This study was reviewed and approved by the Ethics Review Committee at Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, China. A set of pancreatic cancer tissues and matched neighboring healthy pancreatic tissue was obtained from pancreatic cancer patients (n = 30) who underwent surgical treatment at Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, China. None of these patients have received any preoperative chemotherapy or radiotherapy before surgery. Every patient provided their informed consent before they participated in this study.

Human pancreatic cancer cell lines (BxPc-3, PANC-1, SW1990 and AsPC-1), and the normal pancreatic epithelial cell line HPDE6-C7 were obtained from the American Type Culture Collection (Manassas, VA, USA). Cells were cultured in RPMI-1640 medium (Wako, Osaka, Japan) supplemented with 10% fetal bovine serum (FBS, Gibco, Waltham, MA, USA) and maintained at 37°C with 5% CO2.

We generated stable erlotinib-resistant cell lines as previously reported (17). In brief, erlotinib-resistant pancreatic cancer cell line AsPC-1/Erlo was developed from the AsPC-1 cell line by exposing the cells to increasing concentrations of erlotinib (Selleck Chemicals, Houston, TX, USA) for 6 months. Cells were cultured in drug-free media for 4-5 days before experiments to prevent acute drug effects.

Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s instructions, and was reverse-transcribed into cDNA as a template for RNA detection using the PrimeScript RT Master Mix (Takara, Dalian, China). For RNase R treatment, total RNA was incubated for 30 min at 37°C with or without 3 U/mg RNase R (Geneseed, China). Preparation of nuclear and cytoplasmic RNA was performed using a Nuclear/Cytoplasmic Isolation kit (BioVision, San Francisco, USA). Quantitative PCR was run with SYBR Premix EX Taq II (Takara Biotechnology, Tokyo, Japan) on an Mx3000P qPCR system (Agilent Technologies, Santa Clara, CA, USA). For circRNA and mRNA, GAPDH was used as an internal control. The expression of miR-1227 was measured using the mirVana qRT-PCR miRNA Detection Kit (Ambion, Austin, TX, USA). In the miRNA RT-PCR reaction, U6 was used as an internal control. The primers used for real-time PCR were purchased from Ribobio (Guangzhou, China).

The siRNAs, miRNA mimics and miRNA inhibitors used in this study were synthesized by RiboBio (Guangzhou, China). The expression vector encoding E-cadherin was purchased from GeneSeed (Guangzhou, China). For the construction of circ_0013587 overexpression plasmids, the sequence of circ_0013587 was amplified and cloned into a pLCDH-ciR vector by GeneSeed (Guangzhou, China). Stable cells were selected using puromycin (Sigma, St. Louis, MO, USA). Transfection was performed on pancreatic cancer cells using Lipofectamine 3000 reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions.

Cell proliferation and viability was examined using Cell Counting Kit-8 assay (CCK-8, Dojindo, Kumamoto, Japan) as previously reported (18). Pancreatic cancer cells were transfected as indicated and treated with different concentrations of erlotinib for 48 h. Cell survival was determined by the CCK-8 assay.

Transwell cell invasion assays were conducted as previously described (19). Pancreatic cells were resuspended in serum-free RPMI-1640 medium and then seeded into the upper chamber (8 μm pore size; Millipore, Billerica, MA, USA). A 10% FBS-complete RPMI-1640 medium was added to the lower chamber. After 24 h, the invaded cells to the lower face of the filters were fixed, stained with 0.1% crystal violet solution (Sigma, St. Louis, MO, USA), and counted at ×200 magnification in 10 randomly chosen fields. Experiments were repeated three times.

Whole-cell protein extractions were prepared with a RIPA buffer (Beyotime, Beijing, China). The proteins were then electro-transferred onto a polyvinylidene difluoride membrane (Millipore, Bedford, MA, USA) and blocked with 5% non-fat milk in Tris-buffered saline. Western blot analysis was performed with commercially available antibodies, anti-E-cadherin (Cell Signaling, MA), anti-Vimentin (Cell Signaling, MA), anti-caspase-3 (Cell Signaling, MA), anti-cleaved caspase-3 (Cell Signaling, MA), anti-Twist (Santa Cruz Biotechnology, Santa Cruz, CA) and anti-β-actin (Santa Cruz Biotechnology, Santa Cruz, CA).

