CircROR1 binds HNRNPL to regulate FOXO4 pre-mRNA splicing, promoting cutaneous melanoma metastasis and serving as a therapeutic target via RNAi-loaded PEG-LNPs

TL;DR

A circular RNA called circROR1 promotes melanoma metastasis by regulating gene splicing and can be targeted with RNAi-loaded nanoparticles for treatment.

Contribution

Discovery of circROR1 as a novel oncogenic circular RNA and development of a targeted RNAi nanotherapy for melanoma.

Findings

CircROR1 promotes melanoma metastasis by regulating FOXO4 pre-mRNA splicing through HNRNPL.

CircROR1 upregulation increases EMT and resistance to PD-L1 antibody therapy.

RNAi-loaded PEG-LNPs targeting circROR1 reduce tumor metastasis in vivo.

Abstract

Circular RNAs (circRNAs) contribute to gene expression regulation by interacting with splicing factors, a process that is often disrupted in cancers such as cutaneous melanoma (CM).

A circRNA microarray analysis was performed to identify differentially expressed circRNAs. qRT‒PCR was conducted to confirm the expression of circROR1. CCK-8, colony formation, wound healing, and transwell assays were used to analyze proliferation, metastasis and apoptosis in CM cells. Xenograft models and IHC experiments were established to confirm the effects of circROR1 on tumor growth and metastasis in vivo. RNA sequencing and pull-down–MS experiments were performed to identify the mechanisms downstream of circROR1. Nuclear and cytoplasmic fractionation, along with FISH experiments, were conducted to determine the cellular localization of circROR1. To target circROR1 for CM treatment, we used a microfluidic strategy to develop FA-PEG(si-circ) nanoparticles for efficient siRNA delivery.

In CM samples, circROR1 levels were positively correlated with HNRNPL levels and tumor metastasis but negatively correlated with FOXO4 protein levels. CircROR1 was prevalent in CM, and its upregulation increased the levels of factors involved in epithelial–mesenchymal transition, cell migration, and invasion. CircROR1 overexpression conferred resistance to PD-L1-antibody therapy in CM cells by downregulating PD-L1 expression. CircROR1 recruited HNRNPL, influencing its nuclear translocation, and further prevented intron retention in FOXO4 mRNA. In HNRNPL-overexpressing CM cells, circROR1 upregulation inhibited FOXO4α expression and promoted FOXO4ζ expression. Increased FOXO4α expression counteracted circROR1’s effects and suppressed metastatic behaviors. FA-PEG(si-circ) enhanced siRNA stability and efficiency, reducing CM cell lung colonization in vivo.

This study identified circROR1 as an oncogenic circular RNA that plays a crucial role in tumor progression and metastasis. CircROR1-targeted nanotherapy is a promising option for the treatment of metastatic cancer.

The online version contains supplementary material available at 10.1186/s12943-025-02525-1.

Figures

CircROR1 expression is significantly upregulated in metastatic melanoma. a A heatmap generated by the OmicShare platform of circRNA microarray data showing differential expression of circRNAs in HEM, WM35, and WM451 cells (with three independent biological replicates in each group, fold change >|2.0| and *P* < 0.05). The top 8 dysregulated circRNAs, which were highly expressed in WM35 and WM451 cells, but were expressed at much higher levels in WM451 cells, are annotated on the map. b Comparison of circ0000082 expression between nevus tissues (*n* = 8) and CM tissues (*n* = 8). GAPDH was used

CircROR1 inhibits apoptosis and promotes CM cell progression, metastasis, and EMT. a The heatmap displays the top 20 GO analysis results for upregulated genes, conducted using Metascape and OmicShare (-log *P* value ≥2). b, c Wound healing assays of A375 and A2058 cells with stable differential expression of circROR1, at 0, 24, and 48 hours. The migration area was quantified using ImageJ. Statistical significance was calculated using Student’s t test. Data are presented as the means ± SEMs from three independent experiments. d, e Transwell assays of A375 and A2058 cells with stable differentia

CircROR1 stimulates the growth and metastasis of CM in vivo. a The scheme of the in vivo study involved the use of a subcutaneous xenograft model in C57BL/6 mice with B16 cells. b Tumor size of subcutaneous B16 xenograft tumors in mice from the indicated groups (OV-NC *vs*. OV-circ, KD-NC *vs*. KD-circ, n =5 mice per group). c Tumor volume in C57BL/6 mice injected with stable circROR1 overexpression or knockdown B16 cells into dorsal flank (n = 5 mice per group). d The weights of xenograft tumor tissues on day 16 from the indicated groups were measured (OV-NC *vs*. OV-circ, KD-NC *vs*. KD-circ

CircROR1 interacts with and stabilizes HNRNPL. a Flowchart illustrating the selection process for circROR1 RBPs. b Reactome pathway analysis of upregulated proteins identified through circROR1 pull-down and MS assays. c Venn diagram illustrating the analysis of unique and shared RBPs between those associated with prognosis in CM, predicted RBPs based on circROR1 binding sites, and upregulated proteins identified in the MS analysis of A375 cells with circROR1 overexpression. d Volcano plot of differentially expressed proteins identified by the MS analysis of A375 cells with circROR1 overexpress

CircROR1 promotes CM metastasis by triggering FOXO4 isoform switching through interactions with HNRNPL protein. a The Draw Venn Diagram online tool generated the Venn diagram. b Volcano plot of DEGs identified in [GSE8401](GSE8401). c Volcano plot of DEGs identified in [GSE46517](GSE46517). d Images of IHC staining for FOXO4 in primary CM tissues and metastatic CM tissues from tissue chip. e Comparison of FOXO4 IHC scores between primary CM tissues and metastatic CM tissues. f Correlation analysis of circROR1 FISH score with FOXO4 IHC score in CM tissues from tissue chip. g Schematic diagram o

In vivo delivery of anti-CircROR1 suppresses melanoma metastasis. a, b Inhibited effects of si-circROR1-2 on A375 cell proliferation were examined by CCK-8 assay and colony formation assay. The cell-covered area of colony assays was quantified using ImageJ. c, d The inhibitory effects of si-circROR1-2 on A375 cell migration and invasion were examined by transwell assay. The cell-covered area was quantified using ImageJ. e The scheme involved establishing a lung colonization nude mouse model using luciferase (Luc)-expressing A375 cells. IVIS monitoring of lung colonization was performed on the

Schematic diagram of the overall experimental design. CircROR1 binds to the splicing factor HNRNPL, promoting its nuclear translocation and regulating the alternative splicing of FOXO4 pre-mRNA. Upregulation of circROR1 suppresses the generation of the FOXO4α isoform while promoting FOXO4ζ expression, thereby inducing the metastatic potential of CM cells. To achieve circROR1-targeted therapy, a folate-modified polyethylene glycol lipid nanoparticle complex [FA-PEG(si-circ)] was developed using a microfluidic strategy for efficient siRNA delivery and stable gene silencing, effectively suppressi

Full text

Introduction

The incidence and mortality rates of cutaneous melanoma (CM) have increased significantly, indicating that CM poses a serious threat to human health [1]. Clinically, CM is characterized by early metastasis, poor prognosis, and a high mortality rate, with a five-year survival rate of only 16% for patients with metastatic melanoma [2]. Metastasis and dissemination from the primary tumor to secondary organs is a continuous and progressive process involving multiple factors, including tumor cell migration, survival in the circulation, extravasation, and colonization of distant sites [3]. Despite breakthroughs in treating metastatic melanoma with BRAF kinase inhibitors and the CTLA-4 monoclonal antibodies, the prognosis for patients with advanced disease remains poor, primarily due to the emergence of resistance mechanisms in most tumors [4, 5]. Therefore, new treatments for CM are needed, requiring in-depth research into the molecular mechanisms that underlie melanoma invasion and metastasis to identify effective therapeutic targets, reduce metastasis rates, and ultimately improve patient prognosis.

