Circular RNAs (circRNA) are a group of noncoding, covalently uninterrupted loop transcripts, most of which remain to be functionally characterized. Here, we identified circPDIA4 as an oncogenic circRNA in gastric cancer. Clinically, circPDIA4 was significantly upregulated in malignant tissues and was associated with poor survival of patients with gastric cancer. The biogenesis of circPDIA4 was mediated by the RNA-binding protein Quaking, which bound introns 2 and 4 of PDIA4 pre-mRNA to promote backsplicing of exons 3 and 4. Elevated expression of circPDIA4 promoted distant metastasis in various mouse xenograft models in vivo and accelerated cancer cell invasion in vitro. CircPDIA4 functioned through distinct oncogenic mechanisms in the cytoplasm and the nucleus. Cytoplasmic circPDIA4 bound to ERK1/2 and sustained hyperactivation of the MAPK pathway by preventing DUSP6-mediated ERK1/2 dephosphorylation. Notably, circPDIA4 depletion enhanced the sensitivity of gastric cancer cells to ERK inhibitors. In the nucleus, circPDIA4 interacted with DHX9 as a decoy and repressed its inhibitory functions on circRNA biogenesis to boost expression of multiple oncogenic circRNAs, which promoted gastric cancer progression. These findings reveal a dual tumor-promoting mechanism for circPDIA4 by regulating oncogenic circRNA biogenesis and increasing MAPK activity. CircPDIA4 should be investigated further as a potential prognostic biomarker and therapeutic target in gastric cancer.

Significance:

Quaking-regulated circPDIA4 mediates different mechanisms in the nucleus and cytoplasm that coordinate to promote progression and drug resistance in gastric cancer.

Gastric cancer remains the leading cause of cancer-related death worldwide, with an estimated 769,000 deaths according to the GLOBOCAN estimates in 2020 (1). Incidence rates in Eastern Asia and Eastern Europe are generally high (1). Established environmental risk factors of gastric cancer include Helicobacter pylori (H. pylori) infection, alcohol consumption, tobacco smoking, foods preserved by salting, as well as low fruit and vegetable intakes (1, 2). Although chronic H. pylori infection is the main etiology of gastric cancer (1–3), only less than 5% of infected individuals developed gastric cancer, indicating that other factors, especially differences in host genetics, may be crucial during stomach tumorigenesis. Surgical or endoscopic resection is still the major treatment with curative intent for early and locally advanced gastric cancer (4). Advanced gastric cancer is commonly treated with sequential lines of chemotherapy, with a median survival time less than 1 year (4). Hence, it is critical to identify novel curative approaches to improve prognosis and clinical outcomes of patients with gastric cancer.

Epigenetic alterations, including noncoding RNAs, have been established as a fundamental oncogenic mechanism. Circular RNAs (circRNA) are a class of noncoding, covalently closed, single-stranded RNAs with evidently tissue-specific and cell-specific expression patterns (5, 6). Diverse levels of evidences elucidated that circRNAs significantly impacted progression and drug resistance of malignancies in a tumor suppressive or oncogenic manner (7–12). For instance, circURI1 significantly suppresses gastric cancer metastasis via interaction with hnRNPM to modulate alternative splicing of genes controlling the process of cell migration (9). However, the biological importance of circRNAs in invasion and metastasis of gastric cancer cells remains to be explored.

CircRNA biogenesis depends on backsplicing of pre-mRNAs, which has been previously described (13). During the process, a downstream splice donor is covalently joined to an upstream splice acceptor across one or several exon(s) (7, 13, 14). Looping and interacting of the intron downstream of the splice donor site and the intron upstream of the splice acceptor site is required for backsplicing (7, 13, 14). The looping RNA structure can be facilitated by dimerization of RNA-binding proteins (RBP), inverted repeat intron sequences, including Alu elements or non-repetitive complementary intron sequences (15–17). During epithelial–mesenchymal transition (EMT), dimerization of the RBP quaking (QKI) could promote production of certain circRNAs involved in cancer metastasis (18). Similarly, other RBPs, such as HNRNPL, FUS, MBNL1, and RBM20, regulate circRNA biogenesis via binding to intronic RNAs flanking the backsplicing junctions (14, 19–21). On the contrary, DHX9 and ADAR1 suppress Alu-mediated circRNA biogenesis through destabilizing regional double-stranded RNA structures (22–24).

In the current study, we identified a novel QKI-regulated circRNA, circPDIA4, which is backspliced from exons 3 and 4 of PDIA4 pre-mRNA in gastric cancer. CircPDIA4 exhibits markedly higher expression levels in gastric cancer specimens compared with normal tissues and its high expression levels are associated with worse survival of patients with gastric cancer. Consistently, circPDIA4 could promote cell migration and invasion in vitro and gastric cancer metastasis in vivo. CircPDIA4 could competitively interact with phosphorylated ERK1/2 (pERK) to inhibit its dephosphorylation by DUSP6 in cytoplasm, which, in turn, led to activated MAPKs signaling. Interestingly, the depletion of circPDIA4 could significantly promote antineoplastic sensitivities of pERK inhibitors. Here, we also found that circPDIA4 could bind DHX9 in nucleus, repress functions of DHX9 as the negative regulator in circRNA biogenesis, and induce elevated expression multiple oncogenic circRNAs, which, thereby, promote progression of gastric cancer.

Cell culture

Human gastric cancer MKN-45 cells were cultured in RPMI-1640 medium (Gibco). Human gastric cancer HGC-27 and HEK293T cells were cultured in DMEM medium (Gibco). MKN-45, HGC-27, and HEK293T cells were kindly provided by Dr. Yunshan Wang (Jinan Central Hospital, Shandong Province, China). All media were supplemented with 10% FBS (Gibco). Cells were maintained at 37°C in a 5% CO2 incubator and periodically tested Mycoplasma negative using MycoAlert Mycoplasma Detection Kit (Lonza). Cell stocks were conducted within five passages, and all experiments were completed within eight passages.

RT-qPCR

Total RNA was isolated and then reverse transcribed into cDNAs as described previously (25). Expression levels of candidate genes or circRNAs were measured using TB Green Premix Ex Taq II (TaKaRa, RR820A) with indicated primers (Supplementary Table S1). The expression of candidate genes or circRNAs was calculated by using the 2−ΔΔCt method.

RNase R treatment

Total cellular RNA (2 μg) was incubated with 6U RNase R (20 U/mL, Lucigen) at 37°C for 30 minutes. After incubated at 65°C for 20 minutes to inactivate RNase R, RNA was then examined through RT-qPCR.

Patients and tissue specimens

A total of two gastric cancer patient cohorts (Discovery cohort and Validation cohort) were enrolled in this study. The detailed patient characteristics have been reported previously (25). This study was approved by the Institutional Review Board of Shandong Cancer Hospital and Institute. The methods were carried out in accordance with the Declaration of Helsinki. At recruitment, written informed consent was obtained from each subject.

The plasmid constructs

The circPDIA4 expression plasmid was generated by cloning the sequence of exons 3 and 4 of PDIA4 using pCDH-CMV-MCS-EF1-Puro vector. An additional circulation promoter sequence and an AG/GT splicing sequence were added 105bp upstream and 108bp downstream. The plasmid was designated circPDIA4. The human full-length QKI cDNA was directly synthesized and cloned into pCDH-CMV-MCS-EF1-Puro. The plasmid was named as QKI. After one HA-tag sequence was inserted after ATG of the CDS region of DHX9, the cDNA was cloned into the pcDNA3.1 vector to generate the HA-tagged DHX9 plasmid (Full-length). Two truncated DHX9 plasmids (Del_1–329aa or Del_329–1283aa) were mutants of the HA-tagged DHX9 plasmid with CDS region after deletion of 1–329aa or 329–1283aa, separately. Two shRNA hairpins targeting the back splicing sites of human circPDIA4 (shPD-1 or shPD-2) or the control shRNA (Supplementary Table S2) were cloned into the pLKO.1 vector (Addgene_13425). The resultant plasmids were designated shPD-1, shPD-2, or shNC. All the plasmids were synthesized by Genewiz and sequenced to confirm the orientation and integrity.

Lentiviral transduction

As previously reported, recombinant lentiviral particles were produced by transient cotransfection of the shcircPD-1, shcircPD-2 or circPDIA4 plasmids into HEK293T cells (25). Gastric cancer cells were infected with viral supernatant containing 8 μg/mL polybrene. Stably circPDIA4-overexpression (OE) cells were selected using 2 μg/mL puromycin. Stably circPDIA4-knockdown (KD) cells were selected using 10 μg/mL blasticidin. In these cells, circPDIA4 expressions were tested by RT-qPCR.

