Circular RNA circBNC2 inhibits tumorigenesis by modulating ferroptosis and acts as a nanotherapeutic target in prostate cancer

Abstract

Background

Metastasis is a leading cause of cancer-related death in castration-resistant prostate cancer (CRPC) patients. Circular RNAs (circRNAs) have emerged as key regulators of the metastasis of various cancers. However, the functional effects and regulatory mechanisms of circRNAs in metastatic CRPC (mCRPC) remain largely unknown.

Methods

The expression of circBNC2 in prostate cancer (PCa), CRPC and neuroendocrine prostate cancer (NEPC) tissues was analyzed through bioinformatics analysis. Functional assays, including cell proliferation, migration, invasion and ferroptosis, were conducted in vitro and in vivo. The interactions between circBNC2, miR-4298, and ACSL6 were explored via luciferase reporter assays, RNA immunoprecipitation, and western blotting analysis. In addition, for the first time in PCa, we developed novel nanobowls (NBs) loaded with docetaxel (DTX) and circBNC2 (Dc-NBs) and evaluated the antitumor efficacy of Dc-NBs in a photothermal therapy (PTT) strategy.

Results

We identified a novel tumor-suppressive circRNA, circBNC2, in human PCa, CRPC and NEPC samples via bioinformatic analysis. CircBNC2 expression was significantly downregulated in PCa tissues and PCa cell lines. Functional assays demonstrated that circBNC2 inhibited PCa cell proliferation and migration both in vitro and in vivo. Mechanistically, circBNC2 acted as a sponge for miR-4298, and ACSL6 was identified as a direct target of the circBNC2/miR-4298 axis. Moreover, we demonstrated that ACSL6 is essential for mediating circBNC2-regulated ferroptosis in PCa cells. More importantly, we demonstrated the nanodelivery of Dc-NBs, which exhibited significant antitumor effects in both subcutaneous and metastatic PCa models.

Conclusion

This study revealed the tumor-suppressive role of circBNC2 in mCRPC by driving ferroptosis via the circBNC2/miR-4298/ACSL6 axis. Additionally, we developed an efficient and safe PTT strategy based on a nanodelivery system that codelivers circBNC2 and DTX, highlighting its potential as a novel therapeutic approach for mCRPC.

Introduction

The incidence of prostate cancer (PCa) ranks first among all urinary cancers in men, and PCa has become a worldwide health burden [1, 2]. PCa can be treated if it is detected early; however, most cases are diagnosed when the cancer has spread to the bone or lymph nodes, resulting in a poor prognosis [3,4,5]. Different chemotherapeutic drugs, such as mitoxantrone, docetaxel, cabazitaxel, and various combinations of platinum-based anticancer drugs, have been explored for the treatment of metastatic castration-resistant prostate cancer (mCRPC) [6,7,8]. However, all current PCa treatments have limitations, such as resistance to therapy, local relapse, limited drug bioavailability, and side effects on local tissues [9]. Therefore, finding new molecular targets for the diagnosis and treatment of PCa is highly important.

CircRNAs constitute a family of naturally occurring long noncoding RNAs that are prevalently expressed in eukaryotes [10]. It was subsequently discovered that circRNAs exist in the cytoplasm of eukaryotic cells and are expressed in a cell- and organ-specific fashion with significant biological functions [11]. CircRNAs possess a distinctive covalent single-stranded closed-loop configuration, lacking a 5’ cap or 3’ poly (A) tail [12]. Recent investigations have demonstrated that circRNAs play crucial roles in numerous cancers by functioning as microRNA sponges, interacting with RNA-binding proteins, and regulating gene transcription, alternative splicing, and protein translation [13]. More importantly, an increasing number of studies have demonstrated that circRNAs are abnormally expressed and involved in the occurrence and development of different types of cancers [14, 15]. Owing to the resistance of circRNAs to RNase R [16], they may play a role in disease diagnosis and treatment as stable new biomarkers or potential therapeutic targets in cancers [17]. In particular, an increasing number of reports have demonstrated that circRNAs are involved in regulating the invasion, migration and metastasis of PCa [18, 19].

Ferroptosis is a novel form of programmed cell death caused by the accumulation of iron-dependent lipid peroxides [20]. Ferroptosis is involved in the pathophysiological processes of various diseases, including cancers, and can act as a natural barrier to tumor progression [21]. The induction of ferroptosis as a new type of anticancer “medicament” has been widely reported in melanoma [22], hepatocellular [23], gastric [24], ovarian [25], pancreatic [26], breast [27] and colorectal cancers [28]. More intriguingly, an increasing amount of evidence has demonstrated that circRNAs play important regulatory roles in cancer progression via the ferroptosis pathway and might become new diagnostic markers or therapeutic targets of cancers [29, 30]. In hepatocellular carcinoma cells, circPIAS1 overexpression inhibited ferroptosis by competitively binding to miR-455-3p, leading to the upregulation of nuclear protein 1 (NUPR1). NUPR1 subsequently promotes FTH1 transcription, enhancing iron storage in HCC cells and conferring resistance to ferroptosis [31]. More intriguingly, it has been reported that circVPS8 functions as a scaffold, binding to both MKRN1 and SOX15, thereby facilitating the ubiquitination of MKRN1 and the consequent degradation of SOX15. Owing to competitive binding, the ubiquitination capacity of MKRN1 with respect to HNF4A is diminished, resulting in increased HNF4A expression. Increased HNF4A expression, in conjunction with decreased SOX15 expression, synergistically inhibits ferroptosis in glioblastoma [32]. However, the role of circRNAs in regulating PCa progression and metastasis via ferroptosis remains largely unknown.

In the present study, we identified a novel circular RNA, circBNC2 (circBase ID: hsa_circ_0008732), which is significantly downregulated in patients with PCa, especially in those with CRPC and NEPC. Additionally, through both in vitro and in vivo experiments, we demonstrated that circBNC2 can act as a molecular sponge for miR-4298, thereby targeting ACSL6, activating the ferroptosis pathway, and ultimately exerting a tumor-suppressive effect in PCa cells. Moreover, we developed nanoparticles loaded with both circBNC2 and DTX (DTX/circBNC2-co-loaded OA-Gd/PDA NBs, Dc-NBs). The delivery of Dc-NBs significantly attenuated tumor progression and lung metastasis in a PCa cell-xenografted nude mouse model upon photothermal therapy (PTT). Taken together, our study demonstrates that circBNC2 is a potential biomarker for PCa and provides a promising nanotherapeutic strategy for metastatic PCa treatment.

Materials and methods

Bioinformatic analysis

The online tools circBank [33], miRDB [34] and TargetScan [35] were used to predict potential interactions between circRNAs and miRNAs. mirDIP, TargetScan, miRPathDB and miRWalk were utilized to predict miRNA target genes. PRAD cohort datasets from The Cancer Genome Atlas (TCGA) (https://www.cancer.gov/ccg/research/genome-sequencing/tcga), Gene Expression Profiling Interactive Analysis (GEPIA) (http://gepia.cancer-pku.cn/), and the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/) were analyzed to compare gene expression levels between tumor and normal tissues and to assess gene correlations. Gene set enrichment analysis (GSEA) was conducted via the R package “Cluster Profile” to investigate biological functions in circBNC2-overexpressing (OE-circBNC2) and control (OE-Ctrl) PCa cells.