Circular RNAs (circRNAs) are covalently closed, single-stranded circular transcripts that are generated through precursor mRNA back-splicing or exon skipping events and are characterized by the absence of 5′ caps and 3′ poly(A) tails [6]. Due to their unique structure, circRNAs exhibit high stability, abundance, and evolutionary conservation, which confer clear advantages for circRNAs as tumor biomarkers [7, 8]. CircRNAs have been identified as key regulators of various biological processes involved in tumor progression, where they may act as microRNA sponges or RNA-binding protein (RBP) sponges, templates for protein/peptide translation, or regulators of gene transcription and RNA splicing [9]. Among these functions, the activity of circRNAs as RBP sponges has attracted increasing attention in recent years. For example, circANKS1B has been reported to promote breast cancer metastasis, mediated by the splicing factor ESRP1 [10]. Hsa_circ_0000518 promotes the malignant progression of hepatocellular carcinoma by regulating ITGA5, thereby activating the Warburg effect [11]. Circ_0088300 can upregulate the expression of the BOLL protein and further facilitate gastric cancer growth and metastasis through the promotion of mitochondrial metabolic reprogramming [12]. Circ_0053943 facilitates uveal melanoma progression by stabilizing EGFR and further activating the MAPK/ERK signaling pathway [13]. However, the potential functions, biogenesis processes, and associated mechanisms underlying circRNA–RBP complex activity in CM remain largely unexplored.

Alternative splicing during mRNA maturation is a crucial process in the posttranscriptional regulation of gene expression [14]. HNRNPL, a member of the heterogeneous nuclear ribonucleoprotein (hnRNP) family, is recognized primarily as an alternative splicing factor. In various clinical diseases, HNRNPL is responsible for regulating intron retention and exon inclusion and exclusion and consequently defines mRNA splice variants in various disease contexts [15, 16]. For example, HNRNPL plays a crucial role in maintaining vascular integrity by regulating the alternative splicing of TJP1 via its interaction with the lncRNA NTRAS [17]. Moreover, HNRNPL influences B-cell activation, germinal center formation, and antibody responses by modulating transcriptional programs and metabolism through the alternative splicing of histone modifiers [18]. In prostate cancer, the lncRNA SNHG1 competitively binds to hnRNPL, inhibiting the E-cadherin translation. This interaction enhances the role of SNHG1 in the epithelial–mesenchymal transition (EMT) pathway, ultimately facilitating tumor metastasis [19]. Despite the involvement of HNRNPL in alternative mRNA splicing in CM, the underlying mechanisms remain largely unexplored [20].

Our team previously reported that circROR1 is significantly upregulated in CM [21, 22]. These studies established circROR1 as a functionally relevant circRNA in CM, yet the precise molecular mechanisms underlying its activity, especially in the context of splicing regulation and therapeutic targeting, remain unclear. In this study, we extend this previous work, demonstrating that circROR1 promotes the expression of EMT-related factors, cell migration, and invasion. The underlying mechanism likely involves its stabilizing interaction with the splicing factor HNRNPL, which increases FOXO4α mRNA expression. We also developed a therapy combining si-circROR1 with a nanoliposome drug delivery system [FA-PEG(si-circ)] to target circROR1, resulting in the inhibition of CM lung colonization in vivo. Thus, circROR1 may be a promising biomarker for CM metastasis, and the delivery of siRNAs targeting circROR1 represents a crucial strategy for CM treatment. Together, these findings reveal that circROR1 binds to HNRNPL and regulates its splicing of FOXO4 pre-mRNA to promote CM metastasis and present nanoparticle-based circRNA-targeted therapy as a promising strategy for treating CM.

Materials and methods

For a comprehensive description of the methodologies used in this study, please refer to the Supplementary Materials [8, 20, 23–34].

CircRNA microarray

A circRNA microarray analysis was performed to unveil the expression profile of circRNAs in human epidermal melanocytes (HEMs), the low-metastatic melanoma cell line WM35 and the high-metastatic melanoma cell line WM451, each of which was analyzed in three independent biological replicates. The results presented in this section have been previously published by our team [21].

Patient tissue specimens

Tissue samples, including 8 nevus samples, 10 paired CM samples, and their matched adjacent normal tissues, were obtained from patients after surgical resection at Xiangya Hospital of Central South University. These patients did not undergo any anticancer treatment. A melanoma tissue chip was obtained from Shanghai Zhuoli Biotechnology Company Ltd (Shanghai, China). All patients signed informed consent forms, and the protocols were approved by the Ethics Committee of Xiangya Hospital of Central South University (202308636) and Shanghai Zhuoli Biotechnology Company Ltd (LLS M-15-01).

RNA sequencing (RNA-seq) and data analysis

The RNA-seq analysis was performed as described previously. CLC Genomics Workbench (Qiagen Bioinformatics) was used for analysis of the gene ontology according to a standard protocol [21].

Synthesis and identification of FOXO4α and FOXO4ζ alternate splice forms

Quantitative real-time RT‒PCR (qRT‒PCR) of FOXO4 Genomic PCR amplifications were performed with primers shown as follows: Forward primer: 5'-CTGTGGCAGGCTTCACTGAAC-3';

Reverse primer: 5'-GCAAGTGTCAGTCGCTTCTC-3'.

Gel Electrophoresis Genomic PCR products were separated on 2.5% agarose gels. DNA was stained with Vistra Green and visualized using Bio-Rad ChemiDoc XRS with ChemiDoc XRS Imaging System (Bio-Rad Laboratories, California, USA).

A Storm FluorImager system (Molecular Dynamics, Sunnyvale, CA) was used. The sequences of the genomic PCR products were analyzed using ImageJ (National Institutes of Health, Bethesda, MD, USA) [35].

RNA pull-down assays

RNA pull-down assays were performed using a commercial kit (BersinBio, Guangzhou, China). Biotinylated circROR1 probes and negative control probes were designed and synthesized by GenePharm (Shanghai, China). Probes (3 μg) were incubated with 50 μL of streptavidin-coated magnetic beads for 30 minutes at 25 °C. Cultured cells were lysed using lysis buffer containing a protease inhibitor cocktail, and nucleic acids were removed by adding 40 μL of agarose beads. The resulting lysate was then incubated with the streptavidin-coated magnetic beads bound to the biotinylated probe for 6 hours at 4 °C with rotation. The bead‒probe‒protein complexes were washed four times with NT2 buffer (5 minutes per wash at 4 °C with rotation). Pull-down proteins were eluted with protein elution buffer for 2 hours at 37 °C with rotation and subsequently analyzed by mass spectrometry. The sequences of the RNA pull-down probes are listed in Supplementary Table S1.

Mass Spectrometry (MS)

Three independent RNA pull-down experiments were analyzed using MS. Eluted samples from NeutrAvidin beads were separated by SDS-PAGE, stained with Brilliant Blue G-Colloidal Concentrate (Sigma), and gel pieces were excised. These pieces were destained, reduced, alkylated, and subjected to trypsin digestion. The resulting peptide mixtures were extracted, concentrated, and analyzed with an EASY nLC system (Proxeon). Peptides were ionized with a Proxeon ion source and analyzed using a Q Exactive mass spectrometer (Thermo Scientific). Data processing was performed with MaxQuant software (version 1.4.0.3), applying default parameters and a 1% false-discovery rate for peptide and protein identification. Label-free quantitation based on summed peptide ion chromatograms was employed to identify differentially interacting proteins.

Proteins identified by more than one peptide and present in at least two out of the three experiments were included in the analysis. Proteins were ranked according to peptide count and intensity. A 2-fold change in intensity-based absolute quantification (iBAQ) was considered significant, with baseline values calculated as the minimum iBAQ for each sample.

RNA-binding protein immunoprecipitation (RIP)

RIP assays were conducted using the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (BersinBio, Guangzhou, China), in accordance with the manufacturer’s instructions and as described in previous studies [36]. A375 cells, transfected with varying levels of circROR1, were harvested 48 hours post-transfection and lysed on ice for 30 minutes using RIP lysis buffer. Following centrifugation, the supernatant was incubated with 30 μL of Protein-A/G agarose beads and specific antibodies, with immunoglobulin G (IgG) serving as a negative control. After overnight incubation, the immune complexes were centrifuged and washed six times with washing buffer. The proteins bound to the beads were analyzed by western blotting. For RNA analysis, the immunoprecipitated RNA was subjected to qRT‒PCR and electrophoresis.