Cell proliferation and ERK1/2 inhibitor sensitivity analyses

A total of 3 × 104circPDIA4-OE or circPDIA4-KD cells were seeded in 12-well plates and harvested and counted at 24, 48, and 72 hours after seeding. For drug-sensitivity analyses of ERK1/2 inhibitors (MK-8353, Selleck, S870101; ulixertinib, Selleck, S7854), a total of 4,000 stably circPDIA4-OE, circPDIA4-KD or control HGC-27 or MKN-45 cells were seeded per well in 96-well plates. MK-8353 or ulixertinib diluted in DMSO was added to each well to achieve the desired final concentrations. After cells were incubated with MK-8353 or ulixertinib for 48 hours, 20 μL of 5 mg/mL MTT was added to each well. After the cells were incubated with MTT for 4 hours at 37°C in a 5% CO2 incubator, 100 μL DMSO were added to each well. Absorbance in each well was measured at 492 nm using a microplate reader (Spectramaxi3, Molecular Device).

Colony formation assays

A total of 1,000 HGC-27 or MKN-45 cells per well were seeded in 6-well plates. When colonies were visible after 14 days, cells were washed with PBS, fixed with the fixation fluid (methanol:acetic acid = 3:1) and dyed with crystal violet. The colony number in each well was then counted.

Wound-healing and Transwell assays

The wound-healing and Transwell assays were performed as previously reported (25). For the rescue assays, various siRNAs and NC RNA (Genepharma; Supplementary Table S2) were transfected to cells at 24 or 48 hours before the Transwell assays.

Western blot analysis

Western blot analysis was performed following the standard protocol as previously reported (25). In brief, after separated with SDS-PAGE gel, total cellular proteins were transferred to a polyvinylidene fluoride (PVDF) membrane. The PVDF membrane was then incubated with various antibodies (Supplementary Table S3) overnight at 4°C. The ECL Western Blotting Substrate (Pierce, 32106) was used to visualize target proteins.

Xenografts

To examine the in vivo role of circPDIA4 during hematogenous metastases, we inoculated a total of 2 × 106 HGC-27 cells with stable firefly luciferase expression (NC, shcircPD-1, shcircPD-2, vector or circPDIA4) into tail vein of 5-week-old female nude BALB/c mice (Vital River Laboratory; n = 4 per group). To investigate the role of circPDIA4 in abdominal metastases, a total of 1 × 107 HGC-27 cells with stable firefly luciferase expression (vector or circPDIA4) were injected in the middle of the lower abdomen of 5-week-old female nude BALB/c mice (n = 3 per group). Bioluminescent gastric cancer metastases were monitored weekly via the IVIS Spectrum In Vivo Imaging System (PerkinElmer). Mouse lungs and livers with metastasis tumors were formalin-fixed, paraffin-embedded and stained with hematoxylin and eosin (H&E). IHC staining was performed in mice livers with antibodies specific for CEA as previously described (25). All procedures involving mice were approved by the Animal Care Committee of Shandong Cancer Hospital and Institute. All analyses were performed in a blinded fashion with individuals unaware of xenograft types.

Cell transfection

siRNA duplexes for QKI (siQKI-1 and siQKI-2), DHX9 (siDHX9–1 and siDHX9–2), SP1 (siSP1–1 and siSP1–2), MAPK1 (siMAPK1–1 and siMAPK1–2), DUSP6 (siDUSP6–1 and siDUSP6–2), and hsa_circ_0001610 (si1610–1 and si1610–2) were products of Genepharma (Supplementary Table S2). The negative control RNA duplex (NC) for siRNAs (Genepharma) was nonhomologous to any human genome sequence. As reported previously, the INTERFERin reagent (Polyplus) was used to transfect all small RNAs (25). All plasmids were transfected with the jetPRIME reagent (Polyplus).

ChIP-qPCR

The chromatin immunoprecipitation (ChIP) assays were performed using a total of 2.5 × 107 HGC27 or MKN45 cells as previously described (25). Inputs and ChIP products were detected using qPCR.

Dual luciferase reporter assays

We cloned a 1000bp DNA fragment (−1bp∼−1000bp of QKI) into the pGL3-Basic vector (Promega; QKI promoter WT). The mutant QKI promoter–reporter construct (QKI promoter MU) is the QKI promoter WT plasmid with mutated SP1-binding sites (a, b, and c). HGC27 or MKN45 cell lines were transfected with QKI promoter WT or QKI promoter MU and pRL-SV40 (Promega). As reported previously, dual luciferase activities were detected at 48 hours after transfection (26).

circRNA pulldown and RNA pulldown

The circRNA pulldown assays were performed to identify proteins interacting with circPDIA4. In briefs, protein extracts of HGC-27 cells were incubated with biotin-labeled circPDIA4 probes (Supplementary Table S4; Genepharma) at room temperature for 1 hour. The streptavidin magnetic beads were added into the cell lysates with the probes and incubated at 4°C overnight. After beads were washed for three times, proteins bound on the streptavidin magnetic beads were recovered with Elution Buffer following the instruction of Pierce Magnetic RNA-Protein Pull-Down Kit (Thermo Fisher Scientific). The retrieved proteins were analyzed either through LS/MS-MS (Hoogen Biotech Co.) or Western blot analysis. The MaxQuant software (version 1.5.3.30) was used to analyze mass spectra with the UniProtKB human database (Uniport Homo sapiens 188441_20200326; RRID: SCR_004426). For PDIA4 intron 2 RNA or intron 4 RNA pulldown, wild-type PDIA4 intron 2 or 4, mutant PDIA4 intron 2 or 4 with all predicted QKI-binding sites deleted, and SMACRA5 intron 16 as the positive control were subcloned into pcDNA3.1 with inserted T7 promoter to prepare the templates for in vitro RNA synthesis. The RNA pulldown and Western blot assays were performed as reported previously (25).

RNA immunoprecipitation

As reported previously, RNA immunoprecipitation assays were performed with the QKI, DHX9, or ERK1/2 antibodies or IgG isotype-control (25). The protein–RNA complexes were recovered by Dynabeads Protein G beads (Thermo Fisher Scientific). Relative RNA levels in the precipitates were examined by RT-qPCR. A total of 10% of inputs were used for RT-qPCR.

Subcellular fractionation

According to the manufacturer's instructions, the cytosolic and nuclear fractions of HGC-27 or MKN45 cells were separately isolated using the nuclear/cytoplasmic Isolation Kit (Biovision, K266).

Coimmunoprecipitation

Coimmunoprecipitation (Co-IP) was carried out between pERK1/2 and DUSP6 as described previously (25). Gastric cancer cell lysates were incubated with antibodies of pERK1/2, DUSP6 or control IgG overnight at 4°C and with Dynabeads Protein G beads (Thermo Fisher Scientific) at the next day for 2 hours at 4°C. After the beads were washed with the lysis buffer for five times, the recovered proteins were examined using Western blot analysis. A total of 1% of inputs was used for Western blot analysis.

Immunofluorescence

The immunofluorescence assay was performed as reported previously (27). In brief, after fixed in 4% paraformaldehyde for 20 min, HGC-27 or MKN-45 cells were permeabilized with 0.2% Triton X-100, blocked with 1% BSA and incubated with appropriate primary antibodies overnight. Gastric cancer cells were stained with coraLite488-conjugated or coraLite594-conjugated secondary antibodies (Supplementary Table S3), followed by washing with PBS and staining with 4,6-diamidino-2-phenylindole. Images were visualized and recorded with a Zeiss LSM800 confocal microscope (Zeiss, Germany).

RNA-seq

To gain insight into how circPDIA4 regulates gene expression in gastric cancer cells, we performed RNA-seq of HGC-27 and circRNA-seq of MKN-45 cells with stable expression of circPDIA4. Total RNA was isolated from cultured cells using TRizol, and then performed the sequencing using NovaSeq 6000 platform (Illumina). RNA-seq was performed as reported previously (25). For circRNA-seq, the linear RNA was digested with 3U of RNase R per μg of RNA before establishing the cDNA library. Reads count of samples were calculated and converted to FPKM (fragments per kilobase of exon model per million reads mapped). KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway analyses of differentially expressed genes (|log2(FC)|>2, P < 0.05) from RNA-seq were performed.

Statistical analysis

The Student t test was used for calculation of the difference between two groups. The difference between multiple groups was examined using one-way ANOVA analysis with the Dunnett test. The significance of association of expression between different genes was calculated using Spearman's correlation. Impacts of circPDIA4 expression on patients with gastric cancer survival were tested by Kaplan–Meier plots and the median value of candidate gene expression in normal tissues as the cutoff value. The survival durations were analyzed using the log-rank test. A P value of less than 0.05 was used as the criterion of statistical significance. All analyses were performed with SPSS software package (Version 16.0, SPSS Inc.; RRID:SCR_002865) or GraphPad Prism (Version 8.0, GraphPad Software, Inc.).

Data availability

The datasets produced in this study are available in Genome 326 Sequence Archive-human HRA002373 (https://ngdc.cncb.ac.cn/gsa-human/). The datasets in Genome 326 Sequence Archive-human HRA002373, as well as the materials and data used in the study, are available upon request from the corresponding author.

