Circular RNA circCLASP2 promotes nasopharyngeal carcinoma progression through binding to DHX9 to enhance PCMT1 translation

Abstract

Background

Circular RNAs (circRNAs), characterized by their covalently closed-loop structures, constitute a distinct class of non-coding RNAs. They play pivotal regulatory roles within cells and are intricately associated with the progression of malignant tumors. However, their roles and the underlying mechanisms in nasopharyngeal carcinoma (NPC) progression have yet to be fully uncovered and comprehensively understood.

Methods

Employing RNA sequencing technology, high-abundance circular RNAs in NPC were identified. Expression analysis of circCLASP2 in NPC tissues was conducted using quantitative real-time polymerase chain reaction (qRT-PCR) and in situ hybridization experiments. Through in vitro and in vivo functional assays, the influence of circCLASP2 on the proliferation and metastasis of NPC was investigated. LC–MS/MS technology analyzed the binding partners of circCLASP2, its differentially regulated targets, and the associated proteins of PCMT1. Interactions among circCLASP2, DHX9 protein, and PCMT1 mRNA were elucidated through RNA immunoprecipitation and RNA pull-down techniques. The effects of circCLASP2 and DHX9 on RNA G-quadruplex (rG4) structures and PCMT1 mRNA translation were explored through immunofluorescence (IF), ribosomal gradient separation, and dual-luciferase reporter assays. Immunoprecipitation (IP) revealed the downstream effector of the circCLASP2-DHX9-PCMT1 regulatory axis and Phalloidin staining confirmed its ultimate effect on the cytoskeleton. PDS treatment was applied for interventions in NPC, demonstrating potential therapeutic avenues.

Results

Our research revealed that circCLASP2, a novel circRNA that has not been reported in tumors, is upregulated in NPC and fosters cell proliferation and metastasis both in vitro and in vivo. Mechanistically, circCLASP2 acts as a molecular scaffold, facilitating the approximation of DHX9 to PCMT1 mRNA. DHX9 unwinds the inhibitory rG4 structure near the translation initiation site on PCMT1 mRNA, increasing PCMT1 expression. PCMT1 binds to and upregulates cytoskeleton-associated proteins, modulating cytoskeleton strength and dynamics and ultimately driving NPC cell proliferation and metastasis. In both in vitro and in vivo experiments, PDS significantly inhibits NPC growth and metastasis, showcasing promising therapeutic potential.

Conclusions

Our investigation pinpointed a circular RNA, circCLASP2, which is upregulated in NPC and augments cytoskeletal functions via the DHX9-PCMT1 axis, contributing to the malignancy progression of NPC. This pathway holds promise as a potential therapeutic target for NPC. Furthermore, these molecules could also serve as biomarkers for adjunct diagnosis and prognosis assessment in NPC.

Background

Nasopharyngeal carcinoma (NPC) is a malignancy arising from the nasopharyngeal epithelium, categorized under head and neck cancers. Its incidence displays regional discrepancies, with a pronounced prevalence in Southeast Asia, especially in Southern China [1]. NPC typically manifests insidious [2], characterized by high invasiveness and metastatic tendencies, notably involving lymph node metastasis [3]. Currently, radiotherapy and chemotherapy serve as primary treatment approaches [4, 5], yet the prognosis of patients with advanced NPC remains generally unfavorable [6, 7]. The development of NPC is a complex, multi-step process [8,9,10,11], involving numerous genes and pathways [12,13,14]. While the inactivation of tumor suppressor genes and activation of oncogenes are recognized as pivotal in the development of NPC, the precise molecular mechanisms remain incompletely understood. Consequently, elucidating the mechanisms underlying NPC development and identifying more effective therapeutic targets or drugs is urgent and imperative. In recent years, there has been significant interest in exploring the role of circular RNA in NPC.

Circular RNAs (circRNAs) have emerged as a prominent focus of research in recent years. They represent a class of non-coding RNA characterized by a covalently closed-loop structure formed through the back-splicing of pre-mRNA at the 3' and 5' ends [15], exhibiting notable stability [16, 17]. With the advancement of high-throughput sequencing technologies, tens of thousands of circRNAs have been identified. Growing studies have unveiled the critical roles circRNAs play in various disease processes, functioning as molecular sponges, binding to proteins, and even encoding small peptides [18,19,20,21,22,23,24,25,26]. Previous studies have suggested that circRNAs regulate cell proliferation and migration during the malignant progression of NPC [27,28,29,30,31], correlating with the prognosis and TNM staging of NPC patients [32,33,34], thereby holding potential as early diagnostic and therapeutic targets for NPC. Nonetheless, the functional roles of many circRNAs in NPC remain elusive, warranting further exploration.

Through RNA sequencing analysis of NPC cell lines with pronounced metastatic capability, we identified a panel of differentially expressed circRNAs in NPC [35]. In our previous work, we have highlighted several circRNAs, such as circSETD3 [36], circRILPL1 [37], circBART2.2 [38], and circRNF13 [33], which contribute to the malignant progression of NPC. Notably, among our data, circCLASP2, a novel circRNA that has not been reported in tumors, demonstrated significantly elevated expression in NPC, with its function yet to be elucidated.

DExH-Box Helicase 9 (DHX9) is a well-recognized RNA helicase characterized by high sequence conservation [39, 40]. DHX9 modulates the structure of various RNAs, thereby regulating the RNA transcription or translation processes [41,42,43,44,45]. Protein-L-isoaspartate (D-aspartate) O-methyltransferase 1 (PCMT1), a member of the protein methyltransferase family, exhibits increased expression in various cancers and functions as an oncogene [46,47,48]. PCMT1 enhances the expression of its target proteins or assists in restoring the native structure and function of proteins [49,50,51].

In this study, we identified a circRNA, circCLASP2 (circBase ID: hsa_circ_0001280), which is upregulated in NPC. In vivo and in vitro experiments evaluated that circCLASP2 promotes the proliferation and metastasis of NPC cells. Further analysis demonstrated that circCLASP2 acted as a molecular scaffold, facilitating the interaction between DHX9 protein and PCMT1 mRNA. DHX9 enhanced PCMT1 mRNA translation by unwinding the RNA G-quadruplex (rG4) structure on its translation initiation site. Consequently, PCMT1 promoted the expression of cytoskeletal proteins such as α-actinin, β-actin, and CFL1, ultimately prompting the proliferation, migration, and invasion of NPC cells. Intervention with rG4-targeting drug PDS validated the proposed regulatory mechanism and demonstrated significant therapeutic effects on NPC.

Methods

NPC clinical samples

The fresh clinical tissues comprised 22 NPC cases of varying stages and 12 nasopharyngeal chronic inflammation tissues (Supplementary Table 1), from which tissue RNA was extracted for qRT-PCR analysis. Paraffin-embedded tissue section samples were divided into two cohorts: the first cohort included 96 NPC tissues and 42 adjacent non-tumor nasopharyngeal epithelial (NPE) tissues with clinical prognosis information, utilized for in situ hybridization (ISH); the second cohort comprised 34 NPC tissues and 7 adjacent non-tumor NPE tissues, used for ISH and immunohistochemistry (IHC) (Supplementary Table 2). All NPC and non-NPC tissue samples were collected from the Hunan Cancer Hospital (Changsha, China) and underwent pathological diagnosis. This study was approved by the Ethics Committee of Central South University (Changsha, China), and informed consent was obtained from all participants.

Cell culture

The NPC cell lines (HNE2, CNE2, HONE1, C666-1) used in this project were maintained at the Cancer Research Institute, Central South University. All cells were cultured in RPMI-1640 medium (Life Technologies, NY, USA) containing 10% fetal bovine serum (FBS) (Gibco, MA, USA) at 37 °C in a humidified atmosphere with 5% CO₂ in a constant temperature incubator.