Results

Identification of the circular RNA circROR1 and its significant upregulation in metastatic CM

Many studies have demonstrated that tumor metastasis is a key characteristic of malignant CM and is closely associated with poor prognosis [37]. To explore the mechanisms of CM metastasis, we performed a circRNA microarray to compare the levels of circRNAs between the low-metastatic melanoma cell line WM35 and the high-metastatic melanoma cell line WM451 and control human epidermal melanocytes (HEM), each with three independent biological replicates. We identified eight dysregulated circRNAs (fold change >|2.0| and P < 0.05) that were highly expressed in both WM35 and WM451 cells and exhibited significantly higher levels in WM451 cells, namely, hsa_circ_0087960, hsa_circ_0001833, hsa_circ_0000082, hsa_circ_0008545, hsa_circ_0040609, hsa_circ_0072386, hsa_circ_0004305, and hsa_circ_0052130 (Fig. 1a). We then analyzed the expression profiles of these eight circRNAs in 8 CM tissues and 8 nevus tissues using qRT‒PCR (Fig. 1b, Supplementary Fig. S1a). The results for hsa_circ_0087960, hsa_circ_0000082, and hsa_circ_0008545 were consistent with the high-throughput sequencing data. However, only hsa_circ_0000082 was significantly differentially expressed between the 10 CM tissues and adjacent nontumor tissues (Fig. 1c, Supplementary Fig. S1b). Moreover, qRT‒PCR analysis revealed that most of the CM cells (A2058 and A375) expressed higher levels of hsa_circ_0000082, compared to noncancer cell HaCaT cells (Fig. 1d, Supplementary Fig. S1c).Fig. 1. CircROR1 expression is significantly upregulated in metastatic melanoma. a A heatmap generated by the OmicShare platform of circRNA microarray data showing differential expression of circRNAs in HEM, WM35, and WM451 cells (with three independent biological replicates in each group, fold change >|2.0| and P < 0.05). The top 8 dysregulated circRNAs, which were highly expressed in WM35 and WM451 cells, but were expressed at much higher levels in WM451 cells, are annotated on the map. b Comparison of circ0000082 expression between nevus tissues (n = 8) and CM tissues (n = 8). GAPDH was used as an internal reference for performing qRT‒PCR. Statistical significance was calculated by Student’s t test. c Comparison of circ0000082 expression between CM tissues and corresponding normal tissues (n = 10 pairs). GAPDH was used as an internal reference for performing qRT‒PCR. Statistical significance was calculated using the paired t-test. d Differences in circ0000082 expression between various human CM cells (SK-Mel-2, SK-Mel-28, A2058 and A375) and noncancerous HaCaT cells. GAPDH was used as an internal reference for performing qRT‒PCR. Group differences compared with HaCaT cells were analyzed using one-way ANOVA, followed by Dunnett’s multiple comparisons test. Data are presented as the means ± SEMs from three independent experiments. e Location information of circ0000082 (hereafter referred to as circROR1) within the human genome, which is based on the host gene ROR1. f qPCR analysis of circROR1 and ROR1 expression after RNase R digestion. g qPCR analysis of circROR1 and ROR1 expression after A375 cells were cultured with actinomycin D for 0, 4, 8, 12, or 24 hours. Statistical significance was calculated using two-way repeated-measures ANOVA. h Nuclear and cytoplasmic fractionation assays were used to assess the ratio of circROR1 in the cytoplasmic fraction of A375 cells. Actin mRNA and U6 RNA served as subcellular localization markers, with actin mRNA primarily located in the cytoplasm and U6 RNA predominantly enriched in the nucleus. i FISH was used to identify the location of circROR1 in A375 cells. j Representative FISH images of circROR1 in primary CM tissues, metastatic CM tissues, and normal tissues from a tissue microarray. k Box plots showing the expression of circROR1 between normal skin tissues and CM tissues. Statistical significance was determined by the Mann–Whitney U test. l Heatmap displaying the associations of the clinical characteristics of tumors (n = 30) with high and low expression of circROR1. Statistical significance was determined by the chi-square test. m Box plots showing the expression of circROR1 by the presence of metastasis or AJCC stage as a clinical feature. Statistical significance was determined by the Mann‒Whitney U test. NS: not significant, *P < 0.05, ** P < 0.01, and *** P < 0.001. CM, cutaneous melanoma; ANOVA, analysis of variance; NS, not significant; ND, not detected; FISH, fluorescence in situ hybridization; TMA, tissue microarray; AJCC, American Joint Committee on Cancer

Circ0000082 (hereafter referred to as circROR1) is derived from chromosome 1: 64515362–64516388 and consists of two adjacent exons in the ROR1 gene (Fig. 1e). According to annotations from circBase and circPrimer2.0, it is classified as an exonic-type circRNA formed by back-splicing of exon 2 and exon 3 of the ROR1 gene (Supplemental Table S2) [38, 39]. To confirm the ring structure of circROR1, RNase R digestion and actinomycin D RNA stability assays were performed in A375 and A2058 cells. After RNase R digestion, the expression of circROR1 remained nearly unchanged, whereas ROR1 expression decreased significantly (Fig. 1f, Supplementary Fig. S1d). Similarly, treatment with actinomycin D, which globally inhibits transcription, including that of ROR1 pre-mRNA, slightly reduced the half-life of circROR1 compared with that of ROR1, as determined by qPCR. Although this approach does not isolate the effect of precursor depletion, the relatively delayed reduction in circROR1 expression is consistent with its greater stability under transcriptionally repressed conditions (Fig. 1g, Supplementary Fig. S1e). Additionally, nuclear and cytoplasmic fractionation assays confirmed that circROR1 was present in both the nuclear and the cytoplasmic fractions (Fig. 1h, Supplementary Fig. S1f). RNA fluorescence in situ hybridization (FISH) with a Cy3‐labeled circROR1‐specific probe targeting the junction site produced consistent results (Fig. 1i, Supplementary Fig. S1g). Together, these data suggest that circROR1 has a stable circular structure.

To assess the association between circROR1 and metastasis, we performed FISH on a tissue chip containing 12 metastatic CM tissues, 18 nonmetastatic CM tissues, and 7 normal skin tissues. This analysis revealed a greater frequency of positive circROR1 staining in CM tissues than in normal skin tissues (Fig. 1j‒k, P <0.001). Additionally, circROR1 was significantly upregulated in metastatic tumors compared to nonmetastatic tumors (Fig. 1l‒m, P <0.01). Detailed statistical analysis of clinicopathological features in relation to circROR1 expression is presented in Supplementary Table S3. These results indicate that circROR1 is abundant and generally upregulated in melanoma tissues and that its expression is correlated with tumor metastasis.

CircROR1 inhibits apoptosis and promotes CM cell progression, metastasis, and EMT

To explore the biological function of circROR1 in CM cells, we constructed A375, A2058, and B16 cells stably transduced with lentiviruses expressing the circROR1 overexpression vector (OV-circ) or knockdown vector (KD-circ). qRT‒PCR confirmed significant upregulation or downregulation of circROR1 expression in the transduced A375, A2058, and B16 cells, respectively, whereas no significant regulatory effect on ROR1 mRNA was observed (Supplementary Fig. S2a‒b).