Elevated expression of circPDIA4 in gastric cancer tissues was associated with shortened survival time of patients

To identify circRNAs involved in gastric cancer development, we systematically analyzed the cancer-specific circRNA database (http://gb.whu.edu.cn/CSCD; Fig. 1A and Supplementary Fig. S1A; ref. 28). We identified 44 cancer-specific, potential circRNAs in gastric cancer HGC-27, MKN-45 and AGS cell lines (Fig. 1A; Supplementary Fig. S1A). There are 11 novel circRNAs, 17 circRNAs in circBase (http://www.circbase.org/), and 16 undefined RNAs (Fig. 1A; Supplementary Table S5). We then performed RT-PCR, RNase R treatment and Sanger sequencing assays to validate the existence of eleven novel circRNAs and seventeen circRNAs in circBase in gastric cancer cells. It has been found that only circPDIA4 were successfully validated in gastric cancer cells (Fig. 1A). The head-to-tail splicing in the circPDIA4 RT-PCR product was confirmed by Sanger sequencing (Fig. 1B). For PCR verification, circPDIA4 could be amplified by divergent primers (Div-F and Div-R) using cDNA of gastric cancer cells, but not using genomic DNA of the cells (Fig. 1C). Moreover, circPDIA4 was resistant to RNase R treatment, whereas the linear RNAs PDIA4 and β-actin were considerably digested with RNase R (Fig. 1D).

Figure 1.

Identification of a novel circRNA, circPDIA4, derived from PDIA4. A, A schematic view for identification of the candidate circRNAs in gastric cancers. B, The genomic location of circPDIA4 derived from its parent gene PDIA4. The existence of circPDIA4 was validated by Sanger sequencing. Red horizontal arrows, divergent primers used to identify circPDIA4. Light red, junction site of circPDIA4. C, RT-PCR products for detection of circPDIA4 using divergent primers (◂▸) or convergent primers (▸◂) and cDNA or genomic DNA (gDNA) from HGC-27 or MKN-45 as templates. D, RT-qPCR analyses for the expression of circPDIA4 and PDIA4 mRNA after gastric cancer cells treated with or without RNase R. Data represent mean ± SD. The P value was determined by a two-tailed unpaired Student t test. ns, not significant; ***, P < 0.001. E and F, RT-qPCR assay showing the expression levels of circPDIA4 (normalized to β-actin) in gastric cancer tissues and normal tissues in Discovery cohort, Validation cohort, and combined data. Data represent mean ± SD. The P value was determined by a two-tailed paired Student t test. **, P < 0.01; ***, P < 0.001. G, The patients with gastric cancer with higher circPDIA4 expression had a shorter PFS and OS compared with patients with lower circPDIA4 expression (Discovery cohort). The log-rank test was used for survival comparison. The P value was determined by a two-tailed unpaired Student t test. ns, not significant; **, P < 0.01; ***, P < 0.001.

Figure 1.

Identification of a novel circRNA, circPDIA4, derived from PDIA4. A, A schematic view for identification of the candidate circRNAs in gastric cancers. B, The genomic location of circPDIA4 derived from its parent gene PDIA4. The existence of circPDIA4 was validated by Sanger sequencing. Red horizontal arrows, divergent primers used to identify circPDIA4. Light red, junction site of circPDIA4. C, RT-PCR products for detection of circPDIA4 using divergent primers (◂▸) or convergent primers (▸◂) and cDNA or genomic DNA (gDNA) from HGC-27 or MKN-45 as templates. D, RT-qPCR analyses for the expression of circPDIA4 and PDIA4 mRNA after gastric cancer cells treated with or without RNase R. Data represent mean ± SD. The P value was determined by a two-tailed unpaired Student t test. ns, not significant; ***, P < 0.001. E and F, RT-qPCR assay showing the expression levels of circPDIA4 (normalized to β-actin) in gastric cancer tissues and normal tissues in Discovery cohort, Validation cohort, and combined data. Data represent mean ± SD. The P value was determined by a two-tailed paired Student t test. **, P < 0.01; ***, P < 0.001. G, The patients with gastric cancer with higher circPDIA4 expression had a shorter PFS and OS compared with patients with lower circPDIA4 expression (Discovery cohort). The log-rank test was used for survival comparison. The P value was determined by a two-tailed unpaired Student t test. ns, not significant; **, P < 0.01; ***, P < 0.001.

Close modal

The circPDIA4 expression was further analyzed and compared in gastric cancer specimens and normal tissues of patients from Discovery cohort and Validation cohort (Fig. 1E; Supplementary Table S6). circPDIA4 showed evidently increased expression in gastric cancer tissues compared with normal tissues in Discovery cohort (P < 0.01; Fig. 1E; Supplementary Fig. S1B). Consistently, markedly upregulated expression of circPDIA4 in cancerous tissues was observed compared with normal tissues in Validation cohort and combined data of both cohorts (both P < 0.001; Fig. 1E and F; Supplementary Fig. S1B). There were no evident circPDIA4 expression differences between four molecular subtypes (Epstein–Barr virus, microsatellite instability, genomically stable, and chromosomal instability; P = 0.998) or among patients with or without H. pylori infection (P = 0.295; Supplementary Fig. S1C–S1D). No difference of host gene PDIA4 expression was observed between cancer tissues and normal tissues (Supplementary Fig. S1E). Importantly, circPDIA4 levels are significantly associated with progression-free survival (PFS) and overall survival (OS) in patients with gastric cancer of Discovery cohort (PFS: log-rank P = 0.003; OS: log-rank P < 0.001; Fig. 1G). Patients with lower circPDIA4 expression had prolonged time of PFS or OS compared with ones with high circPDIA4 levels (Fig. 1G), suggesting that circPDIA4 might be involved in progression of gastric cancer as a novel oncogene.

circPDIA4 enhanced invasive activities of gastric cancer cells and their formation of metastases in mice

It remains largely unknown how circPDIA4 is involved in gastric cancer development. Therefore, we first generated multiple gastric cancer cell lines with stably silenced circPDIA4 by the lentivirus-packaged shRNA plasmids (NC, shcircPD-1 and shcircPD-2) or forced-expressed circPDIA4 by the lentivirus-packaged overexpression plasmid (Vector and circPDIA4; Fig. 2A). Although circPDIA4 showed no obvious impacts on viability of gastric cancer cells (Supplementary Fig. S1F–S1H), silencing or overexpression of circPDIA4 significantly change migration, invasion, and metastasis capabilities of gastric cancer cells (Fig. 2B and C). The wound-healing assays indicated that stable circPDIA4 KD impaired cell wound healing (Fig. 2B; Supplementary Fig. S2A–S2B). In line with this, stabilized circPDIA4 overexpression accelerated cancer cell migration (Fig. 2B; Supplementary Fig. S2A–S2B). The Matrigel invasion assays elucidated that silencing of circPDIA4 could reduce invasion of gastric cancer cells (Fig. 2C; Supplementary Fig. S2C). In contrast, ectopic circPDIA4 led to increased cancer cell invasion (Fig. 2C; Supplementary Fig. S2C). To explore the mechanistic rationale, we examined the levels of several EMT markers in cells. Intriguingly, knockout of circPDIA4 reduced expression of N-cadherin, VIMENTIN and Snail/Slug/Twist family members and stimulated expression of E-cadherin and ZO-1; whereas forced expression of circPDIA4 could markedly promote expression of N-cadherin, VIMENTIN and Snail/Slug/Twist family members and inhibit expression of E-cadherin and ZO-1 (Fig. 2D; Supplementary Fig. S2D–S2G). In addition, we found significantly negative expression correlations between TJP1 (the gene coding ZO-1) and circPDIA4 or QKI in malignant or normal tissues (Supplementary Fig. S2H–S2K). On the contrary, we observed obviously positive expression correlations between CDH2 (the gene coding N-cadherin) and circPDIA4 or QKI in tissues (Supplementary Fig. S2H–S2K). These results indicated that circPDIA4 evidently enhances EMT, migration, and invasion of gastric cancer cells.

Figure 2.

circPDIA4 promoted migration and metastasis capabilities of gastric cancer cells in vitro and in vivo. A, Relative expression of PDIA4 mRNA or circPDIA4 in gastric cancer HGC-27 and MKN-45 cell lines that stabilized either silenced circPDIA4 (by shRNAs) or overexpressed circPDIA4. Data represent mean ± SD. B, In HGC-27 and MKN-45 cells, circPDIA4 knockdown inhibited wound healing and the stably enforced circPDIA4 expression accelerated wound healing. C, circPDIA4 promoted migration capabilities of HGC-27 and MKN-45 cells. Cells on the lower surface of the chamber were stained by crystal violet. D, In HGC-27 and MKN-45 cells, expression changes of different markers of epithelial-to-mesenchymal transition (E-cadherin, N-cadherin, ZO-1, and VIMENTIN) were examined after overexpression or knockdown of circPDIA4. E and F, Decreased or increased tumor metastases were observed in lungs of nude mice via tail vein injection of circPDIA4-knockdown or circPDIA4-overexpression HGC-27 cells (n = 4). E and F, Luciferase activities of cancer cells were detected at the 45th day (E) or the 35th day (F) after injection. G and H, Representative images of H&E–stained slides of lung metastatic nodules. Scale bar, 50 μm. I–K, circPDIA4 induced liver metastasis in nude mice via intraperitoneal injection of HGC-27 cells (vector or circPDIA4; n = 3). I, Representative images of the liver metastases. J, H&E staining of liver metastases. K, Representative images of IHC staining showing the expression of CEA within liver metastases. Scale bar, 50 μm. The P value was determined by a two-tailed unpaired Student t test. ns, not significant; **, P < 0.01; ***, P < 0.001.