Tissue samples

Seventy-eight pairs of freshly obtained PCa tissues and matched adjacent paracancerous normal prostate tissues were obtained from the Department of Urology of the Fourth Affiliated Hospital of Harbin Medical University for histological analysis. All the tissues were stored in liquid nitrogen until further use. The medical records of the patients were collected, and the following information was obtained: age, sex, pathological stage, T stage, lymph node metastases and distant metastasis. The study was approved by the Institute Research Ethics Committee of the Fourth Affiliated Hospital of Harbin Medical University (2021-WZYSLLSC-31).

Cell culture

Normal human prostate epithelial cells (RWPE-1) and human PCa cell lines (LNCaP, 22Rv1, VCaP, C4-2, DU145, and PC3) were purchased from American Type Culture Collection and authenticated by STR profiling. RWPE-1 cells were cultured in K-SFM (Invitrogen, Carlsbad, CA, USA) medium. VCaP and C4-2 cells were cultured in DMEM (Invitrogen, Carlsbad, CA, USA). LNCaP and 22RV1 cells were cultured in 1640 (Invitrogen, Carlsbad, CA, USA) medium. PC3 and DU145 cells were cultured in F-12 K (Invitrogen, Carlsbad, CA, USA) and MEM (Invitrogen, Carlsbad, CA, USA) media, respectively. All of the culture media were supplemented with 10% foetal bovine serum (FBS) (Invitrogen, Thermo Fisher Scientific, Inc.) and 1% penicillin‒streptomycin (Invitrogen, Carlsbad, CA, USA). All the cell lines were cultured at 37 ℃ in a 5% CO₂ atmosphere. Bimonthly PCR assays were conducted to confirm the absence of Mycoplasma contamination.

Oligonucleotide synthesis and transfection

The oligonucleotides of the miR-4298 mimic/inhibitor and control (NC mimic/inhibitor) were synthesized by RiboBio (Guangzhou, China). Oligonucleotides were transfected via RiboFECT™ CP Reagent (RiboBio, Guangzhou, China) following the manufacturer’s instructions.

Confirmation of the circular structure

Divergent and convergent primers were designed for circRNA validation. The design strategies for divergent or convergent primers can be found in the Supplementary materials and methods. The cDNA samples were retrotranscribed from the total RNA treated with DNase I (Invitrogen, USA) and RNase R (ab286929; Abcam), and the genomic DNA was used as a control. Convergent primers were used as positive controls for linear transcripts, and divergent primers were used to confirm the presence of circular templates. Approximately 20 ng of cDNA or genomic DNA was used with Taq DNA polymerase and 10 × buffer (Takara, Dalian, China) for each PCR amplification, which was performed under the following conditions: 95 ℃ for 3 min; 35 cycles of 94 ℃ for 60 s; 55℃ for 30 s, and 72℃ for 30 s. The PCR products were subjected to gel electrophoresis analysis. The Super DNA Marker was purchased from CWBIO Biotech (Jiangsu, China). The bands were visualized by ultraviolet (UV) irradiation. The bands with sizes similar to those of the expected bands were dissected and purified via an AXYGEN Gel Extraction Kit (Qiagen, CA, USA). The PCR products were sequenced by Shanghai Sangon Co., Ltd.

Lentivirus infection

The lentiviruses used for circBNC2 overexpression (pLV-CMV-hsa_circBNC2-EF1-ZsGreen1-T2A-Puro) or knockdown (pLV-CMV-sh-hsa_circBNC2-EF1-ZsGreen1-T2A-Puro), as well as for ACSL6 overexpression (pLV-CMV-ACSL6-EF1-ZsGreen1-T2A-Puro) and knockdown (pLV-CMV-sh-ACSL6-EF1-ZsGreen1-T2A-Puro), were purchased from Fenghbio (Hunan, China). PCa cells were infected with lentiviruses in the presence of 8 µg/mL polybrene (Sigma-Aldrich). Infected cells were selected with 2 µg/mL puromycin (Merck) for 2 weeks, and successful establishment was confirmed by RT-qPCR and western blotting.

Reverse transcription quantitative PCR (RT-qPCR)

Total RNA was extracted from tissues or cells via TRIzol reagent (Invitrogen, USA). cDNA was subsequently synthesized via a ReverTra Ace reverse transcription (RT) regent kit (Toyobo, Japan) in accordance with the manufacturer’s instructions. RT-qPCR was conducted on a LightCycler 480 (Roche) with SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, USA). The primer sequences for circRNA, miRNA and mRNA were synthesized by Shanghai Sangon Co., Ltd. ACTB was employed as the internal reference for circRNA and mRNA, whereas U6 served as the internal reference for miRNA. The relative expression levels were calculated via the 2^{−ΔΔ Ct} method. The data are from three independent experiments performed in duplicate. The primer sequences are listed in Supplementary Table 1.

RNA sequencing

Total RNA was extracted from cells via TRIzol reagent (Invitrogen, USA). The concentration of RNA was measured via a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA). Subsequent analysis was conducted by Sangon Biotech (Shanghai, China). Messenger RNA (mRNA) was isolated from total RNA via mRNA separation techniques on the basis of the polyA structure at the 3‘end. Sequencing libraries were generated using Hieff NGS™ MaxUp Dual-mode mRNA Library Prep Kit for Illumina^® (12301ES96, YEASEN, China) following manufacturer’s recommendations and index codes were added to attribute sequences to each sample. In order to select cDNA fragments of preferentially 150 ~ 200 bp in length, the library fragments were purified with Hieff NGS™ DNA Selection Beads DNA (12601ES56, YEASEN, China). Then PCR was performed with 2×Super CanaceTM High-Fidelity Mix, Primer mix and Adapter Ligated DNA. At last, PCR products were purified (Hieff NGS™ DNA Selection Beads) and library quality was assessed on the Qubit^®2.0 Flurometer. The libraries were then quantified and pooled. Paired-end sequencing of the library was performed on the HiSeq NovaSeq 6000 sequencers (Illumina, San Diego, CA). High-throughput whole-transcriptome sequencing and bioinformatics analysis were subsequently performed. The protocol of RNA sequencing was presented in the Supplementary materials and methods.

RNase R and actinomycin D treatment

RNase R was employed to identify and affirm the nature of the circRNA. The RNAs extracted from PC3 or DU145 cells were partitioned into two equal portions, one for RNase R and the other for the negative control (Mock). Total RNA (2 µg) was incubated with 3 U/µg RNase for 30 min at 37 °C before reverse transcription. For actinomycin D (A1410, Sigma) treatment, PC3 and DU145 cells were cultured in 6-well plates and treated with 5 µg/mL actinomycin D when the cells reached approximately 60% confluence. After being treated with actinomycin D and RNase R, the RNA expression levels of circBNC2 were ascertained via RT-qPCR assay.