Then, RNA-seq and differential expression analysis were performed on circROR1-overexpressing A375 cells and their respective control cells. A total of 1,405 differentially expressed genes (DEGs) were identified (fold change ≥ 2 or fold change < 0.8 and* P* < 0.05; Supplementary Table S4). GO enrichment analysis of the upregulated genes using Metascape and OmicShare, revealed that these genes were enriched primarily in pathways such as “cell–cell adhesion”, “cell junction organization”, and “positive regulation of cell migration”, among others (Fig. 2a, Supplementary Table S5).Fig. 2. CircROR1 inhibits apoptosis and promotes CM cell progression, metastasis, and EMT.** a** The heatmap displays the top 20 GO analysis results for upregulated genes, conducted using Metascape and OmicShare (-log P value ≥2). b,** c** Wound healing assays of A375 and A2058 cells with stable differential expression of circROR1, at 0, 24, and 48 hours. The migration area was quantified using ImageJ. Statistical significance was calculated using Student’s t test. Data are presented as the means ± SEMs from three independent experiments. d,** e** Transwell assays of A375 and A2058 cells with stable differential expression of circROR1 were performed to evaluate cell migration and invasion. The cell-covered area was quantified using ImageJ. Statistical significance was calculated using Student’s t test. Data are presented as the means ± SEMs from three independent experiments. f,** g** WB analysis of N-cadherin, E-cadherin, vimentin and Snail expression in A375 cells with circROR1 knockdown or overexpression. The histogram analysis of gray values; gray value ratio = the gray value of target proteins/the gray value of GAPDH. Statistical significance was calculated using Student’s t-test. Data are presented as the means ± SEMs from three independent experiments. h Flow cytometry was used to assess apoptosis in A375 cells. In the scatter plots, PI staining is shown on the x-axis and Annexin V staining is shown on the y-axis. The quadrants represent UL, LL, UR, and LR. The percentages of cells in each quadrant are as follows: A375 OV-NC: 0.62/94.44/0.01/4.92; A375 OV-circ: 0.02/56.74/0.08/2.42; A375 KD-NC: 9.61/81.04/8.85/0.49; A375 KD-circ: 7.22/66.14/24.99/1.65(% for UL/LL/UR/LR, respectively). i. Blots are representative of three independent flow cytometry experiments. Statistical significance was calculated using Student’s t-test. j Effects of circROR1 on A375 and A2058 cell proliferation were examined by CCK-8 assay. Statistical significance of the CCK-8 results was calculated using two-way repeated-measures ANOVA. k,** l** Effects of circROR1 on A375 and A2058 cell proliferation were also examined by colony formation assay. The cell-covered area was quantified using ImageJ. Statistical significance was calculated using Student’s t test. Data are presented as mean ± SEM from three independent experiments. NS: not significant, *P < 0.05, **P < 0.01, ***P < 0.001. h, hour; OV-NC, overexpression negative control; OV-circ, circROR1 overexpression; KD-NC, knockdown negative control; KD-circ, circROR1 knockdown; N-cad, N-cadherin; E-cad, E-cadherin; Vim, vimentin

To further investigate the morphological changes associated with circROR1-mediated metastasis, we examined pseudopodium formation in A375 cells with either circROR1 overexpression or knockdown after 12 hours of culture on cell climbing sheets. As shown in Supplementary Fig. S2c, circROR1 expression significantly altered the number and length of pseudopodia, with overexpression promoting more pronounced protrusions and knockdown reducing them. Wound healing and transwell assays were performed to assess the changes in the migration and invasion of A375, A2058, and B16 cells with circROR1 overexpression or knockdown. The results demonstrated that circROR1 upregulation significantly increased the metastatic ability of CM cells, whereas the metastatic potential was markedly reduced by circROR1 knockdown (Fig. 2b‒e, Supplementary Fig. S2d‒e). EMT is known to enable melanoma progression and metastasis [40]. Thus, we further examined the expression of EMT and anti-EMT factors in CM cells with different circROR1 expression levels. In A375 cells, the overexpression of circROR1 promoted the expression of EMT-related factors, including N-cadherin, Snail, and vimentin (to varying degrees), and decreased the expression of the anti-EMT factor E-cadherin, as determined by Western blotting (WB). Additionally, compared with the negative control (NC) group, the circROR1-knockdown group presented low levels of N-cadherin, Snail, and vimentin, and high levels of E-cadherin (Fig. 2f‒g). These data suggest that circROR1 may affect CM metastasis.

In addition, flow cytometry analysis revealed a slight decrease in the apoptosis rate of A375 and A2058 cells following circROR1 overexpression, whereas a more marked increase in apoptosis was observed in A375 cells upon circROR1 knockdown (Fig. 2h‒i, Supplementary Fig. S2f). Moreover, CCK-8 assays were performed to generate growth curves, which demonstrated that circROR1 upregulation significantly enhanced the proliferation and viability of A375, A2058, and B16 cells, whereas circROR1 downregulation inhibited the growth of CM cells (Fig. 2j, Supplementary Fig. S2g). We then examined these effects in greater detail using a colony formation assay. A375, A2058, and B16 cells, either overexpressing circROR1 or transfected with a control vector, were cultured in 6-well plates. Cells were quantified for 14 days in the overexpression group and 21 days in the downregulation group. The results indicated that ectopic expression of circROR1 increased proliferation compared with the expression of the empty vector, whereas the downregulation of circROR1 expression decreased proliferation (Fig. 2k‒l, Supplementary Fig. S2h). These results demonstrate that circROR1 promotes CM cell proliferation and metastasis and inhibits apoptosis in vitro.

CircROR1 influences the growth and metastasis of CM and the PD-L1-antibody therapy response ratein vivo

To evaluate the effects of circROR1 on tumor growth in vivo, murine B16 cells were infected with OV-circ, KD-circ, or the corresponding negative control plasmids and subcutaneously injected under the dorsal skin of 4-week-old female C57BL/6 mice (Fig. 3a). The average tumor volume in the OV-circ group clearly increased from day 12 to day 16, whereas statistically significant inhibition of tumor growth was observed in the KD-circ group during the same period (Fig. 3b‒c). Similarly, the tumor weight in the OV-circ group was clearly greater than that in the OV-NC group, whereas the tumor weight in the KD-circ group was obviously lower than that in the KD-NC group at the experimental endpoint (Fig. 3d). Moreover, a significant decrease in body weight was observed from day 12 to day 16 in the OV-NC group, suggesting that these mice might be cachectic (Fig. 3e). Furthermore, to evaluate the role of circROR1 in the lung colonization phase of the metastatic cascade, the aforementioned stably transfected B16 cells were injected into the blood circulation of C57BL/6 mice through the tail vein. The lungs were collected 16 or 26 days after injection for analysis (Fig. 3f). Interestingly, more nodules were observed in the lungs of the circROR1-overexpression group, whereas fewer nodules were observed in the lungs of the circROR1 knockdown group (Fig. 3g). The body weights of C57BL/6 mice in the circROR1-overexpression group were significantly lower than those in the control group after 12 days (Fig. 3h). Hematoxylin and eosin (HE) staining also revealed that circROR1-overexpression increased the number and total volume of colonized nodes in the lungs of C57BL/6 mice. Immunohistochemical (IHC) staining of lung colonization tumor tissues revealed that circROR1-upregulation significantly increased N-cadherin, vimentin, and Snail expression but decreased E-cadherin expression (Fig. 3i‒j). Taken together, these findings indicate that circROR1 promotes tumor growth and the lung colonization phase of the metastatic progression in vivo.Fig. 3. CircROR1 stimulates the growth and metastasis of CM in vivo. a The scheme of the in vivo study involved the use of a subcutaneous xenograft model in C57BL/6 mice with B16 cells. b Tumor size of subcutaneous B16 xenograft tumors in mice from the indicated groups (OV-NC vs. OV-circ, KD-NC vs. KD-circ, n =5 mice per group). c Tumor volume in C57BL/6 mice injected with stable circROR1 overexpression or knockdown B16 cells into dorsal flank (n = 5 mice per group). d The weights of xenograft tumor tissues on day 16 from the indicated groups were measured (OV-NC vs. OV-circ, KD-NC vs. KD-circ, n = 5 mice per group). e The weights of subcutaneous xenograft C57BL/6 mice model from the indicated groups were measured (OV-NC vs. OV-circ, KD-NC vs. KD-circ, n = 5 mice per group). f The scheme of in vivo study involved the use of a lung colonization C57BL/6 mice model with B16 cells. g Number of lung colonization nodules in C57BL/6 mice injected with stable circROR1 overexpression or knockdown B16 cells via tail vein (n = 3 mice per group). h The weights of lung colonization C57BL/6 mice model from the indicated groups were measured (OV-NC vs. OV-circ, KD-NC vs. KD-circ, n = 3 mice per group). i HE staining and IHC staining for N-cadherin, E-cadherin, vimentin, Snail, HNRNPL and FOXO4 in lung tissues from C57BL/6 mice injected with stable overexpressed-circROR1 B16 cells via tail vein. j The IHC score was quantified using ImageJ. Statistical significance was calculated using Student’s t test. Data are presented as the means ± SEMs from three independent experiments. NS: not significant, *P < 0.05, **P < 0.01, and ***P < 0.001