The luciferase reporter vectors containing the full length of E-cadherin 3′-untranslated region (3′-UTR) or circ_0013587 sequences were obtained from RiboBio (Guangzhou, China). The mutant luciferase reporter vectors were generated using a QuikChange site-directed mutagenesis kit (Stratagene, CA, USA). Then, the luciferase reporter plasmids containing wild-type (WT) circ_0013587 fragment, mutant (MUT) circ_0013587 fragment with a mutated miR-1227 binding site, WT E-cadherin 3′-UTR region, or MUT E-cadherin 3′-UTR region with a mutated miR-1227 binding site were co-transfected with miR-1227 mimic, miR-1227 inhibitor or the respective control, along with the Renilla luciferase plasmid pRL-CMV (Promega, WI, USA) using the Lipofectamine 3000 reagent (Invitrogen, Carlsbad, CA, USA). Following 48 h transfection, the Firefly and Renilla luciferase activities were measured using the Dual-Luciferase Reporter Assay System (Promega).

All procedures were approved by the Animal Research Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, China. The in vivo experiments were performed as previously reported (20). In brief, AsPC-1/Erlo cells stably overexpressing circ_0013587 or AsPC-1/Erlo control cells were subcutaneously injected into the right flank of BALB/c nude mice (HFK Bioscience, Beijing, China), respectively. At 1 week post-transplantation, Erlotinib (50 mg/kg) was given every three days through intraperitoneal injection. Tumor volume (V) was monitored by measuring the length (L) and width (W) and calculated with the formula V = (L × W²) × 0.5. After 30 days, the mice were sacrificed and the weight of the tumor was recorded.

Each experiment was performed in triplicate. The results were expressed as the mean ± standard deviation. Student’s t-tests and one-way ANOVA were performed for the comparisons using Prism 6.0 for Windows (GraphPad, San Diego, CA, USA). P < 0.05 was considered statistically significant.

Human pancreatic cancer cell line AsPC-1 harbors KRAS mutation, p53 mutation and wild-type EGFR, thus representing a malignant phenotype commonly observed in pancreatic cancers (17). To understand the mechanisms of acquired erlotinib resistance in pancreatic cancer cells, we selected erlotinib-resistant AsPC-1/Erlo cells by culturing pancreatic cancer cell line AsPC-1 in increasing concentrations of erlotinib. The sensitivity to erlotinib was examined in each cell line using CCK-8 assays. As expected, the AsPC-1/Erlo cells were more resistant than the parental AsPC-1 cells ( Figure 1A ). Our qRT-PCR assay revealed a significant decrease in circ_0013587 expression in AsPC-1/Erlo cells than in AsPC-1 cells ( Figure 1B ). When we compared the expression of circ_0013587 in pancreatic cancer tissues and adjacent normal tissues, we found that the expression of circ_0013587 was significantly lower in pancreatic cancer tissues compared to their counterpart surrounding tissues ( Figure 1C ). Moreover, circ_0013587 levels in pancreatic cancer cell lines were also decreased compared with that in the normal pancreatic epithelial cell line HPDE6-C7 ( Figure 1D ). Notably, circ_0013587 was expressed more lowly in stage III/IV tissues than in stage I/II samples ( Figure 1E ). Those patients with the high-grade disease and lymph node metastasis had significantly lower circ_0013587 expression ( Figures 1F, G ). The prognostic significance of circ_0013587 expression was analyzed in 30 pancreatic cancer patients with the median as the cutoff value. According to the Kaplan-Meier survival curves, the low circ_0013587 group had shorter overall survival than the high circ_0013587 group ( Figure 1H ). Our results demonstrated that reduced circ_0013587 expression may correlate with the acquired erlotinib resistance in pancreatic cancer cells.