Cell counting kit-8 (CCK-8)

The cells were seeded in 96-well plates (2 × 10³ cells/well) and allowed to adhere. Ten microlitres of CCK-8 solution (CCK-8, Abcam Biotech Co., Ltd.) were added to each well, and the samples were cultured for 2 h. The optical density was recorded at 450 nm with a microplate reader. Proliferation rates were measured at 0, 24, 48, 72 and 96 h after treatment.

Transwell assay

A transwell 24-well chamber (Corning Inc., Corning, NY, USA) was used for the detection of invasion (with Matrigel). Briefly, a cell suspension (2 × 10⁴ cells) in serum-free medium was inoculated into the upper chamber, and medium containing 10% FBS was added to the lower chamber. After incubation for 24 h, the cells passed through the filter due to the attraction of the medium in the lower chamber. The passed cells were fixed and stained with methanol and crystal violet, and then the number of cells was counted under an inverted microscope (Olympus, Tokyo, Japan).

Wound healing assay

A marker was used to draw horizontal lines with the help of a ruler on the back of a 6-well plate. The lines were drawn evenly to a length of 1 cm, and there were at least 3 lines for each well. A total of 5 × 10⁵ cells were inoculated into each well of a 6-well plate. The cells were transfected when they reached a density of approximately 70% confluence. When the cells had just covered the entire well, a 10 µL pipette tip was used to make a scratch, with the help of a ruler, perpendicular to the horizontal lines on the back of the plate. The detached cells were washed with PBS, and then serum-free medium was added to continue the culture for 24 h. Samples were taken at 0 and 24 h, the cells were photographed, and the scratch width was measured.

Dual-luciferase reporter assay

The luciferase reporter vectors containing binding sequences between circBNC2 and miR-4298 and between ACSL6 and miR-4298 and their mutant sequences were constructed by RiboBio Co., Ltd, and the strategies were presented in the Supplementary materials and methods. A mixture of the luciferase reporter vectors and the miR-4298 mimic (RiboBio, Guangzhou, China) was cotransfected into DU145 and PC3 cells. After transfection for 48 h, the luciferase activity was detected via the Dual-Glo^® Luciferase Assay System (Catalog E2920, Promega) according to the manufacturer’s instructions and measured using a dual-luciferase reporter gene detection system (GloMax 96, Promega).

Western blotting assay

Western blotting was performed according to a previous study [36]. Protein lysates were prepared, subjected to SDS‒PAGE, transferred onto NC membranes and blotted according to standard methods via anti-CDK2 (1:2000 dilution; ab32147, Abcam), anti-CDK4 (1:2000 dilution; ab108357, Abcam), anti-Cyclin B1 (1:2000 dilution; ab32053, Abcam), anti-Cyclin D1 (1:2000 dilution; ab134175, Abcam), anti-ACSL6 (1:1000 dilution; ab229958, Abcam), anti-ACSL4 (1:1000 dilution; ab205199, Abcam), anti-COX2 (1:2000 dilution; ab283574, Abcam), anti-NOX1 (1:2000 dilution; ab131088, Abcam), anti-GPX4 (1:2000 dilution; ab125066, Abcam), and anti-FTH1 (1:2000 dilution; ab65080, Abcam), and anti-β-actin monoclonal antibodies (1:5000 dilution; ab8226, Abcam) served as a loading control. Afterward, membranes were incubated with secondary antibodies HRP-conjugated-goat anti-rabbit IgG (H + L) (1:10000 dilution; ab205718, Abcam) at room temperature. β-actin served as an internal control, and signaling was detected by ECL (Sigma-Aldrich, Inc.)

RNA fluorescence in situ hybridization (FISH)

The Cy3-labelled circBNC2 probes and fluorescein amidite (FAM)-labelled miR-4298 probes were designed and synthesized by RiboBio Co., Ltd. (Guangzhou, China). A fluorescence in situ hybridization (FISH) assay was employed to detect the probe signals in PC3 and DU145 cells via a fluorescence in situ hybridization kit (C10910, RiboBio). The cell nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI). Images were captured via a Zeiss confocal microscope.

RNA immunoprecipitation (RIP) assay

RIP experiments were conducted following the instructions of the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Sigma‒Aldrich, USA). DU145 and PC3 cells were treated with RIP lysis buffer containing proteinase (1:100) and RNase inhibitors (1:100). The cell lysate was subsequently incubated with IgG- or anti-AGO2 antibody-coated beads (Millipore) overnight at 4 °C. The following day, proteinase K (HY-108717, MCE) buffer was used for treatment. The RNA from the immunoprecipitate was subsequently extracted via the RNeasy MinElute Cleanup Kit (QIAGEN, Germany), reverse transcribed into cDNA, and finally analyzed by RT-qPCR assay.

Measurement of lipid reactive oxygen species (ROS), MDA, Fe2+ and GSH levels

Intracellular ROS production was measured via a reactive oxygen species assay kit (Meilun, Dalian, China). In brief, 10 µM DCFH-DA diluted in DMEM was added to the PRAD cells, which were incubated for 30 min at 37 °C. The samples were then examined via a ZEISS LSM 800 confocal microscope (Germany). Fluorescence images were captured at excitation/emission wavelengths of 488/525 nm. MDA levels were determined via the Lipid Peroxidation MDA Assay Kit (E-BC-K028-M, Elabscience), and the absorbance was measured at 532 nm. The concentration of Fe²⁺ was assessed via an Iron Content Assay Kit (E-BC-K881-M, Elabscience), and the absorbance was measured at 593 nm. The levels of GSH were determined via a glutathione assay kit (E-BC-K030-M, Elabscience), and the absorbance was measured at 405 nm.

Immunohistochemistry (IHC)

IHC was performed according to a previous study [36]. Briefly, tissue samples were fixed with 4% formalin and embedded in paraffin. Furthermore, thin sections of tissue embedded in paraffin were cut, dewaxed, and rehydrated. The sections were then blocked in 10% goat serum at room temperature for 10 min and incubated overnight at 4 °C with primary antibodies. Specific primary antibodies against ACSL6 (1:1000 dilution; ab229958, Abcam), GPX4 (1:1000 dilution; ab125066, Abcam) and Ki67 (1:1000 dilution; ab15580, Abcam) were used for IHC. Then, the sections were incubated with the secondary antibody (1:500 dilution; Goat Anti-Rabbit IgG H&L (HRP), ab6721, Abcam) as indicated and processed with streptavidin‒peroxidase before counterstaining with haematoxylin. The staining results were scanned with a molecular microscope (Olympus, Tokyo, Japan).

Preparation of OA-Gd/PDA nanoparticles for photothermal therapy of PCa cells

Synthesis of polyacrylic acid nanoparticles (PAA NPs) Polyacrylic acid nanoparticle (PAA NP) PAA NPs were synthesized at room temperature. Firstly, the PAA solution and ammonia water were added to 10 mL of deionized (DI) water and reacted under ultrasonic conditions for 15 min. Then, 40 mL of isopropyl alcohol (IPA) was gradually added dropwise to the mixed solution, during which time the solution changed from clear to milky white. The obtained PAA NPs were stored at room temperature for the next experiment.

