Human immortalized bronchial epithelial cell line (BEAS-2B cell line) was purchased from the Type Culture Collection of the Chinese Academy of Science (Shanghai, China) and cultured using RPMI-1640 supplemented with 10% foetal bovine serum (both Gibco; Thermo Fisher Scientific, Inc.) and 1% penicillin-streptomycin (Beijing Solarbio Science & Technology Co., Ltd.) at 37˚C with 5% CO₂. To establish an in vitro model of asthma epithelial inflammation, cells were treated with 0, 1, 5 and 10 µg/ml lipopolysaccharide (LPS, Sigma-Aldrich; Merck KGaA) for 24 h at 37˚C when cells had reached a confluence of 70-80%. BEAS-2B cells were used as controls following treatment with an equal volume of PBS for 24 h at 37˚C.

Cell activity was measured using CCK-8 assay (Dojindo Laboratories, Inc.) according to the manufacturer's instructions. BEAS-2B cells were resuspended in PBS and the density was adjusted to 2x10⁴ cells/ml before plating onto 96-well plates at a density of 5,000 cells/well. After 48 h at 37˚C, 10 µl CCK-8 reagent was added to each well. After 2 h, absorbance values at 450 nm were measured with a microplate reader (BioTek Instruments, Inc.).

BEAS-2B cells were plated into 6-well plates (5x10⁵ cells/well) and cultured overnight to 60-70% confluency before transfection. All RNA sequences used for cell transfection were obtained from Guangzhou RiboBio Co., Ltd. Transfection complexes containing 50 nM RNAs were transfected into BEAS-2B cells using Lipofectamine 3000^® (Invitrogen; Thermo Fisher Scientific, Inc.) and serum-free Opti-MEM (Thermo Fisher Scientific, Inc.) and incubated for 6 h at 37˚C according to the manufacturer's protocols. Reverse transcription-quantitative (RT-q)PCR or western blot analysis was performed 48 h after transfection to assess transfection efficiency. The sequences of small interfering (si)RNAs and miRNA mimics and inhibitors are listed in Table I.

BEAS-2B cells were treated with LPS at a concentration of 5 µg/ml whereas PBS was used in the control group for 24 h at 37˚C. Total RNA was isolated with RNAiso PLUS (Takara) reagent and mRNA was enriched by Ribo-off rRNA Depletion kit (cat. no. N406; Vazyme Biotech Co., Ltd.). 1% agarose gel electrophoresis and NanoDrop ND-1000 spectrophotometry (Thermo Fisher Scientific, Inc.) were employed to verify quality and integrity of the extracted RNAs. RNA-seq libraries were prepared using the VAHTS^® Stranded mRNA-seq V2 Library Prep kit for Illumina^® (cat. no. NR612; Vazyme Biotech Co., Ltd.) following the manufacturer's recommendations. Briefly, RNA samples were fragmented and RT of first- and second-strand cDNAs was performed using with random hexamer primers; cDNAs were purified and RNA Adapters were ligated at the 5' and 3' ends of DNA fragments. The quality of the established libraries were examined using 8% agarose gel electrophoresis, then quantified by Qubit 2.0 Fluorometer (Invitrogen). The final strand-specific cDNA library (420 µl of a 10 pM library) was subjected to Illumina^® Hiseq X Ten (Illumina, Inc.) platform for 150 bp pair-end sequencing using HiSeq X Ten Reagent Kit (cat. no. FC-501-2501). Sequencing reads from RNA-seq were mapped to the human reference genome (hg19) from UCSC genome database (genome.ucsc.edu/). circRNA identification was performed using CIRI2 software (version 2.0.6; sourceforge.net/projects/ciri/) based on the RNA-seq data.

Total RNA was extracted from BEAS-2B cells placed in a 6-well plate using RNAiso PLUS reagent (Takara, Dalian, China). The cRNA first strand was synthesized by RT using PrimeScript™ RT kit (cat. no. RR037A; Takara Biotechnology Co., Ltd.) under the following reaction conditions: 37˚C for 15 min and 85˚C for 5 sec. qPCR was performed on an MX3000P qPCR system (Agilent Technologies, Inc.) using the ChamQ Universal SYBR qPCR^® Master Mix kit (cat. no. Q711; Vazyme Biotech Co., Ltd.), followed by 40 cycles at 95˚C for 15 sec and 64˚C for 30 sec. U6 and GAPDH were used as internal reference genes to normalize the data. Relative RNA expression levels were calculated using the 2^-ΔΔCq method (26). The primers were synthesized by Shanghai Bioengineering Co., Ltd. The forward and reverse sequences of the primers used for qPCR are shown in Table II.

PARIS kit (Ambion Life Technologies) was used to isolate nuclear and cytoplasmic RNA from BEAS-2B cells following the manufacturer's instructions. The subcellular localization of circDHTKD1 was analysed using RT-qPCR as aforementioned. GAPDH and NEAT1 were used as cytoplasmic and cytosolic controls, respectively.