Plasmids, small interfering RNAs (siRNA), and transfection

The plasmid pcDNA3.1(+) circRNA mini vectors were graciously provided by Professor Yong Li from Baylor College of Medicine. NPC cell RNA underwent reverse transcription to generate cDNA, which served as the template for PCR amplification to obtain the full-length target fragment. This fragment was subsequently inserted into the empty plasmid to produce circCLASP2 and PCMT1 overexpression vectors. The DHX9-Flag overexpression vector was procured from Sino Biological, Inc. (Beijing, China) and validated by sequencing. Dual-luciferase reporter gene vectors and the deletion mutant vectors were procured from Beijing Tsingke Biotech Co., Ltd. (Beijing, China). The siRNAs targeting circCLASP2, DHX9, and PCMT1 were sourced from RiboBio Co., Ltd. (Guangzhou, China). Amplification primers for the target fragments in the overexpression vectors and siRNA sequences are listed in Supplementary Table 3. Plasmids were transfected using Lipofectamine 3000 (Life Technologies, NY, USA), while siRNAs were transfected using Hiperfect (Qiagen, NRW, Germany). PDS was diluted to 50 μM for cell treatment.

RNA extraction and qRT-PCR

After lysing NPC cells or tissues with TRIzol Reagent (Life Technologies, NY, USA), total RNA was extracted through isopropanol precipitation. The extracted total RNA was reverse-transcribed into cDNA using the HiScript cDNA Synthesis kit (Vazyme, Nanjing, China). Subsequently, qRT-PCR analysis was conducted using 2 × SYBR Green (Bimake, TX, USA) on the CFX96™ Real-Time PCR System to detect the expression of various RNA molecules, with GAPDH serving as an internal reference. The primers utilized for qRT-PCR are listed in Supplementary Table 3.

Wound healing assay

Cells were seeded and cultured in 6-well plates. Following transfection, the cells were incubated for 48 h. A sterile 10 μl pipette tip was utilized to create uniform scratches at the bottom of the wells. Images were captured at fixed positions at 0 h, 12 h, and 24 h (HNE2, CNE2, HONE1) or 0 h, 24 h, and 48 h (C666-1) to measure and record the width.

Transwell invasion assay

Matrigel (BD, Shanghai, China) was diluted and evenly applied to the bottom of the upper chamber of Transwell chambers (Millipore, MA, USA). Upon gel solidification, NPC cell suspension was added to the upper chamber, while medium containing 20% FBS was added to the lower chamber. Following a 48 to 60-h incubation period, cells that invaded through the Matrigel were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. Images were captured using an inverted phase contrast microscope, and cell counting was performed.

MTT assay

An MTT assay was conducted in a 96-well plate, with 800 (HNE2, CNE2, HONE1) or 4000 (C666-1) cells seeded per well. Once the cells adhered to the wells, MTT solution (Beyotime, Shanghai, China) was added to each well from day 0 to day 5. The plate was then incubated in the dark at 37 °C for 4 h. Subsequently, the liquid in wells was removed, and dimethyl sulfoxide (DMSO) was added to dissolve the formazan crystals. The plates were shaken steadily for 10 min, and the absorbance was measured at 490 nm using a microplate reader (Molecular Devices, CA, USA), followed by data analysis.

RNase R and actinomycin D treatment

RNA was extracted from NPC cells and incubated with 20 U/μl RNase R (RNR07250, Epicentre, USA) at 37 °C for 30 min, followed by deactivation of RNase R at 70 °C for 10 min. Subsequently, the stability of RNA was assessed using qRT-PCR.

Actinomycin D (Sigma, STL, USA) was added to the NPC cell culture medium at a final concentration of 2 µg/ml. Cells were treated for 0 h, 8 h, 16 h, and 24 h before collection, and RNA was extracted for qRT-PCR analysis.

 Fluorescence in situ hybridization (FISH)

NPC cells were seeded onto slides pre-placed at the bottom of 24-well plates. Upon cell adherence, the cells were fixed with pre-warmed 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. Subsequently, pre-hybridization solution was added and incubated at 37 °C for 2–4 h, followed by the addition of digoxigenin-labeled circCLASP2 specific probes (Sangon Biotech, Shanghai, China) and further overnight incubation at 37 °C. After multiple washes with appropriate concentration saline-sodium citrate (SSC) buffer, blocking solution was added, and the slides were blocked at 37 °C for 30 min. Biotinylated mouse anti-digoxin was added and incubated at 37 °C for 1 h. Following PBS washing, a red fluorescent anti-mouse secondary antibody (LIFE, NY, USA) was added and incubated in the dark at 37 °C for 1 h. Subsequently, DAPI solution (Invitrogen, CA, USA) was added for nuclear staining. The slides were sealed with anti-fluorescence quencher and photographed under a laser scanning confocal microscope (Leica, HE, Germany). The probes utilized for FISH are listed in Supplementary Table 3.

Animal experiments

All animals used (4-week-old female BALB/c nude mice) in the experiments were acquired from the Laboratory Animal Center of Central South University and were housed in a specific pathogen-free (SPF) environment. All animal experimental protocols were approved by the Institutional Animal Care and Use Committee of Central South University (Changsha, China).

Mice were randomly divided into four groups (n = 6). Each mouse received injections of 2 × 10⁶ CNE2 cells transfected with either empty vector (EV), circCLASP2 overexpression plasmid (OE), scrambled siRNA (negative control, NC), or siRNA target to circCLASP2 (si-circCLASP2), respectively. Tumor growth in the subcutaneous tumor xenograft model was monitored every 3 days. After 28 days of subcutaneous injection, the mice were sacrificed by cervical dislocation, and the subcutaneous tumor tissues were collected and weighed. The tumor tissues underwent gradient dehydration and paraffin embedding, and were preserved as sections. In the "footpad-lymph node" metastasis model, photographs of the injection sites on the mice's footpads were taken weekly. After 28 days of injection, the mice were sacrificed by cervical dislocation, and the footpad primary tumors, along with the ipsilateral popliteal lymph nodes, were collected for volume measurement and weighing. These tissues were then preserved as paraffin sections.

In both animal models mentioned above, PDS medication was administered at a dose of 7.5 mg/kg/day. The drug administration commenced 1–2 weeks after model establishment, with PDS being administered via tail vein injection for 5 consecutive days. Mice in the control group received an equivalent volume of saline through tail vein injection. Following euthanasia, the heart, liver, spleen, lungs, and kidneys were collected and processed for paraffin embedding.

Liquid chromatography-mass spectrometry (LC–MS/MS)

Total proteins were extracted and digested with proteases overnight. The resulting peptides were dissolved in 0.1% trifluoroacetic acid (TFA) and subjected to analysis by liquid chromatography-tandem mass spectrometry (LC–MS/MS) using an LTQ Orbitrap Velos Pro mass spectrometer (Thermo Scientific, Bremen, Germany) coupled with an Ultimate 3000 RSLC Nano system (Dionex, CA, USA). The proteins were identified using Proteome Discoverer 1.4 software (Thermo Fisher Scientific, MA, USA) and searched against the UniProt KB/Swiss-Prot database.

RNA pull-down

Biotin-labeled probes targeting the splicing sites of circCLASP2 were designed (Sangon Biotech, Shanghai, China). NPC cells were lysed with the RIP lysis buffer. After centrifugation, the supernatant was incubated with the circCLASP2 probe and Streptavidin Dynabeads (M-280, Invitrogen, USA) at 4 °C for 12–16 h. Following washing, the samples were resuspended in lysis buffer, and the circCLASP2-binding proteins were examined by Western blotting.

RNA immunoprecipitation (RIP)

The Magna RIPTM Kit (17–701, Millipore, USA) was utilized following the manufacturer's guidelines to analyze the interaction between protein and RNAs.