Synthesis of rare earth hydroxide/poly(acrylic acid) NPs (RE(OH)₃/PAA NPs) In a 250 mL round-bottom flask (550 µL of 0.1 mol/L) C₆H₁₁GdO₇ was added to the above PAA NP mixture and stirred for 6 h. Then, the mixture was centrifuged at 8000 rpm for 8 min, washed with DI water 3 times, and the supernatant was discarded to obtain RE(OH)₃/PAA NPs.

Synthesis of RE(OH)₃/PAA@polydopamine (PDA) asymmetric NPs The RE(OH)₃/PAA NPs were dispersed into a 25 mL mixed solution (H₂O: IPA = 1:4). We subsequently adjusted the pH of the mixed solution to 8.8 using an ammonia solution (NH₃·H₂O) (2 mol/L). Next, dopamine (DA) was added, and the mixture was stirred for 16 h to obtain RE(OH)₃/PAA@PDA NPs.

Synthesis of oleic acid-RE/polydopamine NPs (OA-Gd/PDA NBs) The obtained RE(OH)₃/PAA@PDA NPs were mixed with 6 mL of solvent (H₂O: EtOH: OA = 2:3:1) in a 100 mL round-bottom flask and stirred for 15 min. Subsequently, sodium fluoride (NaF) was added to the dispersion, which was further agitated for an additional 15 min. Finally, the mixed solution was transferred to an autoclave and reacted at 180 ℃ for 24 h. Then, the NPs were washed alternately with DI water and ethanol (EtOH) three times to obtain OA-Gd/PDA NPs, which were stored under ambient conditions for subsequent applications or experiments.

Mouse xenografts and treatments

Six-week-old male BALB/c nude mice were selected for subcutaneous tumor experiments. The mice were purchased from Vital River Laboratory Animal Technology Co. Ltd. and housed in an SPF-grade animal facility at the Fourth Harbin Medical University Hospital. The light‒dark cycle was 12 h, and the pathogen-free environment had a relative humidity of 40–60% and a temperature of 25 ± 1 °C. The mice were fed according to the supplier’s instructions. Cultured DU145 cells were collected by centrifugation, washed twice with PBS, and injected subcutaneously into nude mice (1 × 10⁷ cells/mouse). Tumor volume was measured every four days with digital callipers the tumor volume was calculated via the following formula: length × width² × 0.5. After 40 days, the mice were euthanized, and the tumors were dissected and weighed. These tumors were subsequently subjected to haematoxylin and eosin (HE) (G1120, Solarbio, Beijing) and IHC staining. A mouse lung metastasis model was established using DU145-luciferase cells. The DU145-luciferase cell line was purchased from Pricella Biotechnology (Wuhan, China). A suspension of 1 × 10⁶ cells was injected into the tail vein of 6-week-old male BALB/c nude mice. After 40 days, imaging of the mice was performed via the Xenogen IVIS Lumina II imaging system (Calliper Life Sciences, Hopkinton, Massachusetts, USA). D-Luciferin sodium was purchased from MedChemExpress (MCE). The mice were then euthanized, the lung lobes were isolated, and HE staining was performed to analyse lung metastasis in the mice. All animal experimental procedures were approved by the Animal Care and Use Committee of the Fourth Harbin Medical University Hospital and strictly adhered to the guidelines for animal handling and welfare.

Statistical analysis

All experiments were independently repeated at least three times. The quantitative data are presented as the means ± SDs and were compared via Student’s t test or two-way ANOVA. The Kaplan‒Meier method was used to determine the overall survival (OS) rate, and the log-rank test was used to calculate the significance of survival rate differences. All the statistical analyses were performed via GraphPad Prism 9.0 (GraphPad Software, La Jolla, USA). P < 0.05 was considered statistically significant.

Results

CircBNC2 is downregulated in PCa tissues and cell lines

To gain further insight into the role of dysregulated circRNAs in PCa, we reanalyzed published datasets (MiOncoCirc [37, 38], GSE140927 [39], and GSE179321 [40]) from the Gene Expression Omnibus database to confirm that circRNAs are abnormally expressed between PCa and paratumor tissues. Among the abnormally expressed circRNAs (Fig. 1A-D), the downregulated hsa_circ_0008732 (also named circBNC2) attracted our attention. We chose to investigate circBNC2 for the following reasons: a, when overlapping the publicly available MiOncoCirc, GSE140927, and GSE179321 datasets, only circBNC2 was downregulated in these studies (Fig. 1E). Moreover, circBNC2 had relatively low expression levels in PCa tissues compared with normal human prostate tissues (Fig. 1A, C, D). Additionally, the expression of circBNC2 in NEPC is lower than that in CRPC (Fig. 1B). b, we also verified the expression of circBNC2 in PCa and adjacent normal tissue samples from 78 pairs of patient tissues via RT-qPCR. CircBNC2 expression in PCa tissues was lower than that in corresponding adjacent normal samples (Fig. 1F, G). c, circBNC2 had relatively low expression levels in PCa cells (LNCaP, VCaP, 22RV1, C4-2, PC3 and DU145) compared with the human epithelium prostate cell line RWPE-1 (Fig. 1H). d, Functional studies of circBNC2 are rare in PCa. Given the above results, we speculated that circBNC2 may play a tumor-suppressive role in prostate cancer.

Fig. 1

Identification and characterization of circBNC2 in PCa. A Heatmaps show 20 dysregulated circRNAs in PCa tumor and peritumor tissues from MiOncoCirc and Reference 37. The red and blue strips represented high and low expression, respectively. B Heatmaps show 20 dysregulated circRNAs in eight NEPC cases versus 35 CRPC cases with the highest tumor purity from Reference 37. The red and blue strips represented high and low expression, respectively. C and D Heatmaps show 20 dysregulated circRNAs in GSE140927 (C), GSE179321 (D). The red and blue strips represented high and low expression, respectively. E Venn diagram shows the crossed circRNAs among A, B, C and D. F Analysis for mRNA levels of circBNC2 in 78 paired samples of PCa were determined by RT-qPCR. G Histogram of PCa samples in which circBNC2 expression was upregulated (11/78, 14.10%), downregulated (55/78, 70.51%), or no change (12/78, 15.38%), respectively. Log2 (T/N expression) value > 1 as significantly higher expression, <-1 as lower expression, and between − 1 and 1 as no significant change. T, tumor tissue; N, normal tissue. H Analysis for mRNA levels of circBNC2 in PCa cell lines (DU145, PC3, C4-2, 22RV1, VCaP, LNCaP) and normal prostate cell line (RWPE-1). I The back-splice junction site of circBNC2 was identified by Sanger sequencing. J PCR analysis for circBNC2 and BNC2 in complementary DNA (cDNA) and genomic DNA (gDNA). K Analysis for mRNA levels of circBNC2 and the BNC2 gene after treated with RNase R in PC3 and DU145 cells. L RT-qPCR for the abundance of circBNC2 and BNC2 mRNA in PC3 and DU145 cells treated with Actinomycin D at the indicated time points. M Nuclear-cytoplasmic fractionation assay results indicated that circBNC2 was primarily localized in the cytoplasm of PC3 and DU145 cells. The 18s rRNA and the U6 genes were used as cytoplasmic and nuclear controls, respectively. N FISH experiments showed that circBNC2 predominantly located in cytoplasm in DU145 and PC3 cells, the circBNC2 probe was labeled with Cy3(red), while nuclei were stained with DAPI (blue). Scale bar = 25 μm. ns, no significance; *P < 0.05; **P < 0.01; ***P < 0.001