Figure 2.

circPDIA4 promoted migration and metastasis capabilities of gastric cancer cells in vitro and in vivo. A, Relative expression of PDIA4 mRNA or circPDIA4 in gastric cancer HGC-27 and MKN-45 cell lines that stabilized either silenced circPDIA4 (by shRNAs) or overexpressed circPDIA4. Data represent mean ± SD. B, In HGC-27 and MKN-45 cells, circPDIA4 knockdown inhibited wound healing and the stably enforced circPDIA4 expression accelerated wound healing. C, circPDIA4 promoted migration capabilities of HGC-27 and MKN-45 cells. Cells on the lower surface of the chamber were stained by crystal violet. D, In HGC-27 and MKN-45 cells, expression changes of different markers of epithelial-to-mesenchymal transition (E-cadherin, N-cadherin, ZO-1, and VIMENTIN) were examined after overexpression or knockdown of circPDIA4. E and F, Decreased or increased tumor metastases were observed in lungs of nude mice via tail vein injection of circPDIA4-knockdown or circPDIA4-overexpression HGC-27 cells (n = 4). E and F, Luciferase activities of cancer cells were detected at the 45th day (E) or the 35th day (F) after injection. G and H, Representative images of H&E–stained slides of lung metastatic nodules. Scale bar, 50 μm. I–K, circPDIA4 induced liver metastasis in nude mice via intraperitoneal injection of HGC-27 cells (vector or circPDIA4; n = 3). I, Representative images of the liver metastases. J, H&E staining of liver metastases. K, Representative images of IHC staining showing the expression of CEA within liver metastases. Scale bar, 50 μm. The P value was determined by a two-tailed unpaired Student t test. ns, not significant; **, P < 0.01; ***, P < 0.001.

Close modal

We then investigated whether circPDIA4 may impact in vivo metastasis of gastric cancer cells. To determine the in vivo role of circPDIA4 during hematogenous metastases, stably circPDIA4-KD HGC-27 cells, stably circPDIA4-OE cells or NC cells were injected into mice via the tail vein. The mouse model proved that circPDIA4 depletion led to obviously inhibited hematogenous metastasis of gastric cancer cells (both P < 0.001; Fig. 2E). It has been also found that stabilized circPDIA4 overexpression remarkably enhanced distant gastric cancer metastases (P < 0.001; Fig. 2F). These results were further confirmed by the H&E staining analyses of the metastasis tumors of lungs from these mice (Fig. 2G and H). We next examined the in vivo role of circPDIA4 in abdominal metastases via injecting HGC-27 cells into the abdomen of mice. Interestingly, circPDIA4 evidently promoted invasiveness into the liver parenchyma of gastric cancer cells compared with the control cells (Fig. 2I). Consistently, the H&E and IHC stainings with CEA antibody verified the metastatic xenografts in mouse livers (Fig. 2J and K). Taken together, these findings demonstrated that circPDIA4 significantly enhances gastric cancer migration and metastasis in vitro and in vivo.

The RBP QKI promotes circPDIA4 biogenesis

We next explored how circPDIA4 biogenesis is regulated in gastric cancer cells. As shown in Fig. 3A, human PDIA4 RNA exons 3 and 4 circularization forms circPDIA4. Recent reports indicate that multiple RBPs participate in regulation of circRNA formation (14, 18–21). Therefore, we predicted which RBP is involved in regulation of circPDIA4 production through RBPmap (http://rbpmap.technion.ac.il/; ref. 29). Prediction of RBPs binding indicated that there are 128 RBPs might bind both PDIA4 introns 2 and 4 RNA (Supplementary Fig. S3A). After evaluation of the prognostic significance of these RBPs in the The Cancer Genome Atlas (TCGA) stomach adenocarcinoma (STAD) cohort, we found that the expression levels of 10 RBPs are significantly associated with OS of patients with gastric cancer (Supplementary Fig. S3B). Among these RBPs, QKI is the only one that has been reported to control circRNA biogenesis during EMT (18). As shown in Supplementary Fig. S3, there are multiple QKI-binding motifs in both introns 2 and 4 of PDIA4. We therefore chose QKI for further investigation.

Figure 3.

The RBP QKI induced the formation of circPDIA4 in gastric cancer cells. A, A schematic view showing that QKI binds the introns 2 and 4 of PDIA4 pre-mRNA through QKI-binding motifs to induce circPDIA4 formation. B, The expression levels of QKI in HGC-27 or MKN-45 cells silencing with two siRNAs targeting QKI or overexpressing with the QKI expression plasmid (top, RT-qPCR; bottom, Western blot). C and D, The expression levels of circPDIA4 in QKI-knockdown or QKI-overexpressed HGC-27 and MKN-45 cells. Data represent mean ± SD. The P value was determined by a two-tailed unpaired Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001. E, RT-qPCR assay showing the QKI expression levels in gastric cancer tissues and normal tissues in Discovery cohort and Validation cohort. The P value was determined by a two-tailed paired Student t test. ***, P < 0.001. F, Significantly elevated QKI expression in gastric cancer tissues compared with normal tissues in GSE29272. The P value was determined by a two-tailed paired Student t test. ***, P < 0.001. G, Representative QKI protein expression levels in paired gastric cancer and normal tissues. H, Kaplan–Meier plots of OS in patients with gastric cancer from Discovery cohort stratified by the QKI expression level. The log-rank test was used for survival comparison. I, Kaplan–Meier plots of OS in patients with gastric cancer from TCGA STAD cohort stratified by the QKI expression level. J, The log-rank test was used for survival comparison. In gastric cancer or normal tissues of Discovery and Validation cohorts, expression correlation between circPDIA4 and QKI was determined by RT-qPCR, with β-actin serving as an internal control. Statistical analyses were performed with Pearson correlation analyses.

Figure 3.

The RBP QKI induced the formation of circPDIA4 in gastric cancer cells. A, A schematic view showing that QKI binds the introns 2 and 4 of PDIA4 pre-mRNA through QKI-binding motifs to induce circPDIA4 formation. B, The expression levels of QKI in HGC-27 or MKN-45 cells silencing with two siRNAs targeting QKI or overexpressing with the QKI expression plasmid (top, RT-qPCR; bottom, Western blot). C and D, The expression levels of circPDIA4 in QKI-knockdown or QKI-overexpressed HGC-27 and MKN-45 cells. Data represent mean ± SD. The P value was determined by a two-tailed unpaired Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001. E, RT-qPCR assay showing the QKI expression levels in gastric cancer tissues and normal tissues in Discovery cohort and Validation cohort. The P value was determined by a two-tailed paired Student t test. ***, P < 0.001. F, Significantly elevated QKI expression in gastric cancer tissues compared with normal tissues in GSE29272. The P value was determined by a two-tailed paired Student t test. ***, P < 0.001. G, Representative QKI protein expression levels in paired gastric cancer and normal tissues. H, Kaplan–Meier plots of OS in patients with gastric cancer from Discovery cohort stratified by the QKI expression level. The log-rank test was used for survival comparison. I, Kaplan–Meier plots of OS in patients with gastric cancer from TCGA STAD cohort stratified by the QKI expression level. J, The log-rank test was used for survival comparison. In gastric cancer or normal tissues of Discovery and Validation cohorts, expression correlation between circPDIA4 and QKI was determined by RT-qPCR, with β-actin serving as an internal control. Statistical analyses were performed with Pearson correlation analyses.