In situ hybridization (ISH)

Paraffin-embedded tissue sections underwent deparaffinization and rehydration, followed by inactivation of endogenous peroxidases using a 3% hydrogen peroxide. Digestion was carried out with freshly prepared pepsin (in 3% citric acid solution), followed by termination with 0.1 mol/L glycine. The sections were then re-fixed with 4% paraformaldehyde, incubated with pre-hybridization solution at 37 °C for 30 min, and hybridized with digoxigenin-labeled circCLASP2 probes (Sangon Biotech, Shanghai, China) at 37 °C for 12–16 h. Following washing with SSC solution, biotinylated mouse anti-digoxin, streptavidin-biotin complex (sABC), and biotinylated peroxidase were sequentially applied. After washing, the sections were incubated in DAB and hematoxylin for color development, sealed with neutral resin, observed, and photographed under a microscope.

Western blotting

Proteins were extracted using RIPA buffer (Beyotime, Shanghai, China) supplemented with a protease inhibitor cocktail (Roche Applied Sciences, BW, Germany). The samples were then subjected to electrophoresis at 80–120 V for 1.5–2 h on SDS-PAGE gels (8%−12%) to separate proteins of various molecular weights, which were subsequently transferred onto 0.2 μm PVDF membranes (Millipore, MA, USA). Following blocking with 5% non-fat milk for 1–2 h, the membranes were incubated with specific primary antibodies at 4 °C for 12–16 h. After washing with PBST or TBST, the membranes were incubated with corresponding secondary antibodies at room temperature for 2 h. Signals were detected using the ECL detection system (Millipore, MA, USA). The antibodies used are listed in Supplementary Table 4.

Immunoprecipitation (IP)

Primary antibodies were mixed with IP magnetic beads and incubated at room temperature with rotation for 2 h to obtain antibody-conjugated magnetic beads. Meanwhile, NPC cells were treated with GLB + lysis buffer and protease inhibitors, lysed on ice, and then centrifuged at low temperatures to obtain the protein lysate. The lysate was incubated with the antibody-conjugated magnetic beads at 4 °C for 12–16 h. Subsequently, the precipitated complexes were washed with pre-cooled GLB + lysis buffer. After washing, a protein loading buffer was added, and the samples were boiled for 5–10 min before Western blotting analysis. The antibodies used are listed in Supplementary Table 4.

Immunofluorescence (IF)

Cells were fixed with pre-warmed 4% paraformaldehyde, permeabilized with 0.25% Triton X-100, and subsequently blocked with 5% BSA. They were then incubated with primary antibodies overnight at 4 °C, followed by corresponding fluorescent secondary antibodies for 1 h at 37 °C. Cell nuclei were stained with DAPI, and images were captured using a laser scanning confocal microscope (Leica, HE, Germany). The antibodies used are listed in Supplementary Table 4.

Hematoxylin–Eosin staining (H&E)

Mouse tissue paraffin sections were heated at 65 °C for 2 h, dewaxed in xylene, and hydrated through a graded ethanol series. Cell nuclei were stained with hematoxylin solution (Biosharp, Anhui, China), followed by cytoplasm staining with eosin solution (Biosharp, Anhui, China). Finally, the sections were mounted with neutral balsam for preservation.

Immunohistochemistry (IHC)

Immunohistochemistry was conducted on NPC tissues and animal model tissues to analyze protein expression levels using the UltraSensitive™ SP (Mouse/Rabbit) Immunohistochemistry (IHC) kit (MXB, Fujian, China). Following antigen retrieval mediated by EDTA at high temperatures, sections were incubated with specific primary antibodies at 4 °C, followed by staining with DAB and hematoxylin. The tissue sections were then scored and statistically analyzed based on the intensity of staining and the proportion of positive cells. The antibodies used are listed in Supplementary Table 4.

Sucrose density gradient centrifugation for ribosomal fraction separation

Different concentrations of sucrose solutions were layered into centrifuge tubes to prepare a linear density gradient sucrose solution. NPC cells with approximately 80% confluence were incubated with cycloheximide (CHX) (MedChemExpress, Shanghai, China) at 37 °C and 5% CO₂ for 5 min to prevent ribosomal runoff. After washing with ice-cold PBS containing CHX, cells were scraped, centrifuged, and then resuspended in 500 μl of hypotonic buffer (containing 5 mM Tris-HCl, 2.5 mM MgCl₂, 1.5 mM KCl, and 1 × protease inhibitor cocktail without EDTA). Subsequently, 5 μl of 10 mg/ml CHX, 1 μl of 1 M DTT, 25 μl of 10% Triton X-100 (final concentration 0.5%), and 25 μl of 10% sodium deoxycholate (final concentration 0.5%) were added, followed by vortexing for 5 s. The mixture was then centrifuged at 16,000 × g for 7 min at 4 °C. A portion was saved as the input group, and the remaining 500 μl was layered on top of the sucrose gradient and centrifuged at 222,228 × g (36,000 rpm) for 2 h at 4 °C using an ultracentrifuge (Beckman Coulter, CA, USA) with SW41Ti rotor. The fractions were collected separately post-centrifugation, lysed with TRIzol Reagent to extract RNA from each fraction, which was then analyzed by qRT-PCR.

Dual-luciferase reporter assay

The Dual-Luciferase® Reporter Assay System (Promega, WI, USA) was utilized following the manufacturer's guidelines. Relative luciferase activity was determined by normalization to Renilla luciferase activity.

Statistical analysis

Statistical analysis was conducted using GraphPad Prism 9.5.1 software. Comparisons between the two groups were made using the Student's t-test (two-tailed), and for multiple comparisons, one-way ANOVA was employed to assess significant differences. The correlation between circCLASP2 and PCMT1 expression levels and NPC disease stage was statistically analyzed using Chi-square (and Fisher’s exact) tests. Comparison between the two groups in vivo was analyzed using the two-tailed, unpaired, nonparametric Mann–Whitney test. All data are presented as mean ± standard deviation (SD), and a P value < 0.05 was considered statistically significant. Significance levels were denoted as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Results

CircCLASP2 is highly expressed in NPC tissues and is associated with poor prognosis

To investigate the role of circular RNAs in the malignant progression of NPC, we conducted high-throughput and high-coverage RNA sequencing (RNA-seq) analysis of NPC cell lines in our previous work, identifying 6153 circRNAs [35]. Among the top 50 circRNAs with the highest expression abundance (Supplementary Table 5), circCLASP2 was not reported in tumors before and exhibited significantly upregulated in 22 NPC tissue samples compared to 12 nasopharyngeal chronic inflammation tissues (Fig. 1A). CircCLASP2, spanning 449 nucleotides, is spliced from exons 2 to 6 of the cytoplasmic linker associated protein 2 (CLASP2) gene (NM_015097.2). Its splicing site was confirmed by Sanger sequencing (Fig. 1B). ISH detected circCLASP2 expression in 96 NPC and 42 adjacent non-tumor nasopharyngeal epithelium (NPE) paraffin sections, confirming its upregulation in NPC tissues, with patients exhibiting high circCLASP2 expression having a poor prognosis (Fig. 1C-E). Based on the analysis of patients' clinical information, we found that the expression of circCLASP2 was related to the staging of NPC. The expression of circCLASP2 in the NPC tissues of T3-4 stage patients was higher than that in the tissues of T1-2 stage patients (Fig. S1A). Similarly, the expression of circCLASP2 in the tissues of N2-3 stage patients was significantly higher than that in the tissues of N0-1 stage patients. Moreover, the expression of circCLASP2 in the NPC tissues of patients with higher clinical staging was also higher (Fig. S1A). This further suggests that circCLASP2 is closely related to the malignant progression of NPC. RNase R digestion experiments confirmed that circCLASP2 was more resistant to enzyme digestion than CLASP2 mRNA (Fig. 1F, Fig. S1B). At the same time, treatment of NPC cells with actinomycin D demonstrated that circCLASP2 exhibited higher stability than linear CLASP2 mRNA (Fig. S1C). FISH assays revealed the presence of circCLASP2 in both the cytoplasm and nucleus of NPC cells (Fig. 1G). These findings suggest that circCLASP2 is a circular RNA not reported in tumors, stably and highly expressed in NPC, and is closely associated with the prognosis of NPC patients.