On the basis of the circBase annotation (http://www.circbase.org/), circBNC2 originated from exons 2 and 3 of the BNC2 gene (Fig. 1I). CircRNAs and their linear counterparts have identical sequences, except at the junction of the transcript. Therefore, we designed primers (divergent primers) that target the back-splice junction and primers (convergent primers) that target the linear section. The divergent primers were found to amplify only cDNA, confirming the circular structure of the RNA transcript (Fig. 1J). Sanger sequencing of the PCR products also revealed the back-splice junction site (Fig. 1I). The RNase R assay (Fig. 1K) and an actinomycin D assay (Fig. 1L) showed that circBNC2 was more stable than linear BNC2. In addition, the results of the nuclear and cytoplasm fractionation assays (Fig. 1M) and FISH assays (Fig. 1N) revealed that circBNC2 was located mainly in the cytoplasm of these two PCa cell lines. Overall, these results indicate that circBNC2 could act as a ‘protein sponge’ or ‘miRNA sponge’.

The samples were split into two groups to investigate the connection between circBNC2 expression and the clinical significance of PCa. The low circBNC2 expression group was associated with tumor stage (P = 0.0013) and Gleason grade (P < 0.001) but not with age or PSA level (P > 0.05; Table 1). Taken together, these findings indicated that circBNC2 is a stable circRNA located in the cytoplasm that has the potential to be a biomarker for the diagnosis and prognosis of PCa.

Table 1 Clinicopathological parameters of PCa patients and correlation with circBNC2 expression (n = 78)

CircBNC2 inhibits the proliferation, migration and invasion of PCa cells in vitro and in vivo

We investigated the biological role of circBNC2 in PCa via loss- and gain-of-function assays. We used a lentiviral system to overexpress circBNC2 in DU145, PC3, VCaP and LNCaP cells. RT-qPCR analysis revealed that the lentiviral vector system successfully overexpressed circBNC2 (OE-circBNC2) (Fig. 2A; Figure S1A), with no impact on the expression of the BNC2 gene (Fig. 2B; Figure S1B). DU145, PC3, VCaP and LNCaP cells with OE-circBNC2 had lower proliferation rates (Fig. 2C; Figure S1C), a slower migration phenotype (Fig. 2D; Figure S1D) and more inhibited invasive ability (Fig. 2E; Figure S1E). We also used a lentiviral vector system to knockdown the junction site of circBNC2 (Figures S1F-I). Knocking down circBNC2 increased the proliferation, migration and invasion of DU145, PC3, VCaP and LNCaP cells (Figures S1J-O). All of these results indicated that circBNC2 exhibits a potential tumor suppressive role in DU145, PC3, VCaP and LNCaP cells. Specifically, as shown in Fig. 1H, circBNC2 expression was lower in DU145 and PC3 than in VCaP, LNCaP, 22RV1 and C4-2 cells. Therefore, we selected DU145 and PC3 cells, which presented the lowest circBNC2 expression, for subsequent studies. Additionally, we found that OE-circBNC2 inhibited the cell cycle progression of DU145 and PC3 cells, causing cell cycle arrest at the G0/G1 phase (Fig. 2F). Furthermore, western blotting revealed that CDK2, CDK4, Cyclin B1, and Cyclin D1 were also downregulated in OE-circBNC2 PCa cells (Fig. 2G). We then further investigated the role of circBNC2 in vivo. We subcutaneously implanted OE-circBNC2 DU145 cells into nude mice and monitored the growth of xenograft tumors. The Fig. 2H illustrates the experimental strategy. Our results showed that OE-circBNC2 notably inhibited the growth rate (Fig. 2I, J) and the weights of the xenograft tumors (Fig. 2K) compared with the OE-Ctrl group. Moreover, as shown in Fig. 2L, the expression levels of Ki67 were decreased in circBNC2-overexpressing xenograft tumors, which further confirmed that circBNC2 inhibited tumor growth in vivo. Overall, these results indicated that circBNC2 is an important tumor suppressor in PCa.

Fig. 2

CircBNC2 is a potential tumor suppressor in PCa cells. A and B RT-qPCR results showing the expression of circBNC2 (A) and BNC2 mRNA (B) in OE-circBNC2 DU145 and PC3 cells. C-F In OE-circBNC2 DU145 and PC3 cells, cell proliferation was assessed using the CCK-8 assay (C), cell migration was evaluated by the wound healing assay (D) (Scale bar = 200 μm), cell invasion was measured by the transwell assay (E) (Scale bar = 50 μm), and cell cycle changes were analyzed by flow cytometry (F). G Western blotting analyzed the expression of cell cycle-related protein in OE-circBNC2 DU145 and PC3 cells. H Schematic illustration of subcutaneous tumor model establishment and treatment. I Representative images of subcutaneous xenograft tumors (n = 6 for each group). J and K The xenografted tumor growth curve (J) and tumor weight (K) were measured in the OE-circBNC2 group and control group (OE-Ctrl) as indicated. L The representative images of HE staining (Scale bar = 100 μm) and Ki67 immunostaining (Scale bar = 25 μm) analysis of subcutaneous xenograft tumors in OE-circBNC2 or OE-Ctrl group. ns, no significance; *P < 0.05; **P < 0.01; ***P < 0.001

