Human ESCC cell lines TE8 (RBRC-RCB2098) and TE11 (RBRC-2100) (RIKEN BioResource Center, Ibaraki, Japan) were cultured in RPMI-1640 medium supplemented with 10% fetal calf serum (Thermo Fisher Scientific, Waltham, MA, USA) and 100 U/ml penicillin and 100 µg/ml streptomycin (Nacalai Tesque, Kyoto, Japan) in a humidified atmosphere containing 5% CO₂ at 37°C. To generate cisplatin-resistant cell lines, TE8 and TE11 cells were cultured with increasing concentrations of cisplatin for 24 weeks. The established cisplatin-resistant cell lines, TE8R and TE11R, were maintained under a constant cisplatin concentration of 6 µM. TE11 cells were also cultured with increasing concentrations of 5-fluorouracil and docetaxel for 24 weeks. TE11 5-FU was maintained at a 5-fluorouracil concentration of 4.0 µM and TE11 DTX was maintained at a docetaxel concentration of 3.0 ng/ml.

In total, 82 samples of ESCC tissues were collected from patients who were histologically diagnosed with primary thoracic ESCC at Osaka University Hospital (Osaka, Japan) between 2017 April and 2020 September. The 47 samples were obtained via biopsy before treatment, and the 35 samples were obtained by surgical resection after neoadjuvant chemotherapy (NAC). All samples were collected after obtaining written informed consent, and this study was approved by the ethics committee of Osaka University (approval no. 16305-4). Clinicopathologic findings were classified according to the UICC-TNM Classification, seventh edition (21). At our institution, patients with ESCC primarily receive a DCF regimen as NAC, which consisted of docetaxel 70 mg/m² and cisplatin 70 mg/m² on day 1 and 5-FU 700 mg/m²/day on days 1–5, every 3 weeks (22).

Total RNA was extracted from ESCC cells or tissues using TRIzol Reagent (Invitrogen, CA, USA) according to the manufacturer's instructions. The quantity and quality of RNA samples were measured using NanoDrop ND-1000 (Thermo Scientific, DE, USA). The samples were reverse transcribed into cDNA using a reverse transcription system (Promega, Madison, WI) according to the manufacturer's protocol. Reverse transcription quantitative real-time polymerase chain reaction (RT-qPCR) was performed on a ViiA™ 7 real-time PCR system (Applied Biosystems, Foster City, CA, USA) using SYBR Green (TOYOBO, Osaka, Japan) via the following three-step protocol: 95°C for 2 min, (95°C for 15 sec, 65°C for 15 sec, and 72°C for 45 sec) ×40 cycles, melting curve analysis from 60 to 95°C with reads every 0.5°C. PCR was performed in triplicate, with the expression values normalized to the mRNA expression of β-actin calculated using the 2^−ΔΔCt method (23). The genes and primers are specified in Table SI.

Total RNAs were treated using the Epicenter Ribo-Minus rRNA Removal Kit (Illumina, CA, USA) and RNAse R (Epicenter, CA, USA) to remove ribosomal and linear RNA (6). Thereafter, RNA-seq libraries were constructed using TruSeq Stranded mRNA Library Prep (Illumina, CA, USA) following the manufacturer's protocol while skipping the process of purifying the polyA-containing mRNA molecules with oligo-dT attached magnetic beads. The BioAnalyzer 2100 system (Agilent Technologies, Santa Clara, CA, USA) was used to confirm the quality and quantity of the libraries. The RNA-seq library was sequenced using an Illumina HiSeq 3000 instrument (Illumina, San Diego, CA), and 150 bp paired-end reads were generated. Sequencing reads were mapped independently using CIRCexplorer2 (13,14), which is one of the circular RNA detection output tools based on Python, a general-purpose programming language. Cirexplorer uses TopHat (15) and TopHat-Fusion (16) to detect backsplicing junction reads. To compare differences in circRNA expression profiles between TE11 and TE11R, we calculated the fold change between the groups for each circRNA. A t-test was used to estimate the statistical significance of the differences. A significant difference in the expression of circRNA was defined by a fold-change >3.0 or <0.5 with a P-value of <0.05.