Fig. 1

CircCLASP2 is highly expressed in NPC and is associated with poor prognosis in patients. A The differential expression of circCLASP2 in 12 chronic rhinitis epithelial tissues (NPE) and 22 NPC tissues was detected by qRT-PCR, and statistically analyzed by Student's t test (two-tailed). B Schematic representation of circCLASP2. Sanger sequencing confirmed that exons 2–6 of CLASP2 mRNA were back spliced to form circCLASP2. Black arrows indicate the primer locations and orientations for detecting the circCLASP2 splice junction. C-E ISH assays revealed elevated expression of circCLASP2 in 96 NPC paraffin-embedded tissue sections compared to that in 42 NPEs and was associated with adverse prognosis in NPC patients. The ISH results were graded based on the depth of staining and the positivity rate. A comprehensive score greater than or equal to 5 was considered as high circCLASP2 expression, while a score less than 5 was categorized as low circCLASP2 expression. C Representative ISH images, left: scale bar = 100 μm; right: scale bar = 50 μm. D Statistical analysis of ISH scores. Statistical analysis was performed using the unpaired t test. E Kaplan–Meier survival analysis showed significantly reduced overall survival rates in NPC patients with high circCLASP2 expression compared to those with low circCLASP2 expression. F Relative expression levels of circCLASP2 and CLASP2 mRNA in NPC cells after treatment with RNase R. G FISH assays for detecting subcellular localization of circCLASP2 in NPC cells. Nuclei were stained with DAPI (blue), and circCLASP2 was identified with probes (red). Scale bar = 20 μm. Data were presented as mean ± SD. ***, P < 0.001; ****, P < 0.0001

CircCLASP2 promotes the proliferation and metastasis of NPC cells in vitro and in vivo

To elucidate the role of circCLASP2 in the malignant progression of NPC, we constructed a circCLASP2 overexpression vector (empty vector pcDNA3.1(+) was used as negative control) and designed two siRNAs targeting its splicing site (si-circCLASP2-1 and si-circCLASP2-2) to modulate circCLASP2 expression levels in cells. This enabled us to investigate how circCLASP2 influences the proliferation and metastasis of NPC. The qRT-PCR results confirmed successful overexpression and knockdown of circCLASP2 in NPC cells, while also evaluating CLASP2 mRNA expression levels. This validation shows that altering circCLASP2 expression does not impact its parent gene (Fig. S2A-C). MTT assays revealed that circCLASP2 overexpression significantly enhanced the proliferative capacity of NPC cells in vitro, whereas circCLASP2 knockdown resulted in reduced cell proliferation (Fig. 2A, Fig. S3A-B). Transwell invasion assays demonstrated that circCLASP2 overexpression increased the invasive potential of NPC cells in Matrigel, whereas circCLASP2 knockdown led to a notable reduction in invasion (Fig. 2B, Fig. S3C). Wound healing assays indicated that circCLASP2 enhanced the migratory capability of NPC cells, while its knockdown had the opposite effect (Fig. 2C, Fig. S3D). These functional experiments collectively demonstrate that circCLASP2 can promote the proliferation, invasion, and migration of NPC cells in vitro.

Fig. 2

CircCLASP2 promotes proliferation and metastasis of NPC cells in vitro and in vivo. A MTT assay determining the proliferation capability of HNE2 cells with overexpression or knockdown of circCLASP2. B Transwell invasion assay assessing the invasion capability of HNE2 cells with overexpression or knockdown of circCLASP2. Scale bar = 100 μm. The right panel shows statistical analysis. C Wound healing assay measuring the migration ability of HNE2 cells with overexpression or knockdown of circCLASP2, with images taken at the same field of view at 0 h and 24 h. Scale bar = 200 μm. The right panel shows statistical analysis. D-H CNE2 cells transfected with empty vector, circCLASP2 overexpression plasmid, siRNA negative control (NC), or si-circCLASP2 were injected subcutaneously or into the footpad of BALB/c mice (n = 6 per group), with each mouse receiving an injection of 2 × 10⁶ cells. D Subcutaneous tumors in mice from each group (left). Growth curve of volumes (middle) and statistical analysis of the weights (right) of subcutaneous tumors in nude mice. E Representative images of H&E, ISH, and IHC staining of subcutaneous tumor sections. Scale bar = 100 μm. F Representative images of the "footpad-lymph node" metastasis model showing the injection site in the foot and ipsilateral popliteal lymph node. G Popliteal lymph nodes from the "footpad-lymph node" metastasis model (left). Statistical analysis of volumes (middle) and weights (right) of lymph nodes in nude mice. H Representative images of H&E and CK-PAN staining of popliteal lymph node sections from the "footpad-lymph node" metastasis model, showing metastatic tumor cells. Top image scale bar = 200 μm; Bottom image scale bar = 100 μm. Data were presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001

To further assess the impact of circCLASP2 on NPC growth and metastasis in vivo, we established subcutaneous xenograft tumor models and "footpad-lymph node" metastasis models in female BALB/c mice. Results from the subcutaneous xenograft model revealed that, compared to the control group, circCLASP2 overexpression significantly augmented tumor volume and weight, whereas circCLASP2 knockdown notably decreased tumor size (Fig. 2D). ISH staining for circCLASP2 and IHC staining for Ki67 demonstrated a significant increase in Ki67 levels in tumor tissues of the circCLASP2 overexpression group. At the same time, they were reduced in the circCLASP2 knockdown group (Fig. 2E).

In the "footpad-lymph node" metastasis model, localized tumor growth was observed at the footpad injection site, with notably larger tumor volumes in the circCLASP2 overexpression group and apparent enlargement of the ipsilateral popliteal lymph nodes, indicating potential tumor metastasis to the lymph nodes. Conversely, the circCLASP2 knockdown group exhibited contrasting outcomes (Fig. 2F-G, Fig. S4A-B). Histological examination using H&E staining and IHC for CK-PAN revealed metastatic tumor tissues within the lymph nodes of mice in the circCLASP2 overexpression group, which showed a significantly greater extent than those in the control group. Conversely, the lymph nodes of mice in the circCLASP2 knockdown group exhibited fewer metastatic tumor cells compared to those in the control group (Fig. 2H). Altogether, these findings indicate that circCLASP2 promotes the proliferation and metastasis of NPC, both in vitro and in vivo.

CircCLASP2 binds to DHX9 and regulates its subcellular localization

Studies have indicated that circular RNA molecules might function by encoding small peptides [52]. Predictions made using the online tool circRNADb website (http://reprod.njmu.edu.cn/circrnadb) suggest the presence of potential IRES sequences and open reading frames (ORFs) on circCLASP2 (Fig. S5A). However, when this putative ORF was tagged with Flag and transfected into NPC cells (Fig. S5B-C), we were unable to detect any Flag-tagged proteins or peptides by Western blot, thereby ruling out the possibility of circCLASP2 encoding peptides (Fig. S5D).

Binding to proteins is another important way for circRNAs to exert functions [53]. To probe the specific mechanism underlying circCLASP2's promotion of NPC growth and metastasis, we identified 98 potential circCLASP2-binding proteins through RNA pull-down assays combined with LC–MS/MS (Supplementary Table 6). We compare silver staining results with mass spectrometry data to search for valuable target molecules. The result of silver staining of RNA pull-down samples revealed a significant differential band at 140 kDa; the molecular size of DHX9 matches the band and has a high abundance in the mass spectrometry data (Fig. S6A). This is a powerful molecule that caught our attention. Subsequent RNA pull-down assays and RIP experiments confirmed the interaction between circCLASP2 and DHX9 in NPC cells (Fig. 3A-B). Furthermore, FISH-IF experiments corroborated the co-localization of circCLASP2 and DHX9 in NPC cells (Fig. S6B).