CircBNC2 sponges miR-4298 in PCa cells

Considering the subcellular location of circBNC2 in PCa cells, we hypothesized that it may function by sponging miRNA [41]. We used circBank, miRDB and TargetScan as bioinformatics tools to identify potential miRNAs that might interact with circBNC2. As shown in Fig. 3A, there are several potential miRNAs with target sites for circBNC2. We further sought to identify the target of circBNC2, and analyzed the alterations in these microRNAs in OE-circBNC2 DU145 or OE-circBNC2 PC3 cells. We subsequently found that only miR-4298 expression was downregulated in the OE-circBNC2 PCa cells (Figure S2). And we found miR-4298 is overexpression in PCa samples compared to the normal samples on the dataset (GSE113234) (Figure S3A). Therefore, we selected miR-4298 for further study. Additionally, we found that circBNC2 was negatively correlated with miR-4298 in PCa patients (Fig. 3B). Next, we determined the subcellular distribution of circBNC2 and miR-4298 in PCa cells. FISH was performed and revealed that both circBNC2 and miR-4298 were located in the cytoplasm (Fig. 3C). Furthermore, a dual-luciferase reporter assay confirmed that miR-4298 is the sponging target of circBNC2. We constructed plasmids encoding wild-type (WT-circBNC2) and mutated (Mut-circBNC2) forms corresponding to miR-4298 according to the bioinformatics-predicated binding sequence between circBNC2 and miR-4298. These two plasmids were then transfected into control group (NC mimic) or miR-4298-overexpressing (miR-4298 mimic) DU145 and PC3 cells, respectively. As shown in Fig. 3D, WT-circBNC2 was found to suppress luciferase reporter activity in the miR-4298 mimic group compared with the control group (NC mimic), while there was no significant difference in luciferase activity among the miR-4298 mimic or NC mimic groups in the Mut-circBNC2-transfected group. Furthermore, we investigated the regulatory effects of miR-4298 and circBNC2 via a RIP assay. We performed a RIP assay with an antibody against Argonaute 2 (AGO2) in DU145 and PC3 cells. AGO2 has been demonstrated to be a key protein that participates in miRNA function in an RNA-induced silencing complex (RISC)-dependent manner [42]. CircBNC2 and miR-4298 pulled down by anti-AGO2 antibodies were both significantly enriched, indicating that both circBNC2 and miR-4298 are present in RISCs (Fig. 3E). These results suggested that circBNC2 functions as a miRNA sponge through direct targeting of miR-4298.

Fig. 3

CircBNC2 functions as a sponge for miR-4298. A The Venn diagram showing the predicted overlapping potential target miRNAs of circBNC2 by TargetScan, circbank and miRDB. B Correlation analysis showing the mRNA expression between circBNC2 and miR-4298 in 78 PCa patient samples. C The co-localization of circBNC2 and miR-4298 in DU145 and PC3 cells was analyzed by FISH assay. The circBNC2 probe was labeled with Cy3 (red), miR-4298 probes were labeled with FAM (green), and nuclei were stained with DAPI (blue). Scale bar = 10 μm. D The schematic illustration for construction of luciferase reporter plasmids, wild-type (WT) and mutant (Mut)-circBNC2, as indicated (left panel). The luciferase activity was measured in DU145 and PC3 cells which co-transfected with of the luciferase reporter plasmid WT/Mut-circBNC2 and miR-4298 mimic/NC mimic, respectively (right panel). E Fold enrichment of circBNC2 and miR-4298 in PC3 and DU145 cells by the AGO2-RIP assay. F and G RT-qPCR analysis for the expression of miR-4298 in OE-circBNC2 (F) or sh-circBNC2 (G) DU145 and PC3 cells. H - J In OE-Ctrl + NC mimic-, OE-Ctrl + miR-4298 mimic-, OE-circBNC2 + NC mimic- or OE-circBNC2 + miR-4298 mimic- treated DU145 and PC3 cells, cell proliferation, invasion and migration were analyzed by the CCK-8 assay (H), transwell assay (I) (Scale bar = 50 μm) and wound healing assay (J) (Scale bar = 200 μm), respectively. K - M The representative images of subcutaneous xenograft tumors (n = 6/group) (K), tumor growth curves (L), and tumor weight (M) at the endpoint from control and circBNC2 overexpression groups co-treated with NC agomir or miR-4298 agomir. ns, no significance; *P < 0.05; **P < 0.01; ***P < 0.001

We next sought to investigate the role of miR-4298 in PCa cells. We found that miR-4298 was upregulated in PCa cells compared with RWPE-1 cells (Figure S3B). Moreover, the overexpression of miR-4298 (by miR-4298 mimic) (Figure S3C) significantly promoted the proliferation, migration, and invasion of DU145 or PC3 cells compared with the negative control mimic (NC mimic) (Figures S3D-F). In contrast, the proliferation, migration, and invasion of PCa cells were inhibited when PCa cells were treated with a miR-4298 inhibitor (Figures S4A-D). To further evaluate the oncogenic function of miR-4298 in vivo, a subcutaneous xenograft nude mouse model was constructed using DU145 cells in which miR-4298 was silenced with a miR-4298 antagomir, and the NC antagomir was used as the negative control (Ruibo Bio, Guangzhou, China). Subsequently, the silencing of miR-4298 caused a significant decrease in xenograft tumor growth (Figures S4E-H). Additionally, the expression of miR-4298 was significantly upregulated in PCa tissues compared with matched normal adjacent tissues (Figure S4I).

Moreover, we also found that OE-circBNC2 significantly reduced the level of miR-4298 (Fig. 3F), whereas the silencing of circBNC2 (sh-circBNC2) promoted miR-4298 expression in DU145 and PC3 cells (Fig. 3G). We then verified the potential antagonistic effect between circBNC2 and miR-4298 in PCa cells. As illustrated in Fig. 3H‒J, OE-circBNC2 significantly reduced the proliferation, invasion and migration in miR-4298 mimic treated-DU145 or PC3 cells. These findings indicate that OE-circBNC2 attenuated the carcinogenic effect of miR-4298. Moreover, we confirmed the inhibitory effect of circBNC2 on miR-4298 in vivo. We found that OE-circBCN2 reversed the growth of xenograft tumors caused by the overexpression of miR-4298 (miR-4298 agomir) (Fig. 3K-M). In summary, our results suggested that circBNC2 inhibits the aggressiveness of PCa cells by acting as a molecular sponge of miR-4298.

CircBNC2 sponging of miR-4298 results in the upregulation of ACSL6 expression in PCa cells

We then investigated how miR-4298 affects downstream factors. Through cross-analysis of these mirDIP, TargetScan, miRPathDB and miRWalk online databases, we obtained 387 candidate downstream target genes of miR-4298 (Fig. 4A). Then, based on the TCGA database, we analyzed the differential expression of 387 genes in PCa (Fig. 4B). The miRNAs can function by degrading mRNAs or obstructing the translation of targeted genes [43]. Since the miR-4298 was up-regulated in PCa cells, we should selected the down-regulated genes among these predicted 387 genes as the downstream target of miR-4298. Among 387 genes (up-regulation of 12, down-regulation of 41), we found the ACSL6 is the most down-regulated gene among all the candidate genes. Therefore, we selected ACSL6 as the priority research candidate.