Close modal

To verify whether the production of circPDIA4 was dependent on QKI, the effects of siRNA-mediated QKI KD or plasmid-mediated QKI overexpression were tested. In HGC27 and MKN45 cells, KD of QKI obviously suppressed circPDIA4 expression levels (all P < 0.05; Fig. 3B and C). Conversely, overexpression of QKI could significantly upregulate circPDIA4 levels in gastric cancer cells (both P < 0.001; Fig. 3B and D). In Discovery and Validation cohorts, QKI showed markedly elevated expression in gastric cancer tissues compared with normal tissues (both P < 0.01; Fig. 3E; Supplementary Table S7). Transcriptional factor SP1 plays a crucial role in upregulating QKI expression in gastric cancer (Supplementary Fig. S4). In line with these data, evidently increased expression of QKI in cancerous specimens was observed compared with normal tissues in an independent gastric cancer cohort from a high-risk population in China (P < 0.001; Fig. 3F; refs. 30, 31). QKI protein expression in gastric cancer tissues was also found to be upregulated (Fig. 3G). Importantly, high QKI levels were significantly associated with shortened OS time in patients with gastric cancer of Discovery cohort (log-rank P < 0.001) or the TCGA STAD cohort (log-rank P = 0.005; Fig. 3H and I). Similarly, patients with lower QKI expression had prolonged time of PFS compared with ones with high QKI levels (Supplementary Fig. S4G). In addition, there was increased abundance of QKI in tissues of patients with advanced diseases, especially stage III or IV tumors (P = 0.007; Supplementary Fig. S4H), suggesting that QKI significantly contributes to gastric cancer progression. In support of the regulatory relationship between QKI and circPDIA4, noticeably positive expression correlations between QKI and circPDIA4 were observed in gastric cancer and normal tissues from different patient cohorts (Fig. 3J). Taken together, these data elucidated that QKI binds upstream intron 2 RNA and downstream intron 4 RNA of the circRNA-forming exons in PDIA4 to promote circPDIA4 biogenesis.

circPDIA4 sustained high phosphorylation levels of ERK1/2 through interrupting interactions between ERK1/2 and DUSP6

Considering that circRNAs could function through interacting with different proteins, we hypothesized that circPDIA4 might bind certain protein(s) to promote gastric cancer development. To test it, we first detected where circPDIA4 locates in cells and found that it is in both the nuclear fraction and the cytoplasm fraction of gastric cancer cells (Fig. 4A). We found that multiple proteins could be pulled-down by circPDIA4 through RNA pulldown assays using HGC27 cellular extracts (Fig. 4B and C; Supplementary Fig. S5A). Mass spectrometry proteomics revealed that several cancer-related proteins, such as DHX9, ERK1/2, ATXN2L, KIF23, PRMT5, YBX1, TPM3 and HNRPA1, could be pulldown by circPDIA4 (Supplementary Table S8). After validating these candidate proteins in both HGC27 and MKN45 cells through independent RNA pulldown assays, we successfully confirmed DHX9 and ERK1/2 (Fig. 4D). Consistently, RNA immunoprecipitation (RIP) assays also proved more than 100-fold enrichment of circPDIA4 in RNA–protein complexes precipitated with antibody against DHX9 or ERK1/2 as compared with the IgG control in gastric cancer cells (all P < 0.001; Fig. 4E).

Figure 4.

circPDIA4 sustained high phosphorylation levels of ERK1/2 through interrupting binding of DUSP6 with ERK1/2. A, Cellular location of circPDIA4 in HGC-27 and MKN-45 cells. B, Flowchart of identifying circPDIA4-binding proteins by circRNA pulldown and mass spectrometry. C, Silver staining image of circPDIA4 pull-downed proteins of gastric cancer cells. D, circPDIA4 pulldown followed by Western blot analysis validated the interaction between circPDIA4 and ERK1/2, DHX9, or other candidate proteins. E, RIP assays showed association of ERK1/2 or DHX9 with circPDIA4 in HGC-27 and MKN-45 cells. Relative enrichment (means ± SD) represents RNA levels associated with ERK1/2 and DHX9 relative to an input control from three independent experiments. IgG served as the control. F, Western blot analyses of protein expression of ERK1/2 and its phosphorylation levels (pERK1/2) in HGC-27 and MKN-45 cells with stabilized silencing of circPDIA4 (top) or overexpressing circPDIA4 (bottom). GAPDH served as the control. G and H, Interactions between pERK1/2 and DUSP6 in gastric cancer cells were verified via Co-IP and immunofluorescence assays. I and J, circPDIA4 impaired interactions between pERK1/2 and DUSP6 in HGC-27 and MKN-45 cells. K, Knockdown of circPDIA4 increased the antiproliferation effects of the ERK1/2 inhibitor (MK-8353) in gastric cancer cells. Data represent mean ± SD. The P values were determined by a two-tailed unpaired Student t test. **, P < 0.01; ***, P < 0.001.

Figure 4.

circPDIA4 sustained high phosphorylation levels of ERK1/2 through interrupting binding of DUSP6 with ERK1/2. A, Cellular location of circPDIA4 in HGC-27 and MKN-45 cells. B, Flowchart of identifying circPDIA4-binding proteins by circRNA pulldown and mass spectrometry. C, Silver staining image of circPDIA4 pull-downed proteins of gastric cancer cells. D, circPDIA4 pulldown followed by Western blot analysis validated the interaction between circPDIA4 and ERK1/2, DHX9, or other candidate proteins. E, RIP assays showed association of ERK1/2 or DHX9 with circPDIA4 in HGC-27 and MKN-45 cells. Relative enrichment (means ± SD) represents RNA levels associated with ERK1/2 and DHX9 relative to an input control from three independent experiments. IgG served as the control. F, Western blot analyses of protein expression of ERK1/2 and its phosphorylation levels (pERK1/2) in HGC-27 and MKN-45 cells with stabilized silencing of circPDIA4 (top) or overexpressing circPDIA4 (bottom). GAPDH served as the control. G and H, Interactions between pERK1/2 and DUSP6 in gastric cancer cells were verified via Co-IP and immunofluorescence assays. I and J, circPDIA4 impaired interactions between pERK1/2 and DUSP6 in HGC-27 and MKN-45 cells. K, Knockdown of circPDIA4 increased the antiproliferation effects of the ERK1/2 inhibitor (MK-8353) in gastric cancer cells. Data represent mean ± SD. The P values were determined by a two-tailed unpaired Student t test. **, P < 0.01; ***, P < 0.001.

Close modal

To reveal the mechanisms underlying circPDIA4-mediated downstream signaling in gastric cancer, we profiled genome-wide gene expression in HGC27 cells using RNA-seq. Indeed, ectopic circPDIA4 expression induced 258 significantly upregulated or downregulated genes in cells (circPDIA4 vs. Vector; |log2(FC)|>2, P < 0.05; Supplementary Fig. S5B). KEGG pathway analyses of these genes demonstrated that the MAPK signaling is the top changed signaling pathways (Supplementary Fig. S5C). Consistently, although circPDIA4 did not impact total ERK1/2 protein levels in cells (Fig. 4F), silencing of circPDIA4 resulted in diminished pERK1/2 levels in gastric cancer cells (Fig. 4F). On the contrary, overexpressed circPDIA4 obviously upregulated pERK1/2 levels in cells (Fig. 4F).

It has been reported that DUSP6 (also known as MKP3) is a dual-specificity cytoplasmic phosphatase that recognizes and rapidly dephosphorylates pERK1/2, thus preventing nuclear translocation of pERK1/2 and activation of MAPK signaling (32, 33). Considering the importance of DUSP6 in controlling ERK1/2 phosphorylation in cytosol, we first investigated whether DUSP6 is also a partner protein of pERK1/2 in gastric cancer cells. Co-IP assays showed that endogenous DUSP6 precipitated with pERK1/2 and, in contrast, endogenous pERK1/2 also precipitated with DUSP6 in cells under physiological conditions (Fig. 4G). Immunofluorescence assays demonstrated that pERK1/2 and DUSP6 colocalized with each other in gastric cancer cells, especially in cytoplasm (Fig. 4H; Supplementary Fig. S5D–S5E). Interestingly, although circPDIA4 did not impact endogenous DUSP6 expression (Supplementary Fig. S5F), overexpressed circPDIA4 evidently attenuated interactions between DUSP6 and ERK1/2 in cells (Fig. 4I). On the contrary, silencing of circPDIA4 markedly strengthened interactions between DUSP6 and ERK1/2 in gastric cancer cells (Fig. 4J). Silencing of DUSP6 enhanced invasion of circPDIA4-OE cells (Supplementary Fig. S5G–S5H). Hyperactivation of the ERK1/2 phosphorylation has been associated with disease progression and resistance to chemotherapeutics drugs in gastric cancer. MK-8353 is a highly potent, orally bioavailable, dual-specificity ERK1/2 inhibitor (ERKi; refs. 34, 35). Intriguingly, silencing of circPDIA4 expression strikingly reinforced the antineoplastic activities of MK-8353 (P < 0.01; Fig. 4K). Ectopic circPDIA4 expression notably promoted cell viability of gastric cancer cells treated with MK-8353 (P < 0.05; Supplementary Fig. S6A–S6B). Similar results were observed when another ERKi, ulixertinib, was used to treat cells (Supplementary Fig. S6C–S6F). Consistently, silenced circPDIA4 promoted the suppression of cell viability induced by silencing of MAPK1 expression (Supplementary Fig. S7). These results demonstrate that elevated circPDIA4 expression in gastric cancer could weaken interactions between DUSP6 and pERK1/2, enhance ERK1/2 phosphorylation, and, thus, reduce ERKi sensitivities.

circPDIA4 facilitated oncogenic circRNAs formation depending on the RNA helicase DHX9