PD-L1-antibody therapy has become a pivotal focus in the treatment of melanoma, particularly in crafting personalized treatment plans based on the tumor PD-L1 expression characteristics [41]. Therefore, we preliminarily explored whether circROR1 expression influences resistance to PD-L1-antibody therapy in CM cells. First, we investigated the relationships between the DEGs identified through RNA-seq and genes known to be involved in PD-L1 expression and the PD-1 checkpoint pathway, as outlined in the KEGG database. Our findings revealed that the overexpression of circROR1 in A375 cells was strongly linked with the expression of key checkpoint molecules crucial for PD-L1 expression, including CD274 (PD-L1), PTEN, NFκB1, NFκB2, and IFNGR2 (Supplementary Fig. S3a). Additionally, qRT‒PCR and WB assays revealed that circROR1 overexpression markedly decreased both RNA and protein levels of PD-L1 (Supplementary Fig. S3b‒c).

To evaluate the effects of circROR1 on PD-L1-antibody therapy in vivo, we engineered B16 cells expressing either OV-circ or a control plasmid. These cells were subcutaneously injected under the dorsal skin of four-week-old female C57BL/6 mice (Supplementary Fig. S3d). The tumor volumes and body weights of these mice were monitored every three days. By day 13, the tumor volume in the OV-NC group was approximately 100 mm3. Thus, the mice in both groups were administered 150 μg of either PD-L1-antibody or InVivoMAb IgGa-antibody on days 13, 16, and 19. Compared with that in the OV-NC group, the average tumor volume in the OV-NC group treated with the PD-L1-antibody was significantly lower from days 19 to 22. Conversely, the average tumor volume over time in the OV-circ group treated with PD-L1-antibody was not significantly different from that in the group treated with the IgGa-antibody (Supplementary Fig. S3e‒f). Additionally, there was a notable decrease in body weight in the OV-NC group treated with PD-L1-antibody compared with that in the control group (Supplementary Fig. S3g). On day 22, the tumor weights in the OV-NC group treated with PD-L1-antibody were significantly lower than those in the group treated with the IgGa-antibody. However, the tumor weights in the OV-circ group differed between the PD-L1-antibody and IgGa-antibody treatment groups (Supplementary Fig. S3h). Further analysis revealed that the rate of tumor growth inhibition after PD-L1-antibody treatment was significantly lower in the OV-circ group than in the OV-NC group (Supplementary Fig. S3i). These results suggest that the overexpression of circROR1 may confer resistance to PD-L1-antibody therapy in CM cells through the downregulation of PD-L1 expression.

CircROR1 physically interacts with and stabilizes HNRNPL in CM cells

The schematic diagram illustrating the strategy for exploring the mechanisms downstream of circROR1 is shown in Fig. 4a. Considering the cellular localization of circROR1 (as depicted in Fig. 1), we hypothesized that circROR1 might interact with RBPs involved in transcriptional regulation. To identify potential RBP partners of circROR1, a circRNA pull-down assay was performed using biotin-labeled circROR1 probes in A375 cells with stable overexpression of circROR1 or a negative control. The pulled-down proteins were then subjected to MS analysis (Supplementary Fig. S4a). A total of 177 differentially expressed proteins were identified (P < 0.05, Supplementary Table S6). Reactome enrichment analysis of 94 upregulated proteins by OmicShare showed that these proteins may participate in pre-mRNA maturation through “processing of capped intron-containing pre-mRNA”, “mRNA splicing−major pathway”, “mRNA splicing”, and “transport of mature mRNA derived from an intron-containing transcript” pathways, among others (Fig. 4b) [27]. Then, we obtained mRNA expression-related and prognosis-related mutations in RBPs and investigated the potential effects of RBP mutations on splicing patterns in the CM cohort from TCGA SpliceSeq. A total of 19748 different alternative splicing events (ASEs) from 853 RBPs were identified (Fig. 4a, Supplementary Table S7) [20]. Furthermore, the RBPmap platform predicted 59 RBPs with potential circROR1 binding sites (Fig. 4a, Supplementary Table S8). Overlap analysis of these predictions with the 94 upregulated proteins identified by MS analysis yielded 5 RBPs, HNRNPL, IGF2BP3, HNRNPA0, MATR3, and HNRNPM (Fig. 4c‒d).Fig. 4. CircROR1 interacts with and stabilizes HNRNPL.** a** Flowchart illustrating the selection process for circROR1 RBPs. b Reactome pathway analysis of upregulated proteins identified through circROR1 pull-down and MS assays. c Venn diagram illustrating the analysis of unique and shared RBPs between those associated with prognosis in CM, predicted RBPs based on circROR1 binding sites, and upregulated proteins identified in the MS analysis of A375 cells with circROR1 overexpression. d Volcano plot of differentially expressed proteins identified by the MS analysis of A375 cells with circROR1 overexpression; 5 RBPs selected by the overlap analysis are annotated. e qRT‒PCR analysis of HNRNPL expression between CM tissues and corresponding normal tissues (n =10). f Correlation analysis of circROR1 expression with HNRNPL mRNA expression in CM tissues and corresponding normal tissues (n = 20). g Schematic diagram and table illustrating potential circROR1 protein-binding motifs for HNRNPL, as predicted through RBPmap. h Anti-HNRNPL RIP assay showing that circROR1 was pulled down. The results were normalized to those of the input sample, and the IgG group served as the negative control. i Protein lysates prepared from circROR1 overexpression and negative control A375 cells were subjected to WB. Levels of HNRNPL were detected in the cytoplasm and nucleus. j Effect of the protein synthesis inhibitor CHX on HNRNPL protein expression in A375 cells with circROR1 overexpression or knockdown. Band intensities were normalized to the loading control (β-actin) and then to the 0-h value, which was set to 1.0 for each group. Data are presented as the means ± SEMs (n = 3 independent experiments). Statistical analysis was performed using two-way repeated-measures ANOVA followed by Tukey’s multiple comparisons test. *P < 0.05, ** P < 0.01, and *** P < 0.001. siRNA, small interfering RNA

Next, we analyzed the expression patterns of the aforementioned RBPs through in vitro experiments and in tumor samples from a subcutaneous xenograft tumor model (Supplementary Fig. S4b‒d). This analysis revealed that HNRNPL expression was increased in A375 cells overexpressing circROR1, whereas circROR1 knockdown decreased HNRNPL expression at the mRNA level. Similar results were observed in the analysis of HNRNPL mRNA expression in xenograft tumor tissues, consistent with the findings from the pull-down‒MS assays. Notably, HNRNPL is a well-documented splicing factor and has been implicated in cancer metastasis [42, 43].

Data from the GEPIA website indicated that the expression of HNRNPL in CM tissues was higher than that in normal control tissues (Supplementary Fig. S4e). [31] We performed qRT‒PCR analysis using the same CM and paired normal samples as shown in Fig. 1c to validate the bioinformatic results. The results revealed that HNRNPL expression was elevated in CM tissues and was positively correlated with circROR1 expression (Fig. 4e‒f). The RBPmap results identified 14 potential RNA protein-binding motifs for the HNRNPL protein within circROR1 (Fig. 4g, Supplementary Table S8). To verify the interaction between HNRNPL and circROR1, we performed RIP experiments using an HNRNPL-specific antibody, with IgG as a negative control, following standard protocols. Notably, GAPDH was abundant in the input sample but barely detectable in both the IgG and HNRNPL RIP samples, supporting the specificity of the RIP assay (Supplemental Fig. S4f). Under these conditions, we observed that the target RNA circROR1 was significantly enriched in the HNRNPL-RIP group compared with the IgG control group. Moreover, the abundance of circROR1 in HNRNPL-immunoprecipitated complexes was noticeably higher in the OV-circ A375 group than in the OV-NC group (Fig. 4h).