Fig. 4

CircBNC2 regulates ACSL6 expression by sponging in PCa cells. A Venn diagram showing 387 potential target genes which predicted by four databases (TargetScan, miRPathDB, miRWalk, and mirDIP). B Volcano plot showing the profile of differentially expressed genes based on TCGA of PCa. C Correlation analysis revealed negative correlation between the levels of ACSL6 and miR-4298 in 78 PCa patient samples. D The schematic illustration for construction of luciferase reporter plasmids, wild-type (WT) and mutant (Mut)-ACSL6, as indicated (left panel). The luciferase activity was measured in DU145 and PC3 cells which co-transfected with of the luciferase reporter plasmid WT/Mut-ACSL6 and miR-4298 mimic/NC mimic, respectively (right panel). E - H The expression level of ACSL6 was analyzed in DU145 and PC3 cells by RT-qPCR and western blotting after transfection with NC mimic/miR-4298 mimic (E and F) or NC inhibitor/miR-4298 inhibitor (G and H), respectively. I - J The CCK-8 assay (I) and transwell assay (J) showing that miR-4298 mimic or NC mimic affected the proliferative and invasive ability of OE-ACSL6 DU145 and PC3 cells. K and L The expression level of ACSL6 was validated by RT-qPCR (K) and western blotting (L) in OE-circBCN2 DU145 and PC3 cells. M In OE-circBNC2 PC3 and DU145 cells, western blotting analyzed the ACSL6 expression upon miR-4298 mimic or NC mimic treatment. N and O CCK-8 assay (N) and transwell assay (O) showing that sh-ACSL6 knockdown reverse the inhibition of proliferation and invasion in PC3 and DU145 cells caused by circBNC2 overexpression. P - R The representative images of subcutaneous xenograft tumors (n = 6/group) (P), tumor growth curves (Q), and tumor weight (R) at the endpoint from control (OE-Ctrl) and circBNC2 overexpression (OE-circBNC2) groups co-treated with knocking down of control (sh-Ctrl) or knocking down of ACSL6 (sh-ACSL6). S Correlation analysis revealed positive correlation between the levels of ACSL6 and circBNC2 in the 78 PCa patient samples. Scale bar = 50 μm. ns, no significance; *P < 0.05; **P < 0.01; ***P < 0.001

ACSL6 belongs to the acyl-CoA synthetase long-chain (ACSL) family. ACSLs include five isoforms identified as ACSL1, 3, 4, 5, and 6 [44]. They convert free long-chain fatty acids into fatty acyl-CoA esters and play a dominant role in both anabolic (fatty acid synthesis and lipogenesis) and catabolic (lipolysis and fatty acid β-oxidation) pathways, although they have distinct substrate specificities, subcellular localizations, and tissue distributions [45]. Several studies have been demonstrated that ACSL1, 3, 4 affect the behaviour of PCa cells, including proliferation, migration, invasion, apoptosis, and drug resistance [46,47,48]. However, the versatile biological function of ACSL6 remain elusive, especially in PCa. Intriguingly, as shown in Figure S5A, the expression of ACSL6 was found to be downregulated in PCa tissue specimens in the TCGA database. We also detected ACSL6 expression in PCa clinical samples (n = 78) and subsequently found that ACSL6 mRNA was downregulated in PCa tissues compared with normal tissues (Figure S5B, C). In addition, the expression of miR-4298 was inversely correlated with the ACSL6 level in PCa tissues (Fig. 4C). Furthermore, as shown in Figure S5D, bioinformatics illustrated that miR-4298 might share the complementary binding sites with the 3’UTR of ACSL6 mRNA. More importantly, the dual-luciferase reporter assay was subsequently carried out to confirm the regulatory relationship between miR-4298 and ACSL6. The results revealed that luciferase activity was apparently inhibited in the miR-4298 mimic- and ACSL6-WT-cotransfected DU145 and PC3 cells compared with that in the NC mimic- and ACSL6-WT-cotransfected groups, whereas no change was observed in the ACSL6-Mut groups (Fig. 4D). RT-qPCR and western blotting revealed that miR-4298 mimic transfection decreased ACSL6 expression (Fig. 4E, F), while the miR-4298 inhibitor upregulated ACSL6 expression (Fig. 4G, H). However, the function of ACSL6 in PCa is still not clear. We intriguingly found that ACSL6 level was also downregulated in PCa cells compared with RWPE-1 cells (Figure S5E). Meanwhile, we also found the overexpression of ACSL6 (OE-ACSL6) significantly inhibits the cell proliferation, cell migration and invasion of DU145 and PC3 cells (Figures S5F-I).

More importantly, ACSL3 and ACSL4 have been shown to participate in ferroptosis. In addition, ACSL4 is a positive regulator in ferroptosis, whereas ACSL3 contributes to cancer cells acquiring ferroptosis resistance [49]. However, it remains unclear whether ACSL6 is related to ferroptosis in prostate cancer. Thereafter, we also investigated whether ACSL6 regulates ferroptosis in PCa cells. As shown in Figure S5J, OE-ACSL6 induced the MDA, Fe²⁺ levels, while reducing the GSH level. However, we also found the intracellular MDA, Fe²⁺ levels were decreased, whereas the GSH level was increased in knocking down of ACSL6 (sh-ACSL6) PCa cells (Figure S5K). In addition, in DU145 and PC3 cells, OE-ACSL6 reversed the carcinogenic effects caused by transfection with the miR-4298 mimic (Fig. 4I, J). Combined, our results suggested that ACSL6 acts as a tumor suppressor via targeting miR-4298 and may be a driver of ferroptosis in PCa cells.

Furthermore, we also found that OE-circBNC2 meaningfully induced the mRNA and protein expression of ACSL6 (Fig. 4K, L). Consistent with these results, the miR-4298 mimic attenuated the effect of OE-circBNC2 on ACSL6 expression (Fig. 4M). Moreover, to determine whether the tumor-suppressing functions of circBNC2 are dependent on ACSL6, we knocked down ACSL6 to perform several rescue experiments in OE-circBNC2 PCa cells. The inhibitory effect of circBNC2 overexpression on the proliferation, migration and invasion abilities of DU145 or PC3 cells was subsequently reversed in the presence of sh-ACSL6 (Fig. 4N, O; Figure S6). Meanwhile, we demonstrated that OE-circBNC2 inhibited the growth of DU145-xenograft tumors in vivo, whereas silencing ACSL6 attenuated this inhibition (Fig. 4P-R). Furthermore, we found that circBNC2 was positively related to ACSL6 expression in human PCa tissues (n = 78) (Fig. 4S). In summary, these results demonstrated that circBNC2 inhibits proliferation, migration and invasion through the miR-4298/ACSL6 axis.