Fig. 3

CircCLASP2 promotes PCMT1 expression via facilitating the interaction between DHX9 and PCMT1 mRNA in the cytoplasm. A Using a biotin-labeled circCLASP2 probe, RNA pull-down assays enriched circCLASP2 binding proteins in NPC cell lines HNE2, CNE2 and HONE1, with Western blotting showing the interaction between circCLASP2 and DHX9. GAPDH served as a negative control. B RIP assays using DHX9 antibody to detect the interaction between DHX9 and circCLASP2 in HNE2, CNE2, and HONE1. IgG was used as a negative control. C Western blotting assessing the abundance of DHX9 protein in the cytoplasm and nucleus of HNE2, CNE2, and HONE1 cells following overexpression or knockdown of circCLASP2. PARP and β-tubulin served as nuclear and cytoplasmic markers, respectively. D IF experiments examining the effect of circCLASP2 on the subcellular localization of DHX9. E Western blotting measuring PCMT1 protein levels in HNE2, CNE2, and HONE1 following overexpression or knockdown of circCLASP2. F RIP assays were conducted to detect the interaction between DHX9 and PCMT1 mRNA in HNE2, CNE2, and HONE1, using DHX9 antibody. IgG served as a negative control. G RIP assays using DHX9 antibody to assess the effects of overexpressing or knocking down circCLASP2 on the interaction between DHX9 and PCMT1 mRNA. H Western blotting measuring PCMT1 protein levels in HNE2, CNE2, and HONE1 following overexpression or knockdown of DHX9. I Western blotting assessing PCMT1 protein levels in HNE2, CNE2, and HONE1 after simultaneous overexpression of circCLASP2 and knockdown of DHX9. Data were presented as mean ± SD. *, P < 0.05; ***, P < 0.001; ****, P < 0.0001

Using bioinformatics tools catRAPID (http://s.tartaglialab.com/page/catrapid_group) and RNAfold (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi), we predicted three distinct stem-loop structures on circCLASP2 (nt 51–102, nt 130–181, and nt 201–252) that may potentially interact with the DHX9 protein (Fig. S7A-C). RNA pull-down and RIP analyses revealed that the deletion mutant DEL1, lacking the nt 51–102 sequence, failed to bind to DHX9 protein, while deletion mutants DEL2 and DEL3 still exhibited partially binding capacity, indicating that the nt 51–102 region is crucial for the interaction between circCLASP2 and DHX9 protein (Fig. S7D-E). Further deletions were made in three major components of the DHX9 protein (dsRBD, Helicase domain, OB/RGG) (Fig. S8A). RNA pull-down experiments showed that after the dsRBD domain was deleted, DHX9 failed to bind to circCLASP2, while the other two deletion mutants could still be pulled down by circCLASP2, indicating that the dsRBD is an important segment for the interaction between DHX9 and circCLASP2. Additionally, deletion of the Helicase domain affected the expression of DHX9, suggesting that this domain is essential for the stability and function of the DHX9 protein. (Fig. S8B).

To delve deeper into the regulatory relationship between circCLASP2 and DHX9, we overexpressed or knocked down circCLASP2 in cells and observed no significant alteration in DHX9 protein expression (Fig. S9A). Given DHX9's role as a nucleic acid helicase with diverse RNA-binding capabilities, we hypothesized its cytoplasmic activity might be pivotal within circCLASP2's regulatory pathway. To explore this, we fractionated nuclear and cytoplasmic proteins from NPC cells, assessing circCLASP2's impact on DHX9 subcellular localization. Remarkably, circCLASP2 overexpression correlated with heightened cytoplasmic DHX9 levels alongside diminished nuclear DHX9 content, while circCLASP2 knockdown yielded the opposite effect (Fig. 3C). IF assays corroborated these findings, revealing alterations in DHX9's intracellular distribution after circCLASP2 overexpression or knockdown (Fig. 3D, Fig. S9B-C).

The interaction between circCLASP2 and DHX9 may induce conformational changes in DHX9 or alter its functional mode, resulting in an increased cytoplasmic localization of DHX9 upon binding. However, when the binding between circCLASP2 and DHX9 is blocked using circCLASP2-DEL1 or ΔdsRBD, the subcellular localization of DHX9 is no longer affected by circCLASP2 (Fig. S10A-B). These findings imply that circCLASP2, upon binding to DHX9, doesn't modulate DHX9's overall expression but enhances its cytoplasmic distribution.

To test whether DHX9 is a key molecule in the regulation of NPC malignant progression by circCLASP2, we conducted cell functional recovery experiments. The results show that DHX9 can participate in the regulation of the circCLASP2 axis to promote the proliferation, invasion, and migration of NPC cells (Fig. S11A-F). The increased expression of DHX9 in the cytoplasm may enable it to play a more significant role in the regulation of circCLASP2, thereby achieving oncogenic effects. This also suggests that we should pay more attention to the impact of circCLASP2 and DHX9 in the cytoplasm of NPC cells. In the cytoplasm, many circRNAs can regulate post-transcriptional gene expression [54], and DHX9, as a helicase, may also function by modulating the stability or structure of downstream mRNAs.

CircCLASP2 promotes the binding between DHX9 and PCMT1 mRNA, leading to an upregulation of PCMT1 expression

To further explore the downstream targets modulated by the interplay of circCLASP2 with DHX9, we manipulated circCLASP2 levels in NPC cells and conducted LC–MS/MS analysis to identify differentially expressed proteins influenced by circCLASP2. Two independent replicates were performed, resulting in two distinct datasets of upregulated proteins (Supplementary Table 7). To enhance the reliability of our findings, we cross-referenced the outcomes of these two replicates, identifying 24 potential upregulated genes (Fig. S12A). Utilizing RNA hybrid analysis, we assessed putative downstream targets. Among the top 5 molecules exhibiting significant upregulation by circCLASP2, RNA hybrid analysis pinpointed the 5'-UTR region of PCMT1 mRNA as having notable binding affinity with circCLASP2, characterized by the lowest minimum free energy (MFE = −114.1 kcal/mol) and strong sequence coherence (Fig. S12B). Circular RIP (circRIP) provided further evidence confirming the interaction between circCLASP2 and PCMT1 mRNA (Fig. S12C). Based on this evidence, we prioritized PCMT1 for further validation (Fig. S12D).

Of the 24 potential regulated genes by circCLASP2, RNA hybrid prediction revealed that the mRNA of 5 other genes (RAB10, CSNK2A1, TPBG, ARF3, PRPF19) has a higher binding affinity towards circCLASP2 than PCMT1. Consequently, we conducted RIP assays to examine the binding between DHX9 and the mRNAs of these 5 candidate molecules. The results demonstrated that the binding abundance of DHX9 with PCMT1 mRNA was significantly higher than with the other candidate molecules (Fig. S12E), suggesting PCMT1 might play the most critical role in the regulatory system involving circCLASP2 and DHX9 in NPC.

In subsequent experiments, we validated that circCLASP2 overexpression promotes the expression of PCMT1. In contrast, knockdown of circCLASP2 suppresses PCMT1 expression (Fig. 3E). Given the concurrent binding of circCLASP2 with DHX9 and PCMT1 mRNA, we further explored whether these three components form a ternary complex and the dynamics within this system. RIP assay revealed that DHX9 exhibits substantial binding to PCMT1 mRNA across three NPC cell lines (Fig. 3F). Notably, overexpression of circCLASP2 significantly intensified the interaction between DHX9 and PCMT1 mRNA, whereas circCLASP2 knockdown attenuated this interaction. These findings suggest that circCLASP2 presence enhances the binding between DHX9 and PCMT1 mRNA (Fig. 3G).

To explore whether PCMT1 expression is regulated by DHX9, we modulated DHX9 levels and assessed the impact on PCMT1 expression. The results revealed that DHX9 overexpression promotes PCMT1 expression, whereas DHX9 knockdown leads to a reduction in PCMT1 protein levels (Fig. 3H). Furthermore, in NPC cell lines, knocking down DHX9 reversed the increase in PCMT1 levels induced by circCLASP2 overexpression (Fig. 3I).