PCR products were separated by electrophoresis on a 2% agarose gel and gels were cut and sent to Shanghai Sangon Bioengineering Co., Ltd. for sequencing following observation in a UV imaging system.

circRNA fragments containing the full length of circDHTKD1 with either the wild-type (WT) or mutated (MUT) binding site for miR-338-3p were synthesized and inserted into the psiCHECK2 vector (Promega Corporation) to generate psiCHECK2-circ-WT and psiCHECK2-circ-MUT, respectively. BEAS-2B were co-transfected with the aforementioned luciferase reporter plasmid and miR-338-3p mimics/miR-338-3p mimics negative control using Lipofectamine^® 3000 (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's instruction. Following 48 h transfection, the cells were lysed immediately and the relative luciferase activity of firefly and Renilla luciferase was determined using a Dual-Luciferase Reporter assay kit (cat. no. FR201; TransGen Biotech Co., Ltd.) following the manufacturer's protocol.

Total protein was isolated from BEAS-2B cells using RIPA Lysis Buffer (cat. no. 20-188; Merck KGaA), then quantified with BCA Protein Assay (cat. no. P0012; Beyotime Institute of Biotechnology) and the protein was denatured in a 95˚C water bath for 5 min. Proteins (30 µg/lane) were separated by 10% SDS-PAGE, transferred onto PVDF membranes (Millipore; Merck KGaA), blocked with 5% non-fat milk for 2 h at room temperature and incubated overnight at 4˚C with the following primary antibodies: Anti-ETS1 (cat. no. ab220361; Abcam; 1/1,000); anti-phosphorylated (p)-ERK1/2 (cat. no ab201015; Abcam; 1/1,000); anti-ERK1/2 (cat. no. ab184699; Abcam; 1/8,000) and anti-GAPDH (cat. no. ab181602; Abcam; 1/10,000). This was followed by incubation with rabbit anti-mouse horseradish peroxidase-conjugated secondary antibodies (cat. no. ab6721; Abcam; 1/5,000) for 60 min at room temperature. Protein bands were scanned and imaged using Immobilon^® Western HRP Substrate ECL chemiluminescence reagent (cat. no. WBKLS0500; Merck KGaA). Densitometry analysis was performed using ImageJ software (Version 1.53e; National Institutes of Health) with GAPDH as the loading control.

BEAS-2B cells were seeded into 12-well plates (1x10^5 cells/well) and treated with or without PBS for 48 h at 37˚C. Subsequently, the cell supernatant was collected and centrifuged at 1,000 x g for 5 min at 37˚C. According to the manufacturer's instruction, the concentrations of inflammatory cytokines IL-6 (cat. no. PI330), IL-8 (cat. no. PI640), TNF-α (cat. no. PT518) and VEGF (cat. no. PV963) were detected using ELISA kits (Beyotime Institute of Biotechnology).

Cancer-Specific CircRNAs Database (CSCD) (gb.whu.edu.cn/CSCD/) was applied to annotate and identify exonic circRNAs. The circbank (circbank.cn/) and circRNA interactome (circinteractome.nia.nih.gov/) databases were used to predict downstream miRNAs. In addition, the binding site of miR-338-3p ETS1 and ERK2 was predicted using the TargetScan database (targetscan.org/). The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of miR-338-3p target genes was conducted using the Metascape online tool (version 3.5; metascape.org/) (27).

The experimental data from three independent replicates are expressed as the mean ± standard deviation. Data were analysed using SPSS 22.0 (IBM Corp.) and GraphPad Prism 8.0 software (GraphPad Software, Inc.; Dotmatics). Paired Student's t-test was used to assess the difference between two groups, while one-way ANOVA was used to analyse differences between ≥3 groups followed by Tukey's post hoc test for pairwise comparisons. P<0.05 was considered to indicate a statistically significant difference.

CCK-8 assay was performed to evaluate the effect of different concentrations of LPS (0.1, 1.0, 10.0 and 100.0 µg/ml) on the viability of BEAS-2B cells, which were shown to be concentration-dependent. LPS concentration ≥5 µg/ml significantly inhibited viability of BEAS-2B cells after 24 h treatment (Fig. 1A). Furthermore, RT-qPCR and ELISA confirmed that 5 µg/ml LPS increased secretion of pro-inflammatory cytokines (IL-6, IL-8, TNF-α and VEGF) in BEAS-2B cells (Fig. 1B and C).

The transcriptome of LPS-treated BEAS-2B cells was sequenced using CIRI2 software for circRNA expression profiling, followed by selection of exonic circRNAs using CSCD (gb.whu.edu.cn/CSCD/). Finally, the top eight upregulated exonic circRNAs identified in BEAS-2B cells between LPS-induced and normal-control were annotated in the circBase database (circbase.org/; Table III). Changes in expression profile were analysed by RT-qPCR and showed a significant upregulation of circDHTKD1 (hsa_circ_0017724) in LPS-induced BEAS-2B cells compared with that in the control group. The fold change was increased compared with that of other circRNAs (Fig. 2A). Based on University of California, Santa Cruz, genome browser (genome.ucsc.edu/) and circBase databases, circDHTKD1 was shown to be circularised from exons 2 and 3 of the DHTKD1 gene on chromosome 10p14 with a splicing length of 368 bp (Fig. 2B). The back-splice junction of circDHTKD1 was confirmed by Sanger sequencing. The RT-PCR product of circDHTKD1 had head-to-tail splicing of the expected size (137 bp fragment) (Fig. 2C), indicating that the product was a circRNA. In addition, the subcellular localisation of circDHTKD1 in BEAS-2B cells was identified by nucleocytoplasmic separation and circDHTKD1 was found to be significantly enriched in the cytoplasm of BEAS-2B cells (Fig. 2D). These data demonstrated that circDHTKD1 was upregulated in LPS-induced BEAS-2B cells and was present in cytoplasm in a circular form.