ACSL6 is essential for mediating circBNC2-regulated ferroptosis in PCa cells

To further elucidate the mechanism underlying circBNC2-mediated tumor inhibition, we conducted RNA sequencing in OE-cirBNC2 DU145 cells. Our KEGG enrichment analysis revealed that the ferroptosis pathway was the top hit pathway for upregulated genes in OE-circBNC2 DU145 cells (Fig. 5A). A total of 5163 DEGs (OE-circBNC2 vs. OE-Ctrl, 2575 downregulated and 2588 upregulated) were identified for further analysis (see Supplementary Data). The gene‒rank plot revealed a significant difference in gene expression vs. fold change (Fig. 5B). Among the DEGs, six genes were associated with ferroptosis, including the upregulated genes ACSL6, ACSL4, PTGS2 and NOX1 and the downregulated genes GPX4 and FTH1 (Fig. 5B). Therefore, we further verified the regulatory function of circBNC2 in ferroptosis in PCa cells. As illustrated in Fig. 5C-F and Figure S7A, OE-circBNC2 increased the intracellular malondialdehyde (MDA), Fe²⁺ and ROS levels while decreasing the glutathione (GSH) content. Conversely, in sh-circBNC2 PCa cells, the levels of intracellular MDA and Fe²⁺ decreased, and the GSH level increased compared with those in sh-Ctrl-treated cells (Figures S7B-D). Furthermore, we also demonstrated that ferroptosis pathway-related genes, including ACSL6, ACSL4, PTGS2 and NOX1, were upregulated in OE-circBNC2 DU145 and PC3 cells by western blotting and RT-qPCR analysis, respectively (Fig. 5G and H). These results indicate that OE-circBNC2 promotes ferroptosis in PCa cells. Moreover, sh-ACSL6 significantly impaired the intracellular MDA, Fe²⁺ and ROS levels induced by OE-circBNC2 in DU145 and PC3 cells (Fig. 5I-K; Figure S8A). In addition, sh-ACSL6 treatment also increased the GSH level in OE-circBNC2-treated PCa cells (Fig. 5L). In contrast, OE-ACSL6 promoted ferroptosis even in sh-circBNC2 DU145 or PC3 cells (Figures S8B-D). Moreover, a rescue experiment revealed that silencing of ACSL6 reversed the increased expression of ACSL4, COX2 and NOX1 and the decreased expression of GPX4 and FTH1 caused by OE-circBNC2 treatment in PCa cells (Fig. 5M). These findings indicated that ACSL6 is the pivotal factor through which circBNC2 mediates ferroptosis in PCa cells. Collectively, these findings demonstrated that circBNC2 is a tumor-suppressive circRNA that inhibits cancer cell proliferation, migration, and invasion and promotes ferroptosis in PCa cells via the circBNC2/miR-4298/ACSL6 signaling pathway (Fig. 5N).

Fig. 5

CircBNC2 drives ferroptosis through ACSL6 in PCa cells. A KEGG pathway enrichment analysis based on the RNA-seq. Pathway enrichment of differentially expressed genes. The bubble size indicates the number of genes. The color bar indicates the corrected P-value. The threshold for differential was set at 1-fold change, and P < 0.05, as determined by DESeq2. B Volcano plot showing the profile of DEGs based on overexpression of circBNC2. C - F In OE-circBCN2 DU145 and PC3 cells, the concentration of MDA (C), Fe²⁺ (D), ROS (E), and GSH (F) were measured as indicated. G Western blotting showing the expression of ACSL6, ACSL4, COX2, NOX1, GPX4, FTH1 in OE-circBCN2 DU145 and PC3 cells. H RT-qPCR results showing the expression of circBNC2, ACSL6, ACSL4, GPX4, NOX1, FTH1, PTGS2 in OE-circBCN2 PC3 and DU145 cells. I - L Upon sh-Ctrl or sh-ACSL6 treatment, the levels of MDA (I), Fe²⁺ (J), ROS (K) and GSH (L) were measured in OE-Ctrl or OE-circBNC2 PC3 and DU145 cells. M Western blotting results showing the expression of ACSL6, ACSL4, COX2, NOX1, GPX4, FTH1 in OE-circBNC2 PC3 and DU145 cells co-treated with sh-Ctrl or sh-ACSL6. N Schematic diagram illustrating the mechanism by which circBNC2 inhibit PCa proliferation, migration and invasion through the circBNC2/miR-4298/ACSL6 axis. Scale bar = 50 μm. ns, no significance; *P < 0.05; **P < 0.01; ***P < 0.001

Stepwise fabrication of bowl-like structured OA-Gd/PDA NPs

Considering the critical inhibitory role of circBNC2 in PCa, we sought to develop an efficient delivery vehicle for overexpressing circBNC2 as a potential therapy for PCa. Because polyacrylic acid (PAA) acts as a large reservoir capable of absorbing and retaining water molecules within its network structure, we successfully synthesized PAA NPs under room temperature conditions by gradually adding isopropanol (IPA) dropwise to an aqueous PAA solution, which resulted in altered intermolecular forces. Subsequently, rare earth elements were added, which underwent hydrolysis to form RE(OH)₃, ultimately yielding well-dispersed RE(OH)₃/PAA NPs. The reaction system was subsequently adjusted to a pH of 8.8 using NH₃·H₂O, immediately followed by the addition of N, N-dimethylaminoethyl acrylate (DA). Through precise control of the reaction time and temperature conditions, polydopamine (PDA) selectively grew on one side of the RE(OH)₃/PAA NPs, forming a bowl-like structure. This unique growth pattern is likely attributed to the templating effect of RE(OH)₃/PAA NPs, which facilitated island nucleation and anisotropic growth of PDA (Fig. 6A). Finally, we formed bowl-like OA-Gd/PDA NPs with hydrophobic cavities through a solvothermal reaction. Transmission electron microscopy (TEM) revealed that the particle size of these NPs was 180 ± 10 nm (Fig. 6B). We further applied elemental mapping to analyse the elemental distribution of the OA-Gd/PDA NPs. As shown in Fig. 6B, all the elements, including C, N, O, Gd, Na, and F, were located at the corresponding positions of the OA-Gd/PDA NPs, which further proves the successful preparation of NPs with ideal structures. Notably, scanning electron microscopy (SEM) revealed that the NPs exhibited a distinct bowl-shaped structure (Fig. 6C). Hereafter, we refer to the nanobowls as NBs.

Fig. 6

Characteristics and antitumor effects of Dc-NBs in PCa cells in vitro. A Schematic illustration of Step-wise fabrication of bowl-like structured OA-Gd/PDA NPs. B TEM images of the bowl-like structured OA-Gd/PDA NPs (named as NBs in this study) (upper left panel); Elemental mapping of NBs (lower left and right panel). C SEM image of the NBs. D The UV-vis absorbance spectra of NBs. E - F Photothermal temperature changes of NBs with different concentrations (E) and irradiation power densities (F). G Thermal images of water and NBs solution with different concentrations under NIR laser irradiation (808 nm). H Temperature change profiles of NBs with 4 times NIR irradiation cycle. I Heating and cooling curves (gray) and time constant (τs) for heat transfer from the system (pink). J Loading efficiency and encapsulation efficiency and of DTX in NBs nano-carrier. K Zeta potential of the D-NBs and circBNC2 loaded D-NBs. L CCK-8 assay analyzed cell viability of RWPE-1 and DU145 treated with saline (Ctrl) or different concentrations of NBs. M Analysis of cell viability of DU145 cells by CCK-8 assay, upon treatment with saline (Ctrl), DTX loaded NBs (D-NBs), DTX and circBNC2 co-loaded NBs (Dc-NBs) or Dc-NBs combined photothermal therapy (PTT) (Dc-NBs + PTT). N Western blotting results revealed the expression of ACSL6, ACSL4, COX2, NOX1, GPX4, FTH1 in DU145 cells treated with different treatment, including Ctrl, D-NBs, Dc