PCMT1 is associated with the clinical progression of NPC, and involvement in the circCLASP2-regulated pathway of NPC malignancy

To explore the expression relationship of the circCLASP2-DHX9-PCMT1 axis in NPC tissues, we performed ISH for circCLASP2 and IHC for DHX9 and PCMT1 in 34 NPC tissues and 7 adjacent non-tumor NPE tissues. Our findings showed elevated levels of circCLASP2 and PCMT1 in NPC tissues, with a notable positive correlation between their expressions, while the expression of DHX9 does not significantly correlate with that of circCLASP2. However, we observed that as circCLASP2 expression increased, the localization of DHX9 in the cytoplasm became relatively more pronounced (Fig. 4A-B). To investigate the correlation between PCMT1 and NPC progression, we downloaded and analyzed 563 NPC sample data from TCGA, showing that the expression of PCMT1 was positively correlated with both the clinical stage and TNM stage of NPC (Fig. S13A-B).

Fig. 4

PCMT1 is associated with the progression of NPC and involved in the circCLASP2-regulated pathway of NPC malignancy. A-B ISH and IHC analysis of 34 NPC and 7 adjacent non-tumor NPE tissues demonstrated the correlation between the expressions of circCLASP2, DHX9 and PCMT1. A The expression levels of circCLASP2, DHX9 and PCMT1 in NPC and NPE tissues, detected by ISH and IHC. Left: scale bar = 100 μm; right: scale bar = 50 μm. B Correlation analysis of circCLASP2 ISH and PCMT1 IHC scores. C Western blotting detected circCLASP2’s role in rescuing PCMT1 protein expression after knockdown in NPC cells. MTT assay (D), Transwell invasion assay (E), and wound healing assay (F) assessing the impact of circCLASP2-PCMT1 axis on the migration, invasion, and proliferation of NPC cells. Data were presented as mean ± SD. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001

To investigate whether circCLASP2 contributes to NPC progression through PCMT1, we employed siRNA to silence PCMT1 expression in NPC cells. Due to the promotive effect of circCLASP2 on PCMT1 protein expression levels, when circCLASP2 was overexpressed in PCMT1 knockdown cells, the protein level of PCMT1 was restored (Fig. 4C). MTT assays, Transwell invasion assays, and wound healing assays demonstrated that, compared to the control group, PCMT1 knockdown significantly inhibited the migration, invasion, and proliferation abilities of NPC cells in vitro. When circCLASP2 was overexpressed concurrently with PCMT1 knockdown in cells, the suppression of PCMT1 mitigated the migration, invasion, and proliferation induced by circCLASP2 in NPC cells (Fig. 4D-F, Fig. S14A-C). These results suggest that circCLASP2 mediated PCMT1 contributes to the promotion of NPC cell proliferation and metastasis.

These findings suggest that circCLASP2 serves as a molecular scaffold in the cytoplasm of NPC cells, facilitating the interaction between DHX9 and PCMT1 mRNA. DHX9 plays a pivotal role in regulating PCMT1 protein expression, and PCMT1 is involved in the regulation of circCLASP2 to promote the malignant progression of NPC. Thus, both DHX9 and PCMT1 represent a crucial component in the circCLASP2 regulatory axis.

DHX9 unwinds the G4 structure on PCMT1 mRNA, promoting the enrichment of polyribosomes and enhancing translation

To further elucidate the regulatory mechanism following the binding of DHX9 to PCMT1 mRNA, we manipulated DHX9 expression levels and assessed the expression and stability of PCMT1 mRNA. Surprisingly, DHX9 did not exert any influence on either the expression levels or the stability of PCMT1 mRNA (Fig. S15A-B). Consequently, it appears that DHX9's regulation of PCMT1 expression occurs at the translation level. Given DHX9's role as a helicase molecule, we investigated whether DHX9 unwinds the secondary structure of PCMT1 mRNA upon binding to regulate its translation.

QGRS Mapper predicted the existence of a sequence near the translation initiation site of PCMT1 mRNA that could potentially form an RNA G-quadruplex structure (rG4) (Fig. S15C). Previous studies have demonstrated that RNA G-quadruplexes, particularly when located in such positions, play a crucial role in regulating mRNA translation [55, 56]. We drew a schematic of the sequence and its spatial arrangement based on the prediction results (Fig. S15C). To validate the presence of the rG4 structure on PCMT1 mRNA, we utilized the BG4 antibody, which selectively binds to the G-quadruplex structure, and designed a digoxin-labeled probe to target the flanking sequence of the predicted rG4-forming region of PCMT1 mRNA. Cells were treated with pyridostatin (PDS), a statin and a highly selective G4 structure stabilizer. IF results demonstrated that following PDS treatment, higher colocalization signals were observed compared to the control group without PDS treatment, which means rG4 structure on PCMT1 mRNA notably intensified. To investigate whether DHX9 unwinds the rG4 structure on PCMT1 mRNA, we modulated the DHX9 expression. The findings revealed that overexpression of DHX9 led to a reduction in fluorescence colocalization, signifying a notable decrease in the predicted rG4 structure of PCMT1 mRNA; conversely, knocking down DHX9 resulted in increased colocalization signals, thereby significantly enhancing the rG4 structure of the target segment (Fig. 5A, Fig. S15D). These outcomes suggested that after circCLASP2-facilitated DHX9 binding to PCMT1 mRNA, DHX9 can leverage its nucleic acid helicase function to efficiently unwind the rG4 structure near the translation initiation site of PCMT1 mRNA.

Fig. 5

DHX9 enhances PCMT1 mRNA translation by unwinding the G4 structure. A FISH-IF experiments showing the presence of rG4 structure on PCMT1 mRNA in HNE2 cells. Blue: DAPI; Red: Sequences flanking the region on PCMT1 mRNA that potentially form rG4 structures; Green: BG4; Yellow: colocalization of red and green fluorescence. Scale bar = 10 μm. B Dual-luciferase reporter assay system measuring luciferase activity in HNE2 cells after overexpressing circCLASP2 and/or knocking down DHX9. C Dual-luciferase reporter assay system measuring luciferase activity in HNE2 cells after overexpressing circCLASP2 and/or PDS treatment. D Ribosome separation with sucrose density gradient assessing the effects of different treatments on the translation efficiency of PCMT1 mRNA in NPC cells. Treatment conditions included knockdown of circCLASP2, knockdown of DHX9, overexpression of circCLASP2 with concurrent knockdown of DHX9, and overexpression of circCLASP2 with PDS treatment. RNA from various ribosomal components was extracted for qRT-PCR to quantify the enrichment of PCMT1 mRNA on monosomes and polysomes. GAPDH mRNA served as a control. E Western blotting detecting the impact of overexpressing circCLASP2 or DHX9 with concurrent PDS treatment on PCMT1 expression in HNE2, CNE2, and HONE1 cells. Data were presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001

To investigate the impact of DHX9 unwinding of rG4 on the translation efficiency of PCMT1 mRNA, we constructed a wild-type dual-luciferase reporter gene vector (WT), inserting the sequence of PCMT1 mRNA binding to circCLASP2 and the predicted rG4 forming sequence into the upstream of the luciferase gene. A dual-luciferase reporter assay revealed that in NPC cells, circCLASP2 overexpression significantly increased luciferase activity, indicating enhanced luciferase translation compared to the control group. Knockdown of DHX9 reduced fluorescent protein expression, and simultaneous knockdown of DHX9 with circCLASP2 overexpression reversed the circCLASP2-induced increase in fluorescent protein expression. Treatment with PDS also significantly inhibited fluorescent protein expression and reversed circCLASP2-induced luciferase gene translation (Fig. 5B-C, Fig. S16A). Additionally, we constructed two mutant vectors by mutating the GGs of the rG4 forming sequence and the segment of PCMT1 mRNA binding to circCLASP2, respectively. Dual-luciferase reporter assay results showed that the translation-enhancing effect observed with circCLASP2 overexpression in the WT vector disappeared in both mutation vectors, with no statistically significant difference in luciferase activity compared to the control group (Fig. S16B). These findings suggest that DHX9-mediated unwinding of the G4 structure on PCMT1 mRNA is crucial for promoting PCMT1 expression, and circCLASP2 enhances this process via DHX9.