Both loss- and gain-of-function assays were performed to evaluate the effects of circ_0013587 expression on the biological behaviors of pancreatic cancer cells. We knocked down circ_0013587 expression in AsPC-1 cells using siRNAs and also overexpressed circ_0013587 in AsPC-1/Erlo cells using a circ_0013587 overexpression plasmid. Our qRT-PCR results showed that depletion of circ_0013587 with siRNAs resulted in a clear decrease in its levels, and transfection with a circ_0013587 expression plasmid resulted in a robust increase in circ_0013587 levels ( Figures 2A, B ). Cell proliferation and invasion was significantly enhanced following circ_0013587 knockdown, but was significantly attenuated following circ_0013587 overexpression ( Figures 2C–E ). Western blot analysis revealed that knockdown of circ_0013587 in AsPC-1 cells down-regulated E-cadherin protein levels and up-regulated Vimentin protein levels ( Figure 2F ). Conversely, AsPC-1/Erlo cells stably expressing circ_0013587 exhibited higher E-cadherin protein levels and lower Vimentin protein levels compared to control cells ( Figure 2F ). Collectively, these findings suggested that circ_0013587 suppresses proliferation and EMT in pancreatic cancer cells.

To evaluate the role of circ_0013587 in acquired resistance to erlotinib, we used the CCK-8 assay to examine the sensitivity of pancreatic cancer to erlotinib after overexpression or knockdown of circ_0013587. The results suggested that circ_0013587 overexpression significantly restored erlotinib sensitivity in AsPC-1/Erlo cells ( Figure 3A ). Following erlotinib treatment, AsPC-1 cells transfected with circ_0013587 siRNA displayed higher viability than those transfected with the siRNA control ( Figure 3B ). Consistently, overexpression of circ_0013587 in AsPC-1/Erlo cells increased the expression level of an apoptosis-related molecule, cleaved caspase-3 ( Figure 3C ). After transfection with circ_0013587 siRNA, cleavage of caspase-3 was inhibited in AsPC-1 cells exposed to erlotinib ( Figure 3D ). These results indicated that up-regulation of circ_0013587 reverses erlotinib resistance in pancreatic cancer cells.

Circ_0013587 was predominantly located in the cytoplasm of AsPC-1/Erlo cells ( Figure 4A ). Using the CircInteractome database (https://circinteractome.nia.nih.gov/), we have bioinformatically predicted miR-1227 as a target of circ_0013587 ( Figure 4A ). Therefore, we further studied the relationship between circ_0013587 and miR-1227 using luciferase reporter assays. As shown in Figure 4B , transfection with miR-1227 mimic significantly suppressed the luciferase activity of wild type sequence of circ_0013587, and transfection of miR-1227 inhibitor significantly elevated the luciferase activity of wild type sequence of circ_0013587 ( Figure 4B ). Meanwhile, these effects of miR-1227 mimic or inhibitor were significantly abolished in mutant circ_0013587 plasmids with the mutant sequence in the binding site of miR-1227 ( Figure 4B ). These results supported a direct interaction between miR-1227 and circ_0013587. Furthermore, qRT-PCR results demonstrated that overexpression of circ_0013587 decreased, while knockdown of circ_0013587 increased miR-1227 levels in pancreatic cancer cells ( Figure 4C ). Using qRT-PCR assays, we found that the expression of miR-1227 was significantly higher in AsPC-1/Erlo cells than AsPC-1 cells ( Figure 4D ). We confirmed the up-regulation of miR-1227 in pancreatic cancer samples compared to normal tissues ( Figure 4E ). The levels of miR-1227 in pancreatic cancer cell lines were clearly higher than that in normal cells ( Figure 4F ). Using the Kaplan-Meier Plotter database (http://kmplot.com/analysis/), our survival analysis suggested that higher miR-1227 expression was correlated with a worse outcome in pancreatic cancer patients ( Figure 4G ). These findings suggest that circ_0013587 directly sponges miR-1227 in pancreatic cancer cells.

CCK-8 assays showed that overexpression of circ_0013587 sensitized resistant AsPC-1/Erlo cells to erlotinib, while this effect was reversed by restoration of miR-1227 expression ( Figure 5A ). In line with this result, suppression of circ_0013587 contributed to erlotinib resistance in AsPC-1 cells, and transfection with miR-1227 inhibitor was able to overcome erlotinib resistance induced by circ_0013587 silencing ( Figure 5B ). Then, we examined whether the effects of circ_0013587 on cell invasion are mediated by miR-1227 in pancreatic cancer cells. Cell invasion assays revealed that overexpression of circ_0013587 suppressed cell invasion, but cell invasion ability was partially rescued by miR-1227 overexpression ( Figure 5C ). Down-regulation of circ_0013587 increased cell invasion, and this promotion was reversed by a miR-1227 inhibitor ( Figure 5D ). These results demonstrated that up-regulation of circ_0013587 reverses acquired resistance to erlotinib and suppresses cell invasion in pancreatic cancer cells through mediating miR-1227 expression.

