investigate the potential mechanism of chemoresistance, we generated GEM resistance pancreatic cancer cell line (BxPC3^GEM-R/PANC1^GEM-R), and the resistance status was determined (Supplementary Fig. S1A). Both BxPC3^GEM-R and PANC1^GEM-R cells showed increased resistance to GEM compared with parental cells (Supplementary Fig. S1B and S1C).

To identify lncRNAs that may affect the sensitivity of pancreatic cancer cells to chemotherapy, we compared the expression of lncRNAs in parental BxPC3 and BxPC3^GEM-R cells by RNA-seq (Fig. 1A). Meanwhile, the differential gene expression between BxPC3^GEM-R and parental BxPC3 cells was shown in the volcano plots (|log₂FoldChange| >2, P < 0.01; Fig. 1B). As natural antisense of HIF1α, HIF1A-AS1, which was remarkably upregulated in BxPC3^GEM-R cells, drew our additional attention (Fig. 1A and B). HIF1A-AS1 knockdown significantly impaired GEM-R cell survival under GEM-treated conditions (Fig. 1C and D; Supplementary Fig. S1D and S1E; Supplementary Table S2). To mimic the upregulation of HIF1A-AS1 (about 2-fold) between parental and GEM-R cells, we chose a dose of 1.0 μg/well of HIF1A-AS1 overexpression plasmid in 6-well plates for following experiments (Supplementary Fig. S1F and S1G). Conversely, overexpression of HIF1A-AS1 enhanced pancreatic cancer cell viability (Fig. 1E). Meanwhile, knockdown of HIF1A-AS1 had no obvious effect on the viability of parental pancreatic cancer cells (Supplementary Fig. S1H and S1I).

Coding-potential assessment tool, coding potential calculator, and Ribosome profiling data verified that HIF1A-AS1 had a very low protein-coding ability (Supplementary Fig. S2A and S2B). Meanwhile, we confirmed that HIF1A-AS1 had lower evolutionary conservation by using PhyloP analysis (Supplementary Fig. S2C). And HIF1A-AS1 signal was detected mainly in nucleus and partly in cytoplasm (Fig. 1F). Subsequently, combined immunofluorescence and FISH assay further demonstrated that HIF1A-AS1 was located mainly in nucleolus and partly in cytoplasm (Fig. 1G). In addition, HIF1A-AS1 was upregulated in pancreatic cancer tissue compared with adjacent noncancerous tissue (Fig. 1H). The secondary structure of HIF1A-AS1 was predicted by the online tool (Supplementary Fig. S2D). Taken together, these results implied that lncRNA HIF1A-AS1 was highly expressed in GEM-R pancreatic cancer cells and promoted GEM resistance of GEM-R pancreatic cancer cells.

Recent studies have gradually focused on the role of tumor metabolism, especially glycolysis, in promoting tumor drug resistance (19). Indeed, GEM-R pancreatic cancer cells showed increased glucose uptake, lactate production, and cell medium acidification (Fig. 2A). Nevertheless, after the treatment of 2-DG (inhibitor of glycolysis), GEM-R cells failed to resist the toxicity of GEM (Fig. 2B), which establishing that GEM-R pancreatic cancer cells have an increased dependence on glycolysis. Interestingly, HIF1A-AS1 knockdown significantly inhibited, whereas overexpression of HIF1A-AS1 promoted, the glycolysis level of GEM-R pancreatic cancer cells (Fig. 2C and D). Meanwhile, the cell viability of GEM-R pancreatic cancer cells induced by HIF1A-AS1 overexpression could be inhibited by 2-DG, BAY-876 (inhibitor for GLUT1), or oxamate (inhibitor for LDHA; Fig. 2E and F; Supplementary Fig. S2E and S2F). Taken together, these results implied that lncRNA HIF1A-AS1 facilitated GEM resistance of GEM-R pancreatic cancer cells in a glycolysis-dependent manner.