To further support these findings, we utilized the sucrose density gradient method to isolate ribosomes and fractionate them hierarchically into oligoribosome and polysome fractions. The abundance of PCMT1 mRNA in different fractions was assessed via qRT-PCR to evaluate its translational efficiency. Results indicated that knockdown of either circCLASP2 or DHX9 resulted in reduced polysome-associated PCMT1 mRNA levels, implying an inhibition of PCMT1 mRNA translation. Conversely, overexpression of circCLASP2 led to an increase in the enrichment of PCMT1 mRNA on polysomes. However, co-transfection of circCLASP2 overexpression vector and si-DHX9 in NPC cells revealed that DHX9 knockdown could reverse the translation enhancement induced by circCLASP2, suggesting that circCLASP2 regulates PCMT1 mRNA translation through DHX9. To ascertain whether this regulatory process is linked to the rG4 structure, we added PDS in conjunction with circCLASP2 overexpression. Results demonstrated that PDS could attenuate the translation-promoting effect of circCLASP2 on PCMT1 mRNA. As a control, the expression of GAPDH mRNA in different fractions was also examined in each group, indicating that the polysome association remained unaffected by the treatments (Fig. 5D). Furthermore, following PDS treatment, both mRNA and protein levels of PCMT1 were evaluated, revealing that while PDS did not affect PCMT1 mRNA expression (Fig. S16C), whereas it inhibited protein expression (Fig. S16D). PDS treatment was able to reverse the increase in PCMT1 expression induced by circCLASP2 (Fig. 5E).

Collectively, the above experimental results demonstrate that, mediated by the molecular scaffold circCLASP2, DHX9 unwinds the rG4 structure on PCMT1 mRNA translation initiation site to promote the enrichment of polysomes on PCMT1 mRNA, ultimately enhancing PCMT1 translation.

PCMT1 facilitates cytoskeleton polymerization by upregulating the expression of cytoskeleton-related proteins

PCMT1 could exert its biological function through direct binding to its target proteins [57, 58]. To explore the mechanism by which PCMT1 regulates the malignant progression of NPC cells, we conducted IP experiments coupled with LC–MS/MS to identify potential binding proteins of PCMT1, yielding 97 candidate molecules. Among the top 10 scoring proteins, five were members of the actin family, which constitute the cytoskeleton-related proteins α-actinin and β-actin (Supplementary Table 8). PCMT1-IP silver staining revealed distinct bands corresponding to these proteins. Additionally, a differential band near the 20 kDa position was identified as Cofilin1 (CFL1) protein through comparison with PCMT1-IP results (Fig. 6A). IP experiments confirmed the reciprocal binding relationship between α-actinin, β-actin, CFL1, and PCMT1 (Fig. 6B). Overexpression of PCMT1 upregulated the protein levels of α-actinin, β-actin, and CFL1, while PCMT1 knockdown exhibited opposite effects (Fig. 6C). Phalloidin staining of actin filaments revealed that the overexpression of circCLASP2, DHX9 and PCMT1 enhanced filament aggregation in NPC cells, whereas knockdown of these molecules resulted in disorganized and scattered distribution of filaments, indicative of loose and directionless cells [59] (Fig. 6D, Fig. S17A-B).

Fig. 6

PCMT1 modulates cytoskeletal polymerization through regulating cytoskeleton-associated proteins. A PCMT1 binding proteins were captured by IP, which were then separated by SDS-PAGE and visualized using silver staining. B IP detecting the interactions between PCMT1 and α-actinin, β-actin, or CFL1 in NPC cells. Specific primary antibodies were used for enrichment, followed by detection via Western blotting. C Western blotting assessing the impact of overexpressing or knocking down PCMT1 on the protein levels of α-actinin, β-actin, and CFL1. D Phalloidin staining in NPC cells showing actin filament intensity and aggregation after overexpressing or knocking down circCLASP2, DHX9 and PCMT1. Scale bar = 10 μm

The polymerization of actin filaments reflects enhanced cytoskeletal integrity. CFL1, a pivotal regulator of cytoskeletal dynamics, contributes to the balance between cytoskeletal strength and flexibility, facilitating cell movement and deformation [60]. This dynamic cytoskeletal architecture generates coordinated forces within cells, which is crucial for the migration of tumor cells [61]. Moreover, the process of cell mitosis and the distribution of cellular contents heavily rely on the functionality of actin filaments and the cytoskeleton [62]. Consequently, these findings correlated with the augmented proliferation and invasion-migration capabilities of NPC cells observed in vitro and in vivo.

rG4-targeting drug PDS effectively inhibits the proliferation and metastasis of NPC

To explore the impact of PDS on NPC malignant progression mediated by the circCLASP2 regulatory axis, we treated NPC cells with PDS and evaluated their proliferation and metastasis abilities. MTT assays, Transwell invasion assays, and wound healing assays revealed that PDS treatment notably suppressed the proliferation, invasion, and migration capacities of NPC cells. Concurrent administration of PDS with circCLASP2 overexpression significantly attenuated the proliferative, invasive, and migrative effects induced by circCLASP2 in NPC cells (Fig. S18A-C).

To further validate the impact of PDS intervention on NPC malignant progression in vivo, we established animal models. In the subcutaneous xenograft model, mice receiving PDS injections displayed notably reduced mass and volume of subcutaneous xenograft tumors compared to the control group receiving saline injections. Furthermore, mice treated with PDS following circCLASP2 overexpression exhibited slower tumor growth and significantly smaller tumor volume compared to the group solely overexpressing circCLASP2 (Fig. 7A). H&E staining, ISH, and IHC results revealed elevated levels of Ki67 and PCMT1 in tumors of mice overexpressing circCLASP2. However, PDS treatment significantly reduced Ki67 levels in tumor tissues, and also decreased the levels of PCMT1, mitigating the effects of circCLASP2 overexpression (Fig. 7B).

Fig. 7

G4 stabilizer PDS inhibits the proliferation and metastasis of NPC cells both in vitro and in vivo. A-E CNE2 cells (2 × 10.⁶) transfected with empty vectors or circCLASP2 overexpression vectors were injected subcutaneously or into the footpads of BALB/c mice. PDS or saline was administered via the tail vein. A Tumor tissues from in situ xenograft mice models (n = 6 per group) (left). Measurements of tumor volume (middle) and weight (right) in the in situ xenograft model. B Representative images of H&E, ISH, and IHC staining of subcutaneous tumor sections from the xenograft model. Scale bar = 100 μm. C Representative images of the footpad and ipsilateral popliteal lymph node from the "footpad-lymph node" metastasis model. D Images of the popliteal lymph nodes in the "footpad-lymph node" metastasis model (n = 6 per group) (left). Statistical analysis of the weight (middle) and volume (right) of popliteal lymph nodes in the "footpad-lymph node" metastasis model. E Representative images of H&E and CK-PAN staining of popliteal lymph node sections from the "footpad-lymph node" metastasis model, showing metastatic tumor tissue. Top: scale bar = 200 μm; Bottom: scale bar = 100 μm. F Schematic diagram illustrating the circCLASP2-DHX9-PCMT1 axis driving malignant progression of NPC. Data were presented as mean ± SD. *, P < 0.05; **, P < 0.01

In the "footpad-lymph node" metastasis model, mice that overexpressed circCLASP2 displayed accelerated in situ tumor growth at the footpad and larger volumes of popliteal lymph nodes. However, treatment with PDS exhibited a notable therapeutic effect (Fig. 7C-D, Fig. S19A-B). H&E staining and IHC with the CK-PAN antibody of lymph nodes from each group revealed significantly more metastatic tumor tissue in the lymph nodes of mice overexpressing circCLASP2. Conversely, mice treated with PDS showed almost no tumor metastasis to the lymph nodes. Moreover, PDS treatment could reverse the lymph node metastasis induced by circCLASP2 (Fig. 7E). Additionally, H&E staining of the heart, liver, spleen, lungs, and kidneys from each group of mice indicated that PDS did not induce any toxicity in these organs (Fig. S19C). These results demonstrate that PDS treatment significantly inhibits the proliferation and metastasis of nasopharyngeal carcinoma in nude mouse models, further supporting the proposed regulatory mechanism of circCLASP2.