mRNAs involved in the pathways associated with conferring cisplatin resistance to tumor cells were selected for CircRNA-microRNA-mRNA network construction. The binding sites in the 3′-UTR between differentially expressed miRNAs and the selected mRNA were determined using Targetscan (http://www.targetscan.org/) to predict the mRNAs targeted by miRNAs filtered using a cumulative weighted context score of ≤0.8. We used Cytoscape software to visualize the data (24) and the bioinformatic database CircInteractome to predict the sequence of miRNA-binding sites of circ_0004365 (25).

Small interfering RNA (siRNA) of circ_0004365 was synthesized by Sigma-Aldrich (St. Louis, MO) targeting to sequence of the fusion region of circ_0004365. TE11 and TE11R cells were transfected with circ_0004365 siRNAs using the Lipofectamine^® RNAiMAX transfection reagent (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's protocol. Briefly, cells were seeded at 2 ×10⁵ cells/well in 6-well plates added to a final concentration of 50 nM using Lipofectamine^® RNAiMAX (Thermo Fisher Scientific, Waltham, MA, USA) and cultured for 24 h at 37°C in 5% CO₂. The sequence of circ_0004365 siRNA was sense CUGUUUCUCGGAACAGGACdT dT and antisense GUCCUGUUCCGAGAAACAGdTdT.

A total of 4.0×10³ cells per well were seeded in 96-well plates. Cell viability was then assessed using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT) assay 48 h after the addition of cisplatin. Absorbance of the converted dye was measured at a wavelength of 550 nm with background subtraction at 665 nm using an iMark™ Microplate Absorbance Reader (BIO-RAD, Tokyo, Japan). The concentration at which the drug produced 50% growth inhibition (IC50) was estimated using the relative survival curve.

Histopathological tumor response was evaluated according to the histological criteria of the Japanese Esophageal Society (26) and was classified into five categories according to the proportion of tumor degeneration and necrosis: Grade 3 (markedly effective; no viable cancer cells); grade 2 (moderately effective; viable cancer cells accounting for less than 1/3 of the tumor tissue while other cancer cells showed severe degeneration or necrosis); grade 1 (slightly effective; apparently viable cancer cells accounting for 1/3 or more of the tumor tissue, but some evidence of degenerating cancer tissue or cells was present), and grade 0 (ineffective; denoting no discernible therapeutic effect on cancer tissue or cells). Grade 1 lesions can also be subclassified into grade 1a (viable cancer cells accounting for 2/3 or more of the tumor tissue) and grade 1b (viable cancer cells accounting for 1/3 or more, but less than 2/3, of the tumor tissue).

To test for statistically significant differences between the two groups, continuous data were compared using the unpaired Student's t-test. Differences were considered significant at two-sided P-values of <0.05. All analyses were performed using JMP version 14.0 (SAS Institute, Cary, NC).

To identify circRNA expression patterns associated with cisplatin resistance in ESCC cells, we performed RNA-seq of three pairs of TE11 and cisplatin-resistant TE11 (TE11R). A total of 10451 circRNAs were consistent with Circexplorer2 (16). Thereafter, we compared circRNAs expression profiles between TE11 and TE11R to extract candidate circRNA. Accordingly, 4791 circRNAs were expressed both in TE11 and TE11R cells (Fig. 1A). Through expression intensity sorting within TE11 and TE11R cells, the mostly greater and lesser circRNAs in TE11R cells compared to those in TE11 cells are shown via hierarchical clustering (Fig. 1B). Variations in circRNA expression are demonstrated in the volcano plot (Fig. 1C). Among them, we selected 10 candidate circRNAs that were significantly upregulated and downregulated in TE11R relative to TE11 (Tables I and II).