We queried the TargetScan database (http://www.targetscan.org/vert_72/) to identify miR-1227 target genes and discovered a highly-conserved miR-1227 binding site on the 3′-UTR of E-cadherin ( Figure 6A , upper). The western blot assay revealed a reduction in E-cadherin protein expression in AsPC-1 cells harboring miR-1227 overexpression ( Figure 6A , bottom). Transfection with miR-1227 inhibitor increased E-cadherin levels in AsPC-1/Erlo cells ( Figure 6A , bottom). The relative luciferase activity of wild-type E-cadherin 3′-UTR was significantly down-regulated in AsPC-1 cells transfected with miR-1227 mimic than in cells transfected with control mimic ( Figure 6B ). There was a significantly higher relative luciferase activity of wild-type E-cadherin 3′-UTR in AsPC-1/Erlo cells transfected with miR-1227 inhibitor than in cells transfected with control inhibitor ( Figure 6B ). However, the introduction of miR-1227 mimic or inhibitor did not significantly impact the relative luciferase activity of mutant E-cadherin 3′-UTR ( Figure 6B ). We also verified the down-regulation of E-cadherin in AsPC-1/Erlo cells than AsPC-1 cells ( Figure 6C ). The levels of E-cadherin were significantly lower in pancreatic cancer tissues compared with normal tissues ( Figure 6D ). Pancreatic cancer cells expressed lower expression of E-cadherin than normal cells ( Figure 6E ). The findings suggested that miR-1227 can bind to and suppress the expression of E-cadherin.

To access whether miR-1227 regulates erlotinib resistance and cell invasion by modulating E-cadherin expression, we examined the protein levels of E-cadherin, Twist, and Vimentin in AsPC-1 cells co-transfected with miR-1227 mimic (or control mimic) along with E-cadherin expression vector (or control vector), and in AsPC-1/Erlo cells co-transfected with miR-1227 inhibitor (or control inhibitor) along with E-cadherin siRNA (or control siRNA). Our results suggested that miR-1227 mimic induced the expression of Twist and Vimentin, and decreased E-cadherin expression ( Figure 7A ). These impacts of miR-1227 overexpression were largely abolished by ectopic overexpression of E-cadherin in AsPC-1 cells ( Figure 7A ). In addition, inhibition of miR-1227 markedly increased E-cadherin levels and reduced the protein expression of Twist and Vimentin ( Figure 7B ). Co-transfection with E-cadherin siRNA reversed these changes in AsPC-1/Erlo cells ( Figure 7B ). Our CCK-8 and cell invasion assays demonstrated that miR-1227 inhibitor-induced suppression of erlotinib resistance and cell invasion was promoted by E-cadherin knockdown ( Figures 7C, D ). Also, miR-1227 overexpression facilitated erlotinib resistance and cell invasion in AsPC-1 cells, whereas forced overexpression of E-cadherin could significantly suppress erlotinib resistance and cell invasion induced by miR-1227 mimic ( Figures 7E, F ). These findings demonstrated that miR-1227 mediates the acquired resistance to erlotinib and invasive ability of pancreatic cancer cells via E-cadherin.

Nude mouse xenograft models were established using AsPC-1/Erlo cells stably expressing circ_0013587 or AsPC-1/Erlo control cells. Both groups of nude mice received an intraperitoneal injection of erlotinib. As expected, erlotinib alone failed to suppress tumor growth in vivo ( Figures 8A, B ). However, the tumor volume and weight of the circ_0013587-expressing AsPC-1/Erlo cell group was significantly lower than the control group ( Figures 8A, B ). These results suggested that overexpression of circ_0013587 reversed erlotinib resistance in pancreatic cancer in vivo. Collectively, these results indicated that circ_0013587 sponges miR-1227 to enhance E-cadherin level, in turn attenuating the acquired resistance to erlotinib and invasive ability of pancreatic cancer cells ( Figure 8C ).