In summary, circCLASP2 drives NPC malignant progression by serving as an adapter molecule, promoting the interaction between DHX9 and PCMT1 mRNA. DHX9 unwinds the inhibitory G-quadruplex structures on the translation initiation site of PCMT1 mRNA, leading to polyribosome enrichment and facilitating PCMT1 translation. Elevated PCMT1 levels subsequently interact with cytoskeleton-associated proteins, boosting their protein levels and enhancing cell motility and division capabilities, ultimately driving NPC proliferation and metastasis (Fig. 7F).

Discussion

In this study, we discovered a novel circular RNA, circCLASP2, previously unreported in tumors, exhibiting upregulation in NPC and correlation with unfavorable patient prognosis. Serving as a molecular scaffold, circCLASP2 orchestrates the binding between DHX9 and PCMT1 mRNA, thereby facilitating DHX9-mediated unwinding of the G-quadruplex structure on PCMT1 mRNA, resulting in heightened PCMT1 expression. This phenomenon impacts cytoskeletal aggregation and dynamics, ultimately promoting NPC proliferation and metastasis. Notably, existing research underscores the significance of various circular RNAs in NPC malignant progression through diverse mechanisms. For instance, circPVT1 sequesters the E3 ubiquitin ligase β-TrCP, impeding c-Myc proteasomal degradation, thus fueling invasion and metastasis in NPC cells [32]. Similarly, circARHGAP12 directly binds to EZR mRNA's 3'-UTR, bolstering its stability and driving NPC invasion and metastasis [35]. Additionally, EBV-encoded circRPMS1 functions as a molecular sponge for miR-203, miR-31, and miR-451, facilitating NPC onset by mediating epithelial-mesenchymal transition (EMT) [63]. In the mechanistic exploration, our focus was on circCLASP2's pivotal role in binding both DHX9 and PCMT1 mRNA, forming a crucial ternary complex in this complex system.

DHX9, also known as RNA helicase A, is a member of the SF2 superfamily of nucleic acid helicases, pivotal in modulating splicing, transcription, and translation of both DNA and RNA by influencing their secondary structures [41, 64, 65]. Earlier studies elucidated DHX9's capability to unwind various secondary structures in human transcripts. In comparison to WRN, another member of the same helicase family, DHX9 exhibits heightened affinity and accelerated unwinding speed for RNA [66]. At the post-transcriptional level, DHX9's biological function includes activating the translation of the JUND gene and specific retroviral mRNA by binding to their 5'-UTR regions [67, 68]. Additionally, DHX9 regulates the activity of the internal ribosome entry site (IRES) in p53 mRNA, thereby stimulating IRES-mediated p53 translation [69].

In 2018, Shankar Balasubramanian's team elucidated the significance of mRNA's secondary structure in post-transcriptional gene expression regulation. They found that the RNA G-quadruplex structure interacts physically with the helicase family proteins DHX36/DHX9, which helps enrich ribosomes and modulate translation [70]. RNA G-quadruplex structures, as non-classical nucleic acid secondary structures, constitute a class of post-transcriptional regulatory factors relevant to disease gene expression, exerting significant regulatory effects in disease contexts through interactions with RNA-binding proteins [71]. Several studies have validated DHX36's role in translation regulation by unwinding RNA G-quadruplexes. For instance, Wang et al. in 2021 illustrated DHX36's promotion of translation for downstream effector molecule Gnai2 by unwinding the rG4 structure, which is crucial for cell activation and proliferation during skeletal muscle satellite cell regeneration [72]. Subsequently, in 2022, Chun Kit Kwok et al. identified the RNA G-quadruplex structure within the 5'-UTR region of ADAR1 mRNA. They demonstrated DHX36's interaction with this structure, thereby regulating translation in an rG4-dependent manner [73]. However, to date, such mechanisms involving DHX9 have not been reported. Our research, for the first time, unveils DHX9's role in regulating PCMT1 translation by unwinding the rG4 structure near the translation initiation site on PCMT1 mRNA, facilitating ribosome enrichment. Considering the prevalence of rG4 and other secondary structures that are subject to DHX9 regulation within nucleic acid molecules, this regulatory mode may not be unique. By acting as a bridging molecule, circCLASP2 enhances DHX9's ability to influence these structures, potentially facilitating the formation of a synergistic regulatory network that drives the malignant progression of NPC. The implications of this regulatory interaction merit thorough exploration in our subsequent research.

PCMT1, a protein-L-isoaspartate (D-aspartate) methyltransferase, exhibits robust sequence conservation across various organisms [74]. Its role extends across diverse oncogenic regulatory pathways and is notably upregulated in several cancers. For instance, PCMT1 plays a pivotal role in ovarian cancer metastasis [75] and promotes anti-apoptosis and tumor metastasis in hepatocellular carcinoma, while also modulating immune processes [76]. Mammalian systems lacking PCMT1 often exhibit isoaspartate residue accumulation in proteins, rendering them functionally impaired or displaying characteristics associated with aging [77]. In this investigation, we uncovered PCMT1's association with cytoskeleton-associated proteins, resulting in elevated protein levels and enhanced functionality of these proteins. Despite advancements, research into the oncogenic role of aspartate methylation remains in its early stages, and we aim to delve deeper into the intricate mechanisms in our forthcoming studies.

β-actin, among the most abundant intracellular proteins, constitutes the primary framework of cytoskeletal microfilaments [78]. On the other hand, α-actinin, a member of the actin cross-linking protein superfamily, interacts with actin filaments to forge a stable intracellular lattice, pivotal for organizing and fortifying the cytoskeleton [79]. Actin filaments endowed with robust strength furnish cells with internal mechanical support and impetus for motion [80], while their polarity confers cellular dynamism and orientation [81]. Cofilin1, a constituent of the actin-binding protein family, orchestrates the dynamics of the actin-based cytoskeleton [82]. Cofilin1 binding markedly induces the twisting of actin filaments [62], and can also spur their depolymerization and turnover [83], thereby reshaping the structure and function of actin. A cytoskeleton endowed with both resilience and dynamism underpins tumor cells' ability to breach physiological barriers and engage in distant metastasis. Furthermore, during cell proliferation, the concerted action of Cofilin1 and actin propels contractile ring contraction and facilitates cellular content distribution [84, 85]. Phalloidin staining revealed that the circCLASP2-DHX9-PCMT1 regulatory axis significantly amplifies actin filament aggregation, endowing cells with conspicuous orientation.

PDS, a known statin and a potent G-quadruplex structure stabilizer, has garnered attention for its potential therapeutic applications. While not conventionally classified as an anticancer drug, several studies have highlighted its role in modulating the cell cycle and DNA damage repair in Hela cells [86]. Additionally, PDS has shown promise in disrupting breast cancer cell replication, thereby augmenting the efficacy of traditional anticancer agents [87]. Our findings offer novel insights by demonstrating the significant therapeutic efficacy of PDS in treating NPC, shedding light on its potential as a small-molecule anticancer agent. This discovery opens new avenues for exploring targeted therapies and devising combination drug strategies for combating tumors.

Conclusions

Our study has delineated the critical role of a circular RNA, circCLASP2, in the malignant progression of NPC. We characterized the rG4-dependent translational regulatory mechanism of DHX9 on PCMT1, which is mediated by circCLASP2. This axis modulates cytoskeletal dynamics, contributing to NPC proliferation and metastasis. Our findings provide novel insights into the mechanisms underlying NPC progression and underscore the therapeutic potential of specific translation-disrupting small molecules in oncology. The circCLASP2-DHX9-PCMT1 axis may represent a promising target for early diagnosis and therapeutic intervention, with significant implications in clinical practice.