To determine the role of circDHTKD1 in the LPS-induced inflammatory response of BEAS-2B cells, two pairs of siRNAs targeting circDHTKD1 (si-circ#1 and si-circ#2) were designed, both of which effectively interfered with the expression of circDHTKD1. The knockdown of circDHTKD1 caused by si-circ#1 was notably more efficient compared with that caused by si-circ#2 (Fig. 3A). The qPCR results showed that the circDHTKD1 knockdown partially decreased mRNA levels of IL-6 and VEGF but had no significant effect on the mRNA levels of IL-8 and TNF-α (Fig. 3B-E). ELISA confirmed the results of the circDHTKD1 knockdown (Fig. 3F). These results indicated that circDHTKD1 was involved in the regulation of LPS-induced inflammatory responses in BEAS-2B cells.

As circRNAs act as miRNA sponges and circDHTKD1 is primarily localised to the cytoplasm, it was hypothesised that circDHTKD1 may be involved in regulation of LPS-induced BEAS-2B cell inflammatory injury response via a competing endogenous RNA (ceRNA) mechanism. The target miRNAs that bind to circDHTKD1 were predicted through two databases, namely circBank (circbank.cn/) and circRNA interactome (circinteractome.nia.nih.gov/). miR-338-5p and miR-338-3p were shared miRNAs in both databases (Fig. 4A). RT-qPCR results showed that the expression of miR-338-3p significantly decreased following 24 h LPS induction, whereas expression of miR-338-5p did not change significantly (Fig. 4B). The dual-luciferase reporter assay showed that overexpression of miR-338-3p significantly reduced the luciferase activity of plasmid psiCHECK2-circ-WT but failed to decrease luciferase activity of psiCHECK2-circ-MUT (Fig. 4C and D). In addition, expression of miR-338-3p was increased after circDHTKD knockdown, whereas no significant changes in circDHTKD1 expression were detected after miR-338-3p overexpression (Fig. 4E and F). The results suggested that circDHTKD1 interfered with the function of miR-338-3p by acting as a sponge for miR-338-3p.

TargetScan (targetscan.org/) was used to predict the target genes of miR-338-3p while Metascape (metascape.org/) was used for cluster analysis of target genes of miR-338. The results suggested that MAPK signaling pathway was significantly enriched (Fig. 5A). Further searches using Kyoto Encyclopedia of Genes and Genomes database (kegg.jp/) revealed that ERK pathway genes (ERK2 and ETS1) may be regulated by miR-338-3p (Fig. 5B). The present results showed that transfection of miR-338-3p mimics in BEAS-2B cells decreased ETS1 expression levels, while inhibition of miR-338-3p increase ETS1 expression levels (Fig. 5C and D). In addition, western blotting showed that ETS1 protein levels significantly decreased following circRNA knockdown in BEAS-2B cells (Fig. 5E).

To explore the role of ETS1 in LPS-induced BEAS-2B cells, three siRNA pairs targeting ETS1 were designed (si-ETS1#1, si-ETS1#2 and si-ETS1#3) and transiently transfected into BEAS-2B cells. Following 48 h transfection, the mRNA and protein levels of ETS1 were significantly decreased in all transfection groups, with the si-ETS1#2 transfection showing the highest efficiency in knocking down ETS1 (Fig. 6A and B). In addition, western blotting showed that phosphorylation of ERK1/2 in BEAS-2B cells increased after LPS stimulation and knockdown of ETS1 significantly inhibited both the phosphorylation of ERK1/2 induced by LPS and activation of the ERK signalling pathway (Fig. 6C and D). Also, knockdown of ETS1 decreased the expression levels of IL-6 and VEGF (Fig. 6E and F). The present results indicated that ETS1 was involved in the regulation of inflammatory factors IL-6 and VEGF, as well as activation of the ERK pathway.

The functional recovery assay showed that LPS treatment resulted in increased expression of p-ERK compared with that in the negative control and that p-ERK levels were significantly decreased when circDHTKD1 was knocked down. This effect was reversed by transfection with miR-338-3p inhibitor (Fig. 7A and B). In addition, transfection with miR-338-3p inhibitor reversed the decrease in IL-6 and VEGF protein levels that was induced by the circDHTKD1 knockdown (Fig. 7C and D). These results collectively demonstrated that circDHTKD1 enhanced LPS-induced inflammatory responses in BEAS-2B cells via the miR-338-3p/ETS1 axis.