To validate the expression profile established by RNA-seq and CIRCexplorer2 (16), we designed divergent primers of each circRNA candidate to specifically target the circular junction and performed RT-qPCR using ESCC cells and tissues. Consistent with the RNA-seq results, we found that the expression of circ_0004365, mapped on the SEMA3C gene (907 bp) (Fig. 2A, B), was significantly greater in TE11R than in TE11. At the same time, the upregulating expression pattern of circ_0004365 in resistant cells was observed in TE8 and TE8R (Fig. 3A). In the meanwhile, there was no difference in circ_0004365 expression between TE11, TE11 5-FU, and TE11 DTX (Fig. S1). Furthermore, the expression of circ_0004365 was broadly detected in human ESCC and nontumor tissues both before and after NAC (Fig. S2A, B). Interestingly, the relative expression of circ_0004365 was significantly greater in ESCC tumor tissues after NAC than in matched tumor tissues before NAC (Fig. S2C). Thus, the present study focused on the relationship between cisplatin resistance and the expression of circ_0004365 in ESCC.

Firstly, we designed siRNA oligonucleotides targeting the specific junction of circ_0004365. RT-qPCR confirmed that they successfully knocked down circ_0004365 expression (Fig. 3B) without affecting the levels of endogenous linear SEMA3C transcript (Fig. 3C). Thereafter, the MTT assay revealed that TE11 with circ_0003465 knockdown had significantly lower resistance to cisplatin compared to control-transfected cells (Fig. 3D).

Similarly, TE11R cells transfected with circ_0003465 siRNA exhibited significantly lower resistance to cisplatin compared to control-transfected cells (Fig. 3E). Consequently, our findings suggested that circ_0004365 regulates the resistance of ESCC cells to cisplatin.

RT-qPCR revealed that circ_0004365 expression was observed in ESCC tissues and nontumor esophageal tissues. Firstly, no significant difference in the expression level of circ_0004365 was noted between ESCC and nontumor esophageal tissues (Fig. S2A, B). However, a comparison of the 15 pairs of ESCC tissues obtained before NAC and after NAC showed that the expression of circ_0004365 was significantly upregulated in tissues after NAC (P=0.03, Fig. S2C). Next, we analyzed whether circ_0004365 expression was correlated with the characteristics of ESCC patients (Table III). Accordingly, no significant correlation was observed between circ_0004365 expression level before NAC and cStage and pStage (Fig. 4A-D). Interestingly, the expression of circ_0004365 before NAC was significantly upregulated in patients with poor pathological response (P=0.02, Fig. 4E). Furthermore, patients with advanced pT Stage showed an upregulation of circ_0004365 expression after NAC (P=0.02, Fig. 5A). No significant difference was noted between the expression level of circ_0004365 after NAC and pN and pathological response (Fig. 5B, C). Together, these data suggested that increased circ_0004365 expression in ESCC patients was correlated with resistance to chemotherapy, including cisplatin.

Biological prediction and analysis predicted that circ_0004365 binds 33 miRNAs to regulate the expression of 187 mRNAs (Table SII), and the ceRNA network, circRNAs, miRNAs, and mRNAs network was visualized (Fig. 6). CircInteractome (26) predicted that miR-503 had the most binding sites for circ_0004365 (Fig. S3A). The sequence of 3 miRNA-binding sites of circ_0004365 and miR-503 is shown in Fig. S3B. We designed 2 pairs of divergent primers to specifically target the binding sites (Table SI). The one targets binding site 1 and 3 continuously and the other targets binding site 2. RT-PCR products with divergent primers showed binding 3 sites of circ_0004365 (Fig. 2C a, b). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed that hsa04390: Hippo signaling pathway, hsa04650: Natural killer cell mediated cytotoxicity, hsa04623: Cytosolic DNA-sensing pathway, and hsa04514: Cell adhesion molecules were enriched (Table SIII).