To further investigate the interaction between circROR1 and HNRNPL, protein lysates prepared from the cytoplasm and nuclei of OV-circ or OV-NC A375 cells were subjected to WB analysis. We observed increased levels of HNRNPL in the cytoplasm but decreased levels in the nucleus of circROR1-overexpressing cells (Fig. 4i). Next, we performed cycloheximide (CHX) treatment assays to determine whether circROR1 modulates the stability of the HNRNPL protein. We found that the proportion of HNRNPL gradually decreased in the OV-NC group, starting at 6 hours, whereas it remained relatively stable within 10 hours of intervention in the OV-circ group. In the KD-NC group, the HNRNPL protein signal also began to weaken at 6 hours post-treatment; however, in the KD-circ group, the decrease in the protein signal commenced earlier, at approximately 4 hours, with a noticeably sharper decrease in intensity (Fig. 4j).

The transfection efficiencies of the HNRNPL siRNA and the overexpression plasmid were evaluated using qRT‒PCR (Supplementary Fig. S4g‒h). We found that the expression of circROR1 was upregulated in CM cells transfected with the HNRNPL overexpression plasmid, but downregulated in cells transfected with si-HNRNPL (Supplementary Fig. S4i‒j). Considering the complex patterns of circRNA‒protein interactions, these results suggest that circROR1 closely associates with and promotes HNRNPL protein stability and expression. Additionally, circROR1 may reduce HNRNPL nuclear translocation through circRNA adsorption, thereby influencing its role in alternative RNA splicing [44, 45].

CircROR1 promotes CM metastasis by triggering FOXO4 isoform switching through interaction with the HNRNPL protein

As shown in Fig. 4b, the Reactome enrichment results of the circRNA pull-down‒MS assay indicated that circROR1 upregulation could regulate the mRNA alternative splicing activities of RBPs. CircRNAs may regulate mRNA alternative splicing through interactions with the HNRNPL protein and further promote CM metastasis. To verify this hypothesis, we collected possible splicing events regulated by HNRNPL in CM from the TCGA SpliceSeq database (Supplementary Table S9). [20] To find downstream DEGs related to the inhibition of CM metastasis, we extracted the GSE8401 and GSE46517 datasets from the GEO database (Supplementary Table S10) [29, 30]. A Venn diagram showing the overlap of GSE8401, GSE46517, and the HNRNPL AS events identified FOXO4 (Fig. 5a, Supplementary Table S10). Volcano plots showing FOXO4 expression in each GSE dataset are shown in Fig. 5b‒c.Fig. 5. CircROR1 promotes CM metastasis by triggering FOXO4 isoform switching through interactions with HNRNPL protein. a The Draw Venn Diagram online tool generated the Venn diagram. b Volcano plot of DEGs identified in GSE8401. c Volcano plot of DEGs identified in GSE46517. d Images of IHC staining for FOXO4 in primary CM tissues and metastatic CM tissues from tissue chip. e Comparison of FOXO4 IHC scores between primary CM tissues and metastatic CM tissues. f Correlation analysis of circROR1 FISH score with FOXO4 IHC score in CM tissues from tissue chip. g Schematic diagram of the genomic location and splicing pattern of FOXO4 gene and its transcriptional products. h The possible alternative splicing pattern regulated by HNRNPL in CM. i qRT‒PCR and gel electrophoresis analysis of the inclusion of FOXO4 exon 1 in FOXO4α-upregulated A375 cells. j WB analysis of FOXO4 protein expression in FOXO4α-upregulated A375 cells. k Transwell migration and invasion assays of FOXO4α-upregulated A375 cells. l qRT‒PCR analysis of the mRNA expression of EMT and anti-EMT factors in FOXO4α-upregulated A375 cells. m,** n** WB analysis of EMT and anti-EMT factors protein expression in FOXO4α-upregulated A375 cells. o,** p** qRT‒PCR and gel electrophoresis analysis of FOXO4α and FOXO4ζ expression in circROR1 overexpressed or HNRNPL knocked-down A375 cells. WB analysis of FOXO4 protein expression in A375 cells with different levels of circROR1 or HNRNPL expression. q Gel electrophoresis and quantitative analysis of FOXO4α and FOXO4ζ mRNA expression in the HNRNPL overexpressed cells transfected with circROR1 overexpressed vector or negative control vector. r Gel electrophoresis and quantitative analysis of FOXO4α and FOXO4ζ mRNA expression in the circROR1 overexpressed cells transfected with plasmids expressing FOXO4α or negative control. s Transwell migration and invasion assays of circROR1 overexpressed A375 cells transfected with plasmids expressing FOXO4α or negative control. Blots are representative of three independent experiments. GAPDH was used as an internal reference for performing qRT‒PCR. Statistical significance was calculated by Student’s t test. Data are presented as mean ± SEM from three independent experiments. *P < 0.05, ** P < 0.01, and *** P < 0.001. OE-NC, negative control vector; OE-HNRNPL, HNRNPL overexpression vector; OE-FOXO4α: FOXO4α overexpression vector, circROR1 vector, circROR1 overexpression vector; NC-vector, negative control vector

Forkhead box O4 (FOXO4), a member of the FOXO family, has been reported to act as a tumor progression and metastasis suppressor in some cancer types, such as colorectal, gallbladder and prostate cancer, but the role of FOXO4 in CM metastasis has not been well characterized [46, 47]. To confirm the relationship of FOXO4 expression in tumors with CM metastasis, we performed IHC on the same tissue microarray containing 12 metastatic CM tissues, and 18 nonmetastatic CM tissues. Images of IHC staining and scoring revealed more significant downregulation of FOXO4 in metastatic tumors than in nonmetastatic tumors (Fig. 5d‒e,* P* < 0.01). Moreover, the FOXO4 IHC scores were significantly negatively correlated with the circROR1 FISH scores (Fig. 5f, P = 0.018).

As shown in Fig. 5g and Table S8, the FOXO4-8939-Retained Intron (RI) was identified as the possible alternative splicing pattern regulated by HNRNPL (P = 0.0000119). The splicing pattern of FOXO4 gene (also called AFX) and its transcriptional product have been analyzed in previous research. As described by Peters U et al. and Yang Z et al., oligonucleotides for FOXO4 amplification were deposited in GenBank (GenBank accession number Y11284), and two splice variants of the FOXO4 gene were observed, namely, FOXO4α and FOXO4ζ [35, 48]. The mRNA encoding the FOXO4α isoform corresponds to a fusion of GenBank Y11284:1–694, Y11285:101–1153, and Y11286:103–1517. The FOXO4ζ isoform results from a cryptic intron within exon 1 (Y11284:414–578), with a fusion of Y11284:1–413, 579–694, Y11285:101–1153, and Y11286:103–1517 (Fig. 5h).

Given the implication of HNRNPL in the regulation of alternative splicing, we next investigated whether the interaction of circROR1 with HNRNPL could regulate FOXO4 alternative splicing and then promote CM metastasis. To this end, we performed qRT‒PCR using unique primers that generate a 424-bp product for FOXO4α and a 259-bp product for FOXO4ζ (the primers are listed in the Supplemental material). Furthermore, we overexpressed FOXO4α mRNA using an isoform-specific FOXO4α overexpression vector (OE-FOXO4α) (Supplementary Table S1). The effectiveness of OE-FOXO4α was verified via PCR-gel electrophoresis and WB (Fig. 5i‒j). The specific upregulation of the FOXO4α isoform inhibited migration, invasion, and the expression of EMT-related factors in A375 cells (Fig. 5k‒n). These results indicated that FOXO4α could inhibit the metastatic behavior of CM cells.