We then investigated the molecular consequences of the interaction between circPDIA4 and DHX9 in nucleus. As shown in Fig. 5A, the expression levels of DHX9 protein were not disturbed when circPDIA4 was overexpressed or silenced in gastric cancer cells. In addition, circPDIA4 did not impact the nucleoplasm distribution ratio of DHX9 in cells (Fig. 5B). It has been reported that Alu-mediated circRNA biogenesis could be repressed by DHX9 (23). To assess whether production of certain circRNAs is regulated by circPDIA4, we profiled genome-wide circRNA expression in MKN45 cells through circRNAseq. Importantly, circPDIA4 induced a robust increase in levels of 785 circRNAs with circ_0001610 as the most upregulated one (circPDIA4 vs. Vector: log2fold-change>1, P < 0.05; Fig. 5C), suggesting that circPDIA4 might participate in regulating nuclear RNA resolvase functions of DHX9. Therefore, we generated two HA-tagged DHX9 expression constructs with the two Double-stranded RNA-binding motifs (DSRM; Del_329–1283aa) or without these DSRMs (Del_1–329aa; Fig. 5D). Interestingly, RIP assays demonstrated that wild-type DHX9 and Del_329–1283aa DHX9 could significantly enrich circPDIA4 in RNA–protein complexes precipitated with antibody against HA in HGC27 and MKN45 cells (all P < 0.001); whereas Del_1–329aa DHX9 could not enrich endogenous circPDIA4 (Fig. 5E). We then successfully validated the increased abundance of five most significantly upregulated candidate circRNAs (circ_0001610, circ_0005035, circ_0007052, circ_0002968, and circ_0003179) in gastric cancer cells with stably forced expression of circPDIA4 (Fig. 5F). Interestingly, we confirmed that circPDIA4-mediated upregulation of circRNAs, such as circ_0001610, depends on DHX9 protein (Fig. 5G; Supplementary Fig. S8A). Silenced circ_0001610 did not impact levels of total ERK1/2 protein or pERK1/2 in cells (Supplementary Fig. S8B). DHX9 did not change expression of PDIA4 or circPDIA4 in cells (Supplementary Fig. S8C–S8D). These results elucidate that circPDIA4 can competitively bind the RNA-binding motifs of DHX9 as a decoy, repress interactions between DHX9 and its target RNAs, and, thus, boost biogenesis of DHX9-controlled circRNAs.

Figure 5.

circPDIA4 competitively combined with DHX9 as a decoy to promote circRNA biogenesis. A, Western blot analyses of DHX9 protein levels in HGC-27 and MKN-45 cells with stabilized silencing of circPDIA4 (top) or overexpressed circPDIA4 (bottom). B, Western blot analyses of DHX9 protein levels in the cytoplasm and nucleus of HGC-27 and MKN-45 cells with stabilized silencing of circPDIA4 (left) or overexpressed circPDIA4 (right). C, Volcano diagram of differentially expressed circRNAs between MKN-45 cells with stabilized overexpressed circPDIA4 and vector. Differentially expressed circRNAs with q value of <0.05 and |log2(FC)|> 1 were considered significant. The top five upregulated circRNAs have dark spots with their names. D, Schematic structures showing domains of DHX9 protein and DHX9 truncations. Western blot analyses with anti-HA antibody of HGC-27 or MKN-45 cells transfected with HA-tagged plasmids encoding full-length DHX9 or truncated DHX9. E, RIP assays showed relative enrichment of circPDIA4 associated with full-length DHX9 or truncated DHX9 relative to an input control. IgG served as the control. F, RT-qRCR assays verified the levels of the top five upregulated circRNAs in circPDIA4-OE HGC-27 and MKN-45 cells. G, Silencing of DHX9 (siDHX9–1 or siDHX9–2) enhanced circ_0001610 biogenesis, which cannot be rescued by overexpression of circPDIA4 in HGC-27 or MKN-45 cells. Data represent mean ± SD. H and I, Knockdown of circ_0001610 partly reduced migration abilities of HGC-27 and MKN-45 cells that stably overexpressed circPDIA4. Data represent mean ± SD. J, The expression levels of circ_0001610 in cells after silencing with two siRNAs (si1610–1 or si1610–2). Data represent mean ± SD. The P values were determined by a two-tailed unpaired Student t test. ns, not significant; ***, P < 0.001.

Figure 5.

circPDIA4 competitively combined with DHX9 as a decoy to promote circRNA biogenesis. A, Western blot analyses of DHX9 protein levels in HGC-27 and MKN-45 cells with stabilized silencing of circPDIA4 (top) or overexpressed circPDIA4 (bottom). B, Western blot analyses of DHX9 protein levels in the cytoplasm and nucleus of HGC-27 and MKN-45 cells with stabilized silencing of circPDIA4 (left) or overexpressed circPDIA4 (right). C, Volcano diagram of differentially expressed circRNAs between MKN-45 cells with stabilized overexpressed circPDIA4 and vector. Differentially expressed circRNAs with q value of <0.05 and |log2(FC)|> 1 were considered significant. The top five upregulated circRNAs have dark spots with their names. D, Schematic structures showing domains of DHX9 protein and DHX9 truncations. Western blot analyses with anti-HA antibody of HGC-27 or MKN-45 cells transfected with HA-tagged plasmids encoding full-length DHX9 or truncated DHX9. E, RIP assays showed relative enrichment of circPDIA4 associated with full-length DHX9 or truncated DHX9 relative to an input control. IgG served as the control. F, RT-qRCR assays verified the levels of the top five upregulated circRNAs in circPDIA4-OE HGC-27 and MKN-45 cells. G, Silencing of DHX9 (siDHX9–1 or siDHX9–2) enhanced circ_0001610 biogenesis, which cannot be rescued by overexpression of circPDIA4 in HGC-27 or MKN-45 cells. Data represent mean ± SD. H and I, Knockdown of circ_0001610 partly reduced migration abilities of HGC-27 and MKN-45 cells that stably overexpressed circPDIA4. Data represent mean ± SD. J, The expression levels of circ_0001610 in cells after silencing with two siRNAs (si1610–1 or si1610–2). Data represent mean ± SD. The P values were determined by a two-tailed unpaired Student t test. ns, not significant; ***, P < 0.001.

Close modal

To determine whether circPDIA4-regulated circRNAs have any effects on invasive activities of gastric cancer cells, we performed rescue assays using cells with stable circPDIA4 overexpression. We observed a significant increase of invasive activity for cells with ectopic circPDIA4 (both P < 0.001), which could be rescued by silencing of circ_0001610 with siRNAs (all P < 0.001; Fig. 5HJ). Importantly, circ_0001610 did not impact expression of its host gene, TNFRSF21, in cells (Supplementary Fig. S8E). These results suggest the oncogene nature of circ_0001610 in gastric cancer. We further detected expression of the five candidate circRNAs in paired gastric cancer and normal tissues of Discovery cohort and Validation cohort. We observed evident upregulation of these circRNAs in gastric cancer specimens compared with normal tissues of both cohorts (all P < 0.001; Fig. 6A and B). Notably, we found significantly positive expression correlations between circPDIA4 and circ_0001610, circ_0005035, circ_0007052, circ_0002968, or circ_0003179 in either malignant tissues or normal gastric tissues in both cohorts (Fig. 6CF). Collectively, these results suggested that circPDIA4-controlled biogenesis of oncogenic circRNAs may underline mechanisms of how circPDIA4 drives invasive and metastatic processes of gastric cancer cells.

Figure 6.

The expression of the top five upregulated circRNAs in tissue specimens. A and B, RT-qPCR assays show the expression levels of circ_0001610, circ_0005035, circ_0007052, circ_0002968, or circ_0003179 (normalized to β-actin) in gastric cancer and paired normal tissues in Discovery cohort (A) and Validation cohort (B). Data represent mean ± SD. The P value was determined by a two-tailed paired Student t test. ***, P < 0.001. C–F, Expression correlations between circPDIA4 and circ_0001610, circ_0005035, circ_0007052, circ_0002968, or circ_0003179 in gastric cancer or normal tissues of Discovery cohort and Validation cohort. circRNA expression levels were determined by RT-qPCR, with β-actin as the control. Statistical analyses were performed with Pearson correlation analyses.

Figure 6.

The expression of the top five upregulated circRNAs in tissue specimens. A and B, RT-qPCR assays show the expression levels of circ_0001610, circ_0005035, circ_0007052, circ_0002968, or circ_0003179 (normalized to β-actin) in gastric cancer and paired normal tissues in Discovery cohort (A) and Validation cohort (B). Data represent mean ± SD. The P value was determined by a two-tailed paired Student t test. ***, P < 0.001. C–F, Expression correlations between circPDIA4 and circ_0001610, circ_0005035, circ_0007052, circ_0002968, or circ_0003179 in gastric cancer or normal tissues of Discovery cohort and Validation cohort. circRNA expression levels were determined by RT-qPCR, with β-actin as the control. Statistical analyses were performed with Pearson correlation analyses.