Antisense lncRNA can regulate the expression of the sense gene via various mechanisms (20). As a natural antisense RNA of HIF1α mRNA (Fig. 3A), HIF1A-AS1 had been reported to be involved in oxidative regulation of HIF1α in human-induced pluripotent stem cell–derived endothelial cells under oxidative stress (21). Thus, we speculated that HIF1α was a critical target of HIF1A-AS1 in GEM-R pancreatic cancer cells. Interestingly, we observed that GEM-R cells expressed a significantly higher level of HIF1α protein, even under normoxia conditions (Fig. 3B). Knockdown or overexpression of HIF1A-AS1 had no significant influence on HIF1α mRNA (Fig. 3C). The mRNA stability assays revealed that the half-life of HIF1α mRNA had no obvious alteration when depleting or overexpressing HIF1A-AS1 (Supplementary Fig. S3A). Nevertheless, HIF1A-AS1 knockdown significantly decreased, whereas overexpression of HIF1A-AS1 increased the expression of HIF1α protein (Fig. 3D). Therefore, these results implied that HIF1α was modulated by HIF1A-AS1 post-transcriptionally.

With treatment of MG132, the proteasome inhibitor, we found that knockdown or overexpression of HIF1A-AS1 altered the expression of HIF1α protein in GEM-R or parental cells (Fig. 3E; Supplementary Fig. S3B). In addition, E3 ubiquitin ligase von Hippel–Lindau (VHL), the key regulator of HIF1α protein, showed no significant difference between parental and GEM-R cells under normoxia or hypoxia (Fig. 3F; Supplementary Fig. S3C). Meanwhile, HIF1α protein stability assays displayed that the degradation rate of HIF1α remained unchanged in HIF1A-AS1 knockdown GEM-R cells and HIF1A-AS1–overexpressed parental cells (Fig. 3G; Supplementary Fig. S3D). More importantly, HIF1A-AS1 knockdown decreased, whereas overexpression of HIF1A-AS1 increased the level of polysomes-conjunct HIF1α mRNA in GEM-R pancreatic cancer cells (Fig. 3H; Supplementary Fig. S3E), indicating that HIF1A-AS1 increased the translation of HIF1α mRNA. And we also found the level of polysomes-conjunct HIF1α mRNA was increased in GEM-R cells than the parental cells (Fig. 3I). Moreover, targets of HIF1α, including key glycolytic proteins such as GLUT1, HK2, PKM2, and LDHA, were regulated by HIF1A-AS1, which further verified that HIF1α was a critical target of HIF1A-AS1 (Fig. 3J; Supplementary Fig. S3F). Furthermore, there was no significant difference in HIF1α mRNA level between GEM-R cells and parental cells (Supplementary Fig. S3G).

As expected, overexpression of HIF1α could rescue the downregulated chemotherapy resistance and glycolysis level caused by HIF1A-AS1 knockdown (Fig. 3K–M; Supplementary Fig. S4A), whereas knockdown of HIF1α could reverse the upregulated chemotherapy resistance and glycolysis level caused by overexpression of HIF1A-AS1 (Supplementary Fig. S4B–S4E). Together, these results revealed that HIF1α was the vital target of HIF1A-AS1 for exerting function in GEM-R pancreatic cancer cells, and post-transcriptionally regulated by HIF1A-AS1 in translation level but not through protein degradation.

We then focused on the potential mechanism of upregulated translation of HIF1α induced by HIF1A-AS1. Combined immunofluorescence and FISH analysis showed that the colocalization of HIF1A-AS1 and HIF1α mRNA was mainly in the nucleolus and partly in the cytoplasm (Fig. 4A). And the MS2-tagged RNA affinity purification analysis was performed to confirm the association between HIF1A-AS1 and HIF1α mRNA (Fig. 4B). The relative enrichment of HIF1A-AS1 in the pull-down beads of the HIF1A-AS1-MS2 group was measured by qPCR (Supplementary Fig. S5A). The HIF1α mRNA that interacted with HIF1A-AS1 was pulled down by the HIF1A-AS1-MS2 compounds (Fig. 4C). These results indicated that HIF1A-AS1 can directly interact with HIF1α mRNA.

Recently, lncRNAs have been reported to be involved in regulating variable signaling pathways by interacting with RNA binding proteins (22, 23). Specifically, it has been reported that Y-box binding protein 1 (YB1) can interact with HIF1α mRNA and activate its translation in high-risk sarcomas (24). Therefore, we hypothesized that HIF1A-AS1 enhanced the translation of HIF1α by interacting with YB1 in GEM-R pancreatic cancer cells. HIF1A-AS1 had a high protein–RNA interaction propensity to YB1 according to catRAPID (Fig. 4D; Supplementary Table S3; ref. 25). Similarly, our results also showed that YB1 knockdown inhibited HIF1α protein expression but had no influence on HIF1α mRNA (Supplementary Fig. S5B). Coincidently, YB1 was detected in the biotin-labeled HIF1A-AS1 pull-down compounds (Fig. 4E). YB1 could also be pulled down by biotin-labeled HIF1α mRNA (Fig. 4F). Furthermore, the deletion-mapping analyses indicated that YB1-binding site was located in the 1–216nt of HIF1A-AS1 (Fig. 4G). And extensive enrichments of HIF1α mRNA and HIF1A-AS1 were precipitated by the antibody against YB1, which verified that YB1 can interact with HIF1A-AS1 and HIF1α mRNA (Fig. 4H and I). HIF1α mRNA-MS2 compounds also pulled-down the YB1 and HIF1A-AS1 (Supplementary Fig. S5C), implying that interaction between HIF1A-AS1/HIF1α mRNA hybrids and YB1. By using truncated YB1-expressing plasmids, we ascertained that the cold shock domain (CSD) of YB1 was a necessary site that binds to HIF1α mRNA and HIF1A-AS1 (Fig. 4J; Supplementary Fig. S5D).

There was no significant difference in YB1 expression between GEM-R cells and parental cells (Supplementary Fig. S5E). Results from MS2-RIP assays showed that neither knockdown nor overexpression of YB1 can affect the binding between HIF1A-AS1 and HIF1α mRNA (Supplementary Fig. S5F and S5G). Nevertheless, the knockdown of HIF1A-AS1 dramatically decreased the interaction between YB1 and HIF1α mRNA, which was rescued by overexpression of YB1 (Fig. 4K). Coincidently, knockdown of YB1 reduced the HIF1A-AS1–induced interaction between YB1 and HIF1α mRNA (Fig. 4L). Dual-luciferase reporter assay was conducted by transfecting with a dual-luciferase reporter containing full-length HIF1α 5′-UTR sequence to validate the effect of HIF1A-AS1 on YB-1-driven translation of HIF1α. The results showed that the luciferase density decreased in BxPC3^GEM-R cells after the knockdown of HIF1A-AS1, whereas it was restored by overexpression of YB1 (Fig. 4M). Then, two truncated mutants (M1, 1–216nt; M2, 217–652nt) of HIF1A-AS1 were transfected into cells, the luciferase density between the M1 and WT HIF1A-AS1 groups showed no difference, but decreased luciferase density was detected in the M2 group. Meanwhile, no obvious alteration was observed in BxPC3^GEM-R cells transfected with the siYB1 (Fig. 4N). The results further indicated that 1–216nt of HIF1A-AS1 was critical for YB-1 to drive the translation of HIF1α.

Polysomal fractionation assays showed that overexpression of YB1 rescued the downregulation of polysomes-conjunct HIF1α mRNA induced by siHIF1A-AS1, and vice versa (Fig. 4O; Supplementary Fig. S5H). Consistently, the expression of YB1 altered the level of HIF1α protein regulated by HIF1A-AS1, instead of the mRNA level (Fig. 4P; Supplementary Fig. S5I and S5J).

We further explored the role of HIF1A-AS1/YB-1/HIF1α axis in the progression of pancreatic cancer chemoresistance. Knockdown of YB1 decreased, whereas overexpression of YB1 increased the effect of HIF1A-AS1/HIF1α axis on the GEM responsiveness, IC₅₀, and glycolysis level (Supplementary Fig. S6A-S6D). Taken together, our results intensively indicated that HIF1A-AS1 facilitated HIF1α translation by recruiting YB1 to HIF1α mRNA in GEM-R pancreatic cancer cells.

A recent study had demonstrated that the phosphorylation status of YB1 protein affected its ability to binding mRNAs (26). Therefore, we wondered whether HIF1A-AS1 enhanced the interaction between YB1 and HIF1α via regulating the phosphorylation status of YB1. Interestingly, that neither knockdown nor overexpression of HIF1A-AS1 affected the mRNA and protein level of YB1 under normoxia or hypoxia (Supplementary Fig. S7A-S7D). Nevertheless, phosphorylation of YB1 was significantly increased by overexpression of HIF1A-AS1 and remarkably decreased by HIF1A-AS1 knockdown (Fig. 5A and B). In addition, the RIP assay showed obvious enrichment of HIF1α mRNA by anti–phospho-YB1 (p-YB1) antibody, which was enhanced by overexpression of HIF1A-AS1 but impaired by HIF1A-AS1 knockdown (Fig. 5C and D). Meanwhile, HIF1α mRNA pulled down phospho-YB-1 (S102D), instead of dephospho-YB-1 (S102A; Supplementary Fig. S7E). These results implied that phosphorylation status was pivotal for YB1 to bind to HIF1α mRNA.

It has been shown that the CSD of YB1 could be phosphorylated by AKT serine/threonine kinase (AKT) at Ser102 and that disruption of this specific site could inhibit tumor cell growth (27). We pretreated BxPC3^GEM-R cells with AKT activator IGF1 or inhibitor API-2, and found that IGF1 remarkably enhanced, whereas API-2 substantially reduced the phosphorylation level of AKT and YB1 (Supplementary Fig. S7F). Therefore, we presumed that HIF1A-AS1 promoted phosphorylation of YB1 in an AKT-dependent manner. Coincidently, RNA pull-down assay showed that HIF1A-AS1 could interact with AKT (Fig. 5E). RIP assay using anti-AKT antibody also showed the accumulation of HIF1A-AS1 (Fig. 5F). Furthermore, the deletion-mapping analysis suggested that 1–126nt of HIF1A-AS1 was essential for the interaction between HIF1A-AS1 and AKT (Fig. 5G). We also constructed truncated AKT-expressing plasmids and the results revealed that the protein kinase B-like pleckstrin homology (PH) domain of AKT interacted with HIF1A-AS1 (Fig. 5H). Moreover, the co-IP assays showed that HIF1A-AS1 promoted the interaction between phospho-AKT (p-AKT) and p-YB1 protein, which was remarkably impeded by silencing of HIF1A-AS1 (Fig. 5I and J). These results indicated that HIF1A-AS1 regulated the phosphorylation of YB1 by enhancing the interaction between p-AKT and p-YB1. Meanwhile, API-2 decreased the p-YB1 induced by HIF1A-AS1 overexpression, whereas IGF1 recovered the p-YB1 that was inhibited by siHIF1A-AS1, suggesting that AKT activity was critical for the HIF1A-AS1–induced phosphorylation of YB1 (Fig. 5K). In addition, API-2 abrogated the induction of HIF1α protein regulated by HIF1A-AS1 (Fig. 5L), whereas IGF1 rescued the siHIF1A-AS1–inhibited HIF1α protein expression (Fig. 5M). Similar results were observed in RIP assays (Supplementary Fig. S7G and S7H). Taken together, it was suggested that AKT-dependent phosphorylation of YB1 was critical for the regulation of HIF1A-AS1 on HIF1α expression.

We then explored the potential molecule that regulated the expression of HIF1A-AS1 in GEM-R pancreatic cancer cells. We observed that GEM-R pancreatic cancer cells expressed a higher level of HIF1A-AS1 compared with parental cells (Fig. 6A). In addition, the Cancer Cell Line Encyclopedia (CCLE) database showed a positive correlation trend between HIF1A-AS1 and HIF1α mRNA in various cancer cells, including pancreatic cancer cells (Fig. 6B). Meanwhile, HIF1A-AS1 expression was increased gradually in GEM-R pancreatic cancer cells up to 24 hours after exposure to hypoxia condition or treated with CoCl₂ (Fig. 6C). Thus, we speculated that HIF1A-AS1 could be regulated by HIF1α. Indeed, HIF1A-AS1 showed a dose-dependent increase with HIF1α overexpression (Fig. 6D). Conversely, the depletion of HIF1α significantly eliminated HIF1A-AS1 expression (Fig. 6E). The sequence analysis showed that the HIF1A-AS1promoter contains 3 putative hypoxia-responsive elements (HRE; Fig. 6F). Meanwhile, ChIP assay showed obvious enrichment of HRE3 by anti–HIF1α antibody but not HRE1 or HRE2 (Fig. 6G). Then, dual-luciferase reporter plasmids containing WT or mutant HRE3 sequence (MUT) of promoter were designed and then transfected into BxPC3^GEM-R cells. The results showed that overexpression of HIF1α significantly increased luciferase intensity of the WT group, whereas no obvious change was observed in the MUT group (Fig. 6H). Nevertheless, the ChIP and luciferase assay further displayed that the silencing of HIF1α reduced the binding and transcription activity of HIF1α under hypoxia conditions (Fig. 6I and J). Therefore, these results intensively indicated that HIF1A-AS1 was transcriptionally regulated by hypoxia-induced HIF1α in GEM-R pancreatic cancer cells.

We performed xenograft mouse models to further reveal the effects of the HIF1A-AS1/HIF1α pathway on chemotherapy resistance of pancreatic cancer in vivo. Consistent with our previous in vitro results, overexpression of HIF1A-AS1 significantly abolished the inhibitory effects of GEM treatment in parental pancreatic cancer cells derived groups, as indicated with the increased tumor size and tumor weight (Fig. 7A–C). Nevertheless, treatment of 2-DG suppressed the GEM resistance induced by overexpression of HIF1α. Similarly, HIF1A-AS1 knockdown inhibited GEM resistance in GEM-R pancreatic cancer cells-derived xenografts, which could be reversed by overexpression of HIF1α. IHC assays of GEM-R pancreatic cancer cell–derived xenografts showed that expression of HIF1α could be downregulated by HIF1A-AS1 knockdown (Supplementary Fig. S8A). Furthermore, HIF1A-AS1/HIF1α axis was involved in regulation of proliferation, apoptosis, and glycolysis of GEM-R groups, as indicated by increased Ki67 (marker for cell proliferation), decreased Cleaved Caspase 3 (marker for apoptosis), and increased GLUT1 (marker for glycolysis), respectively.

To validate the relationship between the HIF1A-AS1/HIF1α pathway and GEM resistance progression of pancreatic cancer, we detected HIF1A-AS1 and HIF1α expressions in 69 pancreatic cancer tissues. HIF1A-AS1 levels were remarkably higher in tissues from 45 patients who were resistant for GEM treatment compared with 24 patients that were sensitive to GEM treatment (Fig. 7D). Meanwhile, higher HIF1α protein levels were observed in GEM-R pancreatic cancer group relative to the GEM-sensitive group (Fig. 7E). Moreover, HIF1A-AS1 and HIF1α expressions showed a positive correlation in patients with pancreatic cancer with GEM resistance (Fig. 7F; Supplementary Fig. S8B). GEM-R patients with pancreatic cancer with higher HIF1A-AS1 or HIF1α expression displayed a poorer overall survival (OS; Supplementary Fig. S8C and S8D). Further survival analysis showed that low level of both HIF1A-AS1 and HIF1α was much more beneficial to OS of GEM-R patients with pancreatic cancer, whereas high level of both HIF1A-AS1 and HIF1α was unfavorable for prognosis (Fig. 7G). FISH assay and IHC analysis displayed obvious upregulation of HIF1α in GEM-R pancreatic cancer tissue, which was further increased in those with high expression of HIF1A-AS1 (Fig. 7H and I). Together, these results further confirmed that the HIF1A-AS1/HIF1α signaling contributed to chemoresistance of pancreatic cancer (Fig. 7J).

Variable lncRNAs were involved in GEM resistance of pancreatic cancer. For example, lncRNA PVT1 promoted GEM resistance of pancreatic cancer via activating the Wnt/β-catenin and autophagy pathway (28). LncRNA SLC7A11-AS1 induced GEM resistance by blocking SCF^β-TRCP-mediated degradation of NRF2 (29). In addition, the role of glycolysis in promoting drug resistance of pancreatic cancer has been gradually revealed (19, 30). Dai and colleagues (31) displayed that glycolysis inhibitor, 2-DG, significantly increased the sensitivity of pancreatic cancer cells to GEM. Xi and colleagues (9) found that hENT1 reversed GEM resistance by inhibiting glycolysis and altering glucose transport mediated by HIF1α in pancreatic cancer. However, an association between lncRNAs and glycolysis-dependent GEM resistance of pancreatic cancer has not yet been well established. Specifically, our results showed that HIF1A-AS1 was significantly upregulated lncRNAs in GEM-R pancreatic cancer cells and prompted glycolysis-associated GEM resistance by regulation translation of sense gene HIF1α. Coincidently, we verified that depletion of HIF1A-AS1 significantly reversed the GEM resistance of GEM-R cells both in vitro and in vivo experiments. Thereby, our research provides new data for lncRNAs facilitating the interaction between glycolysis and chemoresistance in pancreatic cancer.

Research had demonstrated that HIF1α–dependent glycolysis promoted GEM sensitivity of pancreatic cancer cells. For example, lncRNA PVT1 promoted GEM resistance of pancreatic cancer via activating the Wnt/β-catenin and autophagy pathway (28). LncRNA SLC7A11-AS1 induced GEM resistance by blocking SCF^β-TRCP-mediated degradation of NRF2 (29). Coincidently, our study identified that HIF1α is a critical target for HIF1A-AS1 exerting effects on GEM resistance of pancreatic cancer, and we further explored the mechanism for the HIF1A-AS1 regulating HIF1α expression. Specifically, we discovered that HIF1A-AS1 upregulated HIF1α expression at protein level but not mRNA level. Meanwhile, HIF1A-AS1 did not affect the stability of HIF1α protein, but increased the polysome-conjunct HIF1α mRNA. These results implied that HIF1A-AS1 regulated HIF1α at transnational level in GEM-R pancreatic cancer cells. Differently, research showed that lncRNA DANCR stabilized HIF1α mRNA by interacting with NF90/NF45 complex (32), and lncRNA GEHT1 blocked VHL-mediated degradation of HIF1α and enhanced the protein level of HIF1α (33). Therefore, it is noteworthy that our present study implicates a novel mechanism of lncRNA-mediated regulation of HIF1α translation.

YB1 was a highly conserved protein that could bind to both DNA and RNA to play diverse regulatory roles in numerous cellular processes (34). Importantly, YB1 was found to enhance the translation of HIF1α mRNA and contributed to metastasis of high-risk human sarcomas (24). It aroused our interest in whether YB1 was involved in HIF1A-AS1–mediated HIF1α translation. Exactly, we validated HIF1A-AS1 recruited YB1 to HIF1α mRNA, by which promoting the translation process. Nevertheless, depletion of YB1 visibly decreased HIF1α protein level that was induced by overexpression of HIF1A-AS1. These results intensively suggested that YB1 was necessary for HIF1A-AS1–mediated HIF1α translation in GEM-R pancreatic cancer cells.

Previous studies have explicated a crucial role of YB1 phosphorylation at Ser102 in multiple malignant tumors (27, 35, 36). Meanwhile, results have revealed that the AKT-mediated YB-1 phosphorylation could activate the translation of mRNA (37). Interestingly, we identified that the HIF1A-AS1–induced phosphorylation of YB1 was achieved by enhancing the interaction between p-AKT and p-YB1. Moreover, the expression of p-YB1 and HIF1α protein induced by HIF1A-AS1 was downregulated or enhanced by AKT inhibitor or activator, respectively. These results further suggested that AKT-dependent phosphorylation of YB1 was crucial for HIF1A-AS1 to regulate HIF1α translation in the progression of GEM resistance of pancreatic cancer.

HIF1α was found to function as a key oncogene by forming feedback loops with various molecules. For example, hypoxia-induced feedback of HIF1α and lncRNA-CF129 contributed to pancreatic cancer progression (38). Coincidently, we revealed that HIF1α could bind to HRE of HIF1A-AS1 promoter and consequently activated HIF1A-AS1 transcription, then forming a feedback loop of the HIF1A-AS1/HIF1α pathway. Coincidently, high expression of both HIF1A-AS1 and HIF1α was associated with the poor prognosis in patients with pancreatic cancer who accepted GEM treatment.

Collectively, the present study implicates a reciprocal feedback of HIF1A-AS1 and HIF1α that promotes chemoresistance of pancreatic cancer via translational activation of HIF1α regulated by p-YB1. It is suggested that HIF1A-AS1 is a potentially valuable therapeutic target to enhance the treatment response in GEM-R pancreatic cancer.

No disclosures were reported.

F. Xu: Validation, visualization, methodology, writing–original draft, writing–review and editing. M. Huang: Validation, methodology, writing–original draft, writing–review and editing. Q. Chen: Validation, investigation, project administration. Y. Niu: Funding acquisition, validation, investigation, project administration. Y. Hu: Validation, writing–review and editing. P. Hu: Validation. D. Chen: Validation. C. He: Validation. K. Huang: Validation. Z. Zeng: Validation. J. Tang: Validation. F. Wang: Validation. Y. Zhao: Validation. C. Wang: Supervision. G. Zhao: Conceptualization, resources, data curation, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.

This research was supported by the Clinical Research Physician Program of Tongji Medical College, Huazhong University of Science and Technology and the National Natural Science Foundation of China (grant numbers: 81672406 and 81872030 to Gang Zhao; 81802377 to Yi Niu). The authors thank Jinghui Zhang for the excellent technical assistance. They are very grateful to the patients and mice for their contributions and sacrifices to this study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.