As a next step, we analyzed the inclusion of FOXO4 intron 1 in exon 1, which increases the ratio of FOXO4α to FOXO4ζ, in A375 cells expressing different levels of circROR1 or HNRNPL. Compared with control cells, OV-circ A375 cells expressed the mRNA encoding FOXO4α at higher levels but expressed the mRNA encoding FOXO4ζ at lower levels. A similar trend was also observed in A375 cells transfected with si-HNRNPL. We then tested the regulatory effects of circROR1 and HNRNPL on FOXO4 protein synthesis. HNRNPL knockdown in A375 cells significantly decreased FOXO4 protein expression; similarly, circROR1 overexpression in A375 cells reduced FOXO4 expression (Fig. 5o‒p). We then transfected the circROR1 overexpression vector (circROR1 vector) or negative control vector (NC vector) into A375 cells overexpressing HNRNPL and observed that the upregulation of circROR1 weakened the effects of HNRNPL on FOXO4α expression (Fig. 5q). We subsequently transfected OE-FOXO4α into A375 cells with elevated circROR1 expression and observed that the upregulation of FOXO4α could reverse the upregulation of cell migration and invasion induced by circROR1 overexpression (Fig. 5r‒s). Collectively, our findings suggest that switch in splicing mediated by the circROR1/HNRNPL interaction decreases FOXO4α expression, whereas the upregulation of FOXO4α could counteract this effect and inhibit the metastatic behaviors of CM cells.

PEGylated siRNA lipoplexes targeting circROR1 affect siRNA biodistribution and gene-silencing effectsin vitro

Efficient intracellular delivery of siRNA plays a critical role in the RNA interference process [49, 50]. Given the critical role of circROR1 upregulation in CM, we sought to develop an efficient nanotechnology-based delivery vehicle for siRNA targeting circROR1 as a potential therapy for CM. The knockdown efficiency and in vivo effect of si-circ delivered by Lipo3000 were analyzed by qRT‒PCR which showed that si-circ-2 significantly reduced circROR1 expression and inhibited A375 cell proliferation (Fig. 6a‒b) as well as migration and invasion (Fig. 6c‒d) in vitro (see also Supplementary Fig. S5a).Fig. 6. In vivo delivery of anti-CircROR1 suppresses melanoma metastasis. a,** b** Inhibited effects of si-circROR1-2 on A375 cell proliferation were examined by CCK-8 assay and colony formation assay. The cell-covered area of colony assays was quantified using ImageJ. c,** d** The inhibitory effects of si-circROR1-2 on A375 cell migration and invasion were examined by transwell assay. The cell-covered area was quantified using ImageJ. e The scheme involved establishing a lung colonization nude mouse model using luciferase (Luc)-expressing A375 cells. IVIS monitoring of lung colonization was performed on the fourteenth day, coinciding with the initiation of FA-PEG(siRNA) injections. The treatment concluded fourteen days later. f The critical reagents and synthesis protocol of positively charged lipoplexes for FA-PEG. g,** h** Gel retardation electrophoresis of FA-PEG and siRNA. FA-PEG and siRNA were mixed at different N/P ratios to form complexes, which were then analyzed using gel electrophoresis. Lane 1: DNA marker; Lane 2: PBS; Lanes 3–8: N/P ratios of 1, 2, 4, 6, 8 and 10; Lane 9: Lipo 3000(siRNA); Lane 10: Only FA-PEG; Lane 11: Only siRNA; Lane 12: DNA marker. i Representative transmission electron microscopy (TEM) images of FA-PEG(si-circ). j Uptake of Lipo3000(si-circ) and FA-PEG(si-circ) into A375 cells. The siRNA was labeled with Cy5 dye (red), and nuclei were counterstained with 4,6-diamidino-2-phenylindole (DAPI, blue). Scale bars represent 100 nm. k Biodistribution of FA-PEG(si-circ) complex in vivo. BALB/c nude mice bearing A375 lung metastases were treated with different Cy5-siRNA formulations via tail vein injection. Fluorescent images of mice were obtained with in vivo fluorescence imaging system after 6 h (n = 4). l Representative images of metastases in whole lungs from lung colonization nude mice using Luc-A375 cells after 14-day treatment of FA-PEG(si-NC), free siRNA, or FA-PEG(si-circ) (n = 4). The numbers of metastatic modules from the indicated groups are shown. m Bioluminescence images of colonization in whole mice and lungs before and after tail vein injection of FA-PEG(si-NC), free siRNA, or FA-PEG(si-circ). The luminescence values of pulmonary nodules from the indicated groups are shown. n, o HE staining of tissue sections from major organs (lung, liver, and kidney) of mice was performed after the treatment. Statistical significance was calculated by Student’s t test. Data are presented as the means ± SEMs from three independent experiments. *P < 0.05, ** P < 0.01, and *** P < 0.001. ND, not detected; FA-PEG, DSPE-PEG2000-FA nanoliposomes; Luc, luciferase; NS, not significant

The selective delivery of nanoformulations can also be improved using specific ligand‒receptor interactions followed by binding and uptake through receptor-mediated endocytosis. Folate receptor α (FRα), a glycosylphosphatidylinositol glycoprotein that is anchored to the cell surface, is overexpressed in various epithelial malignant cancers, including CM, but is expressed at only limited levels in normal tissues [51, 52]. In our research, we established a lung colonization in nude mouse model using stable luciferase (Luc)-expressing A375 cells to further validate the ability of si-circROR1 to inhibit CM metastasis (Fig. 6e, Supplementary Fig. S5b). Lung tissues with metastases were collected, and the percentage of FRα-positive cells was analyzed using IHC. The percentage of FRα-positive cells was significantly greater in metastatic lung tissue than in paracancerous lung tissue, which is consistent with similar findings in previous studies on cancer (Supplementary Fig. S5c) [51, 52]. Thus, in our research, DSPE-PEG2000-FA was used as a stabilizer to prepare siRNAs targeting circROR1 (si-circ) and meant to provide the resultant PEGs with the ability to activate target FR-overexpressing tumors and long blood circulation times. The N/P ratio is defined as the molar ratio of the nitrogen content of DSPE-PEG2000-FA nanoliposomes (hereafter referred to as FA-PEG) to the phosphorus content of the anionic siRNA. To determine the optimal N/P ratio for FA-PEG(si-circ) complex synthesis, lipoplexes of FA-PEG formulations were prepared with si-circ at different N/P ratios (1 to 10), using a concise and scalable microfluidic-mediated strategy for one-step formation of the targeted liposomes (Fig. 6f) [33]. Electrostatic interactions between the negatively charged siRNA and cationic liposomes as a function of N/P ratio were assessed by the electrophoretic gel retardation assay. Complete retardation of si-circ migration was observed in all the FA-PEG(si-circ) complexes with an N/P ratio of 8 or higher (Fig. 6g‒h). Transmission electron microscopy was performed to examine the structural characteristics of the complex and revealed that the FA-PEG(si-circ) particles possessed a compact core‒shell morphology with a uniform diameter of approximately 70 nm (Fig. 6i). We further evaluated the serum stability of the complex by incubating free siRNA and FA-PEG(si-circ) (equivalent to 100 nM siRNA) in DMEM supplemented with 10% FBS for 0, 2, 4, and 6 hours at 37 °C. After incubation, siRNA integrity was analyzed by agarose gel electrophoresis. The results indicated that siRNA complexed with FA-PEG remained stable with minimal degradation for up to 6 hours, whereas free siRNA exhibited substantial degradation within 2 hours. These findings confirm that FA-PEG(si-circ) protected the siRNA from nuclease-mediated degradation in serum (Supplementary Fig. S5d).

To further evaluate the cellular uptake efficiency of the different delivery systems, we compared the internalization of FA-PEG(si-circ) and Lipo3000(si-circ) (100 nM siRNA equivalent) in A375 cells. After incubation with different Cy5-siRNA formulations for 6 h, FA-PEG(si-circ) entered the A375 cells in slightly greater quantities than Lipo3000(si-circ) (P =0.0632). The results suggested that, to some extent, FA-PEG(si-circ) complex promoted siRNA internalization and increased cellular uptake compared with Lipo3000(si-circ) (Fig. 6j). We also tested the biodistribution of free si-circ and FA-PEG(si-circ) 6 h after intravenous injection of free siRNA and FA-PEG(si-circ) (equivalent to 5 nmol of siRNA per mouse). Fluorescence imaging indicated that FA-PEG(si-circ) exhibited better targeting of tumor sites than free siRNA did (Fig. 6k).