Close modal

CircRNAs function as a new layer of regulating tumorigenesis and aggressiveness of cancer cells. In this study, we found that circPDIA4 was highly expressed in gastric cancer and associated with poor prognosis of patients. QKI controlled biogenesis of circPDIA4 via backsplicing of exons 3 and 4 of PDIA4 RNA in gastric cancer. circPDIA4 could promote gastric cancer cell migration, invasion, and metastasis in vitro and in vivo. Mechanistically, circPDIA4 bound to ERK1/2 in cytoplasm and sustained ERK1/2 activation by preventing DUSP6-mediated dephosphorylation of pERK. Notably, circPDIA4 depletion significantly enhanced ERKi drug sensitivities in gastric cancer. More importantly, we found that circPDIA4 interacted with DHX9 in nucleus, repressed its functions, and stimulated expression of multiple oncogenic circRNAs, which, thereby, promoted gastric cancer progression. These results reveal a novel role of circPDIA4 in circRNA biogenesis as a decoy, ERKi resistance and gastric cancer progression (Fig. 7).

Figure 7.

Graphical representation of the regulation and functions of circPDIA4 in gastric cancer. QKI controlled biogenesis of circPDIA4 via backsplicing of exons 3 and 4 of PDIA4 RNA in gastric cancer. circPDIA4 could promote gastric cancer cell migration, invasion, and metastasis in vitro and in vivo. Mechanistically, circPDIA4 bound to ERK1/2 and sustained ERK1/2 activation by preventing DUSP6-mediated dephosphorylation of pERK in cytoplasm. Notably, we found that circPDIA4 interacted with DHX9 in nucleus, repressed its functions, and induced elevated expression of multiple oncogenic circRNAs, which, thereby, promoted gastric cancer progression.

Figure 7.

Graphical representation of the regulation and functions of circPDIA4 in gastric cancer. QKI controlled biogenesis of circPDIA4 via backsplicing of exons 3 and 4 of PDIA4 RNA in gastric cancer. circPDIA4 could promote gastric cancer cell migration, invasion, and metastasis in vitro and in vivo. Mechanistically, circPDIA4 bound to ERK1/2 and sustained ERK1/2 activation by preventing DUSP6-mediated dephosphorylation of pERK in cytoplasm. Notably, we found that circPDIA4 interacted with DHX9 in nucleus, repressed its functions, and induced elevated expression of multiple oncogenic circRNAs, which, thereby, promoted gastric cancer progression.

Close modal

It has been found that QKI regulates formation of a large number of circRNAs (18). As a dimer, QKI is capable of binding two well separated regions with QKI-binding motif(s) of a single RNA molecule, bringing the circle-forming exons into close proximity and promoting circRNA biogenesis (18). For example, QKI binds upstream and downstream of the circRNA-forming exons and/or introns in SMARCA5, ZEB1, and NDUFB2 RNA to promote circRNA formation in lung cancer, prostate cancer, and hepatocellular carcinoma (18, 36–38). However, it remains unknown how QKI contributes to circRNA biogenesis in gastric cancer. Indeed, we disclosed that QKI binds to introns flanking the circPDIA4-forming exons of PDIA4 pre-mRNA to promote circPDIA4 formation. Our data develop the knowledge about the effects of QKI-controlled circRNA on gastric cancer development.

ERK1 and ERK2 are two main members of MAPKs and constitutive activation of ERK1/2 occurs in cancers (32, 33). The physiological outcomes of MAPK signaling depend on both the magnitude and duration of kinase activation whose regulation is diverse and complex. Protein phosphatases are a class of proteins currently known to play an important role in the negative regulation of MAPK signaling (32, 33). In particularly, cytoplasmic phosphatase DUSP6 preferentially inactivates ERK1/2 (32, 33). Importantly, we observed direct interactions between circPDIA4 and ERK1/2 in gastric cancer cells, which leading to reduced binding of DUSP6 with ERK1/2, elevated levels of pERK1/2, activated MAPK signaling and gastric cancer pathogenesis. Multiple ERKi to suppress aberrant activation of ERK1/2 and malignancies have been successfully developed (34, 35). In line with the role of circPDIA4 in pERK1/2 maintenance, we first report that circPDIA4 KD also helped to overcome resistance of gastric cancer to ERKi. Our findings provide a previously unrecognized mechanism for ERK1/2 phosphorylation regulation and novel evidences for circRNAs participating in regulating protein posttranslational modifications, especially their implications for protein phosphorylation (39).

Presence of Alu complementary repeats in the introns flanking the downstream splice donor and upstream splice acceptor sites may bring these splice sites into close proximity and lead to backsplicing, which eventually generating a covalently closed circRNA (22–24). DHX9 represses Alu-mediated circRNA biogenesis by destabilizing double-stranded RNA structures formed as a result of the interactions between the Alu transposable elements. However, the functions and underlying mechanisms of circRNA in regulating circRNA formation via DHX9 remain elusive. We for the first time demonstrated that circPDIA4 could enhance biogenesis of other circRNAs via acting as an oncogenic decoy that prevents DHX9 binding to double-stranded RNA with Alu elements, thus antagonizing DHX9-induced suppression of circRNA formation. Our findings broaden the understanding of mechanisms by which circRNAs could be generated.

In summary, we provide clear evidences that circPDIA4 is a novel oncogene in gastric cancer, with markedly upregulated expression in malignant tissues and associated with poor prognosis of patients. On one hand, circPDIA4 acts through regulation of interactions between DUSP6 and ERK1/2, declined dephosphorylation of pERK1/2, activated MAPK signaling, and reduced ERKi sensitivities. On the other hand, circPDIA4 cooperates with DHX9 to promote progression of gastric cancer through facilitating biogenesis of other oncogenic circRNAs, though no such role of circRNAs has been reported during human carcinogenesis. Our findings extend the present understanding of circRNAs and imply the clinical potential of circPDIA4 as novel therapeutic targets for gastric cancer.

N. Zhang reports grants from National Natural Science Foundation of China and Natural Science Foundation of Shandong Province during the conduct of the study. M. Yang reports grants from National Natural Science Foundation of China, Natural Science Foundation of Shandong Province, Major Scientific and Technological Innovation Project of Shandong Province, Taishan Scholars Program of Shandong Province, and Program of Science and Technology for the youth innovation team in University of Shandong Province during the conduct of the study. No disclosures were reported by the other authors.

Y. Shen: Resources, data curation, formal analysis, validation, investigation, writing–original draft. N. Zhang: Resources, software, funding acquisition, validation, visualization, methodology, writing–review and editing. J. Chai: Resources. T. Wang: Resources, software, validation, methodology. C. Ma: Resources, data curation, software, methodology. L. Han: Resources. M. Yang: Conceptualization, resources, software, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