To further explore the CM cell inhibitory effects mediated by FA-PEG(si-circ), we extended our studies to other lung colonization models in nude mice (Fig. 6e). Two weeks after the tail vein injection of A375-luc cells, an in vivo imaging system (IVIS) was used to track tumor metastasis. FA-PEG(si-NC), free si-circ and FA-PEG(si-circ) (equivalent to 5 nmol of siRNA per mouse) were subsequently injected into the mice via the tail vein [34]. After another 2 weeks of treatment, the mice were imaged through bioluminescence imaging to visualize the lung colonization, and the main organs were collected. The results revealed that, compared with free siRNA and FA-PEA(si-NC) groups, the FA-PEG(si-circ) group exhibited a more significant decrease in tumor colonization and a more significant increase in body weight (Fig. 6l‒m, Supplementary Fig. S5e). HE staining of the tumor sections revealed few metastatic foci and obvious necrotic areas in the FA-PEG(si-circ) group (Fig. 6n).

The viability of HaCaT and SV-HUC cells after incubation with FA-PEG for 24 h or 48 h was measured by performing a CCK-8 assay. Treatment with 0 to 200 μg/100 μL FA-PEG for 24 h or 48 h did not significantly decrease the cell viability. However, treatment with 250 μg/100 μL FA-PEG markedly decreased the survival rates of HaCaT cells at 24 h or 48 h (Supplementary Fig. S5f). These results indicated that FA-PEG(siRNA) has low cytotoxicity and is suitable for biomedical applications. To evaluate the toxicity and side effects of the FA-PEG(siRNA) complex in vivo, the HE staining was performed in liver and kidney tissue sections collected from mice after multiple dosing treatments. The results revealed no noticeable damage or pathological changes between the FA-PEG(siRNA) group and the control groups (Fig. 6o). The results also demonstrate the low cytotoxicity and good biocompatibility of FA-PEG(siRNA), highlighting its potential for biomedical applications.

To investigate the broader oncogenic relevance of circROR1 and evaluate the potential utility of FA-PEG(si-circ) across tumor types, we assessed circROR1 expression in a panel of cancer cell lines using qRT‒PCR. Notably, circROR1 expression was markedly elevated in bladder cancer cell lines (5637, SW780, and EJ) compared with nontumorigenic SV-HUC ureteral epithelial cells. Elevated levels of circROR1 were also observed in tumor-derived cell lines including Cal27 and SCC4 (tongue squamous cell carcinoma), Caco-2, SW480 (colorectal adenocarcinoma), and KHM-5M (thyroid carcinoma) cells, compared with the normal skin keratinocyte line HaCaT (Supplementary Fig. S5g). These results suggest that FA-PEG(si-circ) may have treatment value in targeting circROR1 in various cancers.

Discussion

CircROR1 is a circular RNA encoded on chromosome 1: 64515362–64516388 and comprises two adjacent exons within the ROR1 gene. To elucidate the specific role of circROR1, we conducted a series of clinical sample analyses as well as invivo and in vitro experiments. We identified circROR1 as a key promoter of the expression of EMT-related factors and demonstrated its role in enhancing CM cell migration and invasion. The underlying mechanism appears to involve interaction with the splicing factor HNRNPL, which regulates FOXO4α mRNA expression. Furthermore, we developed a nanoliposome-based drug delivery system (FA-PEG(si-circ)) containing siRNA against circROR1. This approach effectively targeted circROR1 and inhibited CM metastasis in vivo. Thus, circROR1 represents a potential biomarker for CM metastasis, and targeting circROR1 with siRNA delivery is a promising strategy for CM treatment.

To date, many studies have shown that cancer metastasis involves EMT and cellular dysregulation to achieve the migration of tumor cells from primary locations to adjacent organs, and this process involves the multifaceted roles of circRNAs as epigenetic factors [53]. In the present study, we show that circROR1 promotes CM cell progression and metastasis in vivo and in vitro, highlighting the role of this circROR1 in cancer biology, and its potential as a therapeutic target and biomarker. Many other studies have also highlighted the role of circRNAs in promoting tumor metastasis. For example, the upregulation of circ-101,882 is beneficial for promoting the malignancy and progression of gastric tumor cells. Circ-101882 enhances cancer cells invasiveness and promotes vimentin, N-cadherin, and Snail expression while decreasing E-cadherin expression to stimulate EMT [54]. Our research involved similar experiments. More research has focused on the interactions between circRNAs and miRNAs. Circ-0030018 is an oncogenic factor in esophageal cancer and sponges miR-599 to increase the expression of ENAH. The positive association of circ-0030018 with ENAH expression is vital for inducing EMT and increasing tumor cell invasion and metastasis [55]. These observations led us to hypothesize that circROR1 may regulate HNRNPL and EMT in other ways.

The RNA-binding protein HNRNPL is a core component of the exon junction complex and is considered an important regulator of posttranscriptional processes including mRNA splicing, transport, translation, and surveillance [15, 18]. In the present study, we discovered that elevated levels of circROR1 expression can bind more HNRNPL protein, which in turn exacerbates the expression of EMT-related factors and enhances cell migration and invasion abilities. Research has shown that some circRNAs are regulated during EMT through interactions with various factors [40]. For example, EIF4A3 binds to the MMP9 mRNA, facilitating circMMP9 cyclization and enhancing circMMP9 expression in glioblastoma multiforme [56]. Circ-0001756 promotes RAB5A expression by inducing IGF2BP2 expression. RAB5A then stimulates the EGFR/MAPK axis to mediate EMT, thereby enhancing the progression and metastasis of ovarian tumor cells [57]. The unique structure of circRNAs determines their multifaceted and complex interactions with splicing factors and further influences downstream molecule maturation, making it a valuable subject for further investigation [6].

While we elucidated the oncogenic role of circROR1 and its molecular mechanisms in CM both in vitro and in vivo, several limitations warrant discussion. First, our data indicate that circROR1 is significantly associated with HNRNPL and that the stability of HNRNPL is reduced in A375 cells with circROR1 knockdown. HNRNPL, a splicing factor, may also influence the maturation and formation of circROR1. The interaction between circROR1 and HNRNPL may represent a positive feedback loop that further promotes tumor progression. Moreover, this research focused on FOXO4 splicing, but the circROR1–HNRNPL complex likely regulates multiple splicing targets that contribute to the observed phenotypes, which remains to be further explored. In addition, the molecular mechanism underlying the association of this complex with specific regions of the FOXO4 pre-mRNA is not yet clear. Moreover, although the expression of FOXO4ζ is lower than that of FOXO4α, the function of FOXO4ζ requires further investigation. Additionally, although in vivo bioluminescence imaging was performed to dynamically monitor tumor progression, the changes in circROR1 expression in lung tumor tissues require further validation. Finally, in terms of animal model selection, we utilized an experimental (nonspontaneous) metastasis model, which simulates the final step of the metastatic cascade—lung colonization—rather than the complete process of spontaneous dissemination from a primary tumor site.

In conclusion, our study identified circROR1 is an oncogenic circular RNA that plays a crucial role in tumor progression and metastasis. The mechanism underlying its function involves interaction with the splicing factor HNRNPL, which regulates FOXO4 mRNA expression. This study is the first to demonstrate a direct and effective approach for suppressing metastatic cancer by targeting circROR1, via the FA-PEG(si-circ) complex. Together, these findings suggest that circROR1-targeted nanotherapy is a promising option for the treatment of metastatic cancer (Fig. 7).Fig. 7. Schematic diagram of the overall experimental design. CircROR1 binds to the splicing factor HNRNPL, promoting its nuclear translocation and regulating the alternative splicing of FOXO4 pre-mRNA. Upregulation of circROR1 suppresses the generation of the FOXO4α isoform while promoting FOXO4ζ expression, thereby inducing the metastatic potential of CM cells. To achieve circROR1-targeted therapy, a folate-modified polyethylene glycol lipid nanoparticle complex [FA-PEG(si-circ)] was developed using a microfluidic strategy for efficient siRNA delivery and stable gene silencing, effectively suppressing melanoma metastasis in vivo

Supplementary Information

Supplementary Material 1. Supplementary Material 2.