This work was supported by the National Natural Science Foundation of China (82103291, 31871306, and 82173070); Natural Science Foundation of Shandong Province (ZR2021LZL004 and ZR202102250889); Major Scientific and Technological Innovation Project of Shandong Province (2021ZDSYS04); Taishan Scholars Program of Shandong Province (tsqn20161060); Program of Science and Technology for the youth innovation team in universities of Shandong Province (2020KJL001).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Sung
H
,
Ferlay
J
,
Siegel
R
,
Laversanne
M
,
Soerjomataram
I
,
Jemal
A
, et al
.
Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries
.
CA Cancer J Clin
2021
;
71
:
209
49
.
2.
Plummer
M
,
Franceschi
S
,
Vignat
J
,
Forman
D
,
de Martel
C
.
Global burden of gastric cancer attributable to Helicobacter Pylori
.
Int J Cancer
2015
;
136
:
487
90
.
3.
Hooi
J
,
Lai
W
,
Ng
W
,
Suen
M
,
Underwood
F
,
Tanyingoh
D
, et al
.
Global prevalence of helicobacter pylori infection: systematic review and meta-analysis
.
Gastroenterology
2017
;
153
:
420
9
.
4.
Smyth
E
,
Nilsson
M
,
Grabsch
HI
,
Van Grieken
NC
,
Lordick
F
.
Gastric cancer
.
Lancet
2020
;
396
:
635
48
.
5.
Kristensen
L
,
Andersen
M
,
Stagsted
L
,
Ebbesen
K
,
Hansen
T
,
Kjems
J
.
The biogenesis, biology, and characterization of circular RNAs
.
Nat Rev Genet
2019
;
20
:
675
91
.
6.
Patop
I
,
Wüst
S
,
Kadener
S
.
Past, present, and future of circRNAs
.
EMBO J
2019
;
38
:
e100836
.
7.
Kristensen
L
,
Jakobsen
T
,
Hager
H
,
Kjems
J
.
The emerging roles of circRNAs in cancer and oncology
.
Nat Rev Clin Oncol
2022
;
19
:
188
206
.
8.
Li
F
,
Yang
Q
,
He
A
,
Yang
B
.
Circular RNAs in cancer: limitations in functional studies and diagnostic potential
.
Semin Cancer Biol
2021
;
75
:
49
61
.
9.
Wang
X
,
Li
J
,
Bian
X
,
Wu
C
,
Hua
J
,
Chang
S
, et al
.
CircURI1 interacts with hnRNPM to inhibit metastasis by modulating alternative splicing in gastric cancer
.
Proc Natl Acad Sci USA
2021
;
118
:
e2012881118
.
10.
Wei
L
,
Sun
J
,
Zhang
N
,
Zheng
Y
,
Wang
X
,
Lv
L
, et al
.
Noncoding RNAs in gastric cancer: implications for drug resistance
.
Mol Cancer
2020
;
19
:
62
.
11.
Dragomir
M
,
Kopetz
S
,
Ajani
J
,
Calin
G
.
Non-coding RNAs in GI cancers: from cancer hallmarks to clinical utility
.
Gut
2020
;
69
:
748
63
.
12.
Liu
M
,
Yang
J
,
Xu
B
,
Zhang
X
.
Tumor metastasis: mechanistic insights and therapeutic interventions
.
MedComm
2021
;
2
:
587
617
.
13.
Chen
R
,
Chen
X
,
Xia
L
,
Zhang
J
,
Pan
Z
,
Ma
X
, et al
.
N6-methyladenosine modification of circNSUN2 facilitates cytoplasmic export and stabilizes HMGA2 to promote colorectal liver metastasis
.
Nat Commun
2019
;
10
:
4695
.
14.
Ashwal-Fluss
R
,
Meyer
M
,
Pamudurti
N
,
Ivanov
A
,
Bartok
O
,
Hanan
M
, et al
.
circRNA biogenesis competes with pre-mRNA splicing
.
Mol Cell
2014
;
56
:
55
66
.
15.
Starke
S
,
Jost
I
,
Rossbach
O
,
Schneider
T
,
Schreiner
S
,
Hung
L
, et al
.
Exon circularization requires canonical splice signals
.
Cell Rep
2015
;
10
:
103
11
.
16.
Zhang
X
,
Wang
H
,
Zhang
Y
,
Lu
X
,
Chen
L
,
Yang
L
.
Complementary sequence-mediated exon circularization
.
Cell
2014
;
159
:
134
47
.
17.
Liang
D
,
Wilusz
J
.
Short intronic repeat sequences facilitate circular RNA production
.
Genes Dev
2014
;
28
:
2233
47
.
18.
Conn
S
,
Pillman
K
,
Toubia
J
,
Conn
V
,
Salmanidis
M
,
Phillips
C
, et al
.
The RNA-binding protein quaking regulates formation of circRNAs
.
Cell
2015
;
160
:
1125
34
.
19.
Khan
M
,
Reckman
Y
,
Aufiero
S
,
van den Hoogenhof
M
,
van der Made
I
,
Beqqali
A
, et al
.
RBM20 regulates circular RNA production from the titin gene
.
Circ Res
2016
;
119
:
996
1003
.
20.
Fei
T
,
Chen
Y
,
Xiao
T
,
Li
W
,
Cato
L
,
Zhang
P
, et al
.
Genome-wide CRISPR screen identifies HNRNPL as a prostate cancer dependency regulating RNA splicing
.
Proc Natl Acad Sci USA
2017
;
114
:
E5207
E15
.
21.
Errichelli
L
,
Dini Modigliani
S
,
Laneve
P
,
Colantoni
A
,
Legnini
I
,
Capauto
D
, et al
.
FUS affects circular RNA expression in murine embryonic stem cell–derived motor neurons
.
Nat Commun
2017
;
8
:
14741
.
22.
Rybak-Wolf
A
,
Stottmeister
C
,
Glažar
P
,
Jens
M
,
Pino
N
,
Giusti
S
, et al
.
Circular RNAs in the mammalian brain are highly abundant, conserved, and dynamically expressed
.
Mol Cell
2015
;
58
:
870
85
.
23.
Aktaş
T
,
Avşar Ilık
İ
,
Maticzka
D
,
Bhardwaj
V
,
Pessoa Rodrigues
C
,
Mittler
G
, et al
.
DHX9 suppresses RNA processing defects originating from the Alu invasion of the human genome
.
Nature
2017
;
544
:
115
9
.
24.
Ivanov
A
,
Memczak
S
,
Wyler
E
,
Torti
F
,
Porath
H
,
Orejuela
M
, et al
.
Analysis of intron sequences reveals hallmarks of circular RNA biogenesis in animals
.
Cell Rep
2015
;
10
:
170
7
.
25.
Zheng
Y
,
Lei
T
,
Jin
G
,
Guo
H
,
Zhang
N
,
Chai
J
, et al
.
LncPSCA in the 8q24.3 risk locus drives gastric cancer through destabilizing DDX5
.
EMBO Rep
2021
;
22
:
e52707
.
26.
Zhang
N
,
Li
Y
,
Xie
M
,
Song
Y
,
Liu
J
,
Lei
T
, et al
.
DACT2 modulated by TFAP2A-mediated allelic transcription promotes EGFR–TKIs efficiency in advanced lung adenocarcinoma
.
Biochem Pharmacol
2020
;
172
:
113772
.
27.
Yuan
J
,
Song
Y
,
Pan
W
,
Li
Y
,
Xu
Y
,
Xie
M
, et al
.
LncRNA SLC26A4-AS1 suppresses the MRN complex-mediated DNA repair signaling and thyroid cancer metastasis by destabilizing DDX5
.
Oncogene
2020
;
39
:
6664
76
.
28.
Xia
S
,
Feng
J
,
Chen
K
,
Ma
Y
,
Gong
J
,
Cai
F
, et al
.
CSCD: a database for cancer-specific circular RNAs
.
Nucleic Acids Res
2018
;
46
:
D925
D9
.
29.
Paz
I
,
Kosti
I
,
Ares
M
,
Cline
M
,
Mandel-Gutfreund
Y
.
RBPmap: a web server for mapping binding sites of RNA-binding proteins
.
Nucleic Acids Res
2014
;
42
:
W361
7
.
30.
Wang
G
,
Hu
N
,
Yang
H
,
Wang
L
,
Su
H
,
Wang
C
, et al
.
Comparison of global gene expression of gastric cardia and noncardia cancers from a high-risk population in china
.
PLoS ONE
2013
;
8
:
e63826
.
31.
Li
W
,
Hu
N
,
Burton
V
,
Yang
H
,
Su
H
,
Conway
C
, et al
.
PLCE1 mRNA and protein expression and survival of patients with esophageal squamous cell carcinoma and gastric adenocarcinoma
.
Cancer Epidemiol
2014
;
23
:
1579
88
.
32.
Owens
D
,
Keyse
S
.
Differential regulation of MAP kinase signalling by dual-specificity protein phosphatases
.
Oncogene
2007
;
26
:
3203
13
.
33.
Courtney
T
,
Deiters
A
.
Optical control of protein phosphatase function
.
Nat Commun
2019
;
10
:
4384
.
34.
Moschos
S
,
Sullivan
R
,
Hwu
W
,
Ramanathan
R
,
Adjei
A
,
Fong
P
, et al
.
Development of MK-8353, an orally administered ERK1/2 inhibitor, in patients with advanced solid tumors
.
JCI insight
2018
;
3
:
e92352
.
35.
Boga
S
,
Deng
Y
,
Zhu
L
,
Nan
Y
,
Cooper
A
,
Shipps
G
, et al
.
MK-8353: discovery of an orally bioavailable dual mechanism ERK inhibitor for oncology
.
ACS Med Chem Lett
2018
;
9
:
761
7
.
36.
Han
J
,
Meng
J
,
Chen
S
,
Wang
X
,
Yin
S
,
Zhang
Q
, et al
.
YY1 complex promotes quaking expression via super-enhancer binding during EMT of hepatocellular carcinoma
.
Cancer Res
2019
;
79
:
1451
64
.
37.
Chen
D
,
Chou
F
,
Chen
Y
,
Tian
H
,
Wang
Y
,
You
B
, et al
.
Targeting the radiation-induced TR4 nuclear receptor-mediated QKI/circZEB1/miR-141–3p/ZEB1 signaling increases prostate cancer radiosensitivity
.
Cancer Lett
2020
;
495
:
100
11
.
38.
Li
B
,
Zhu
L
,
Lu
C
,
Wang
C
,
Wang
H
,
Jin
H
, et al
.
circNDUFB2 inhibits non–small cell lung cancer progression via destabilizing IGF2BPs and activating anti-tumor immunity
.
Nat Commun
2021
;
12
:
295
.
39.
Su
T
,
Wang
T
,
Zhang
N
,
Shen
Y
,
Li
W
,
Xing
H
, et al
.
Long non-coding RNAs in gastrointestinal cancers: implications for protein phosphorylation
.
Biochem Pharmacol
2022
;
197
:
114907
.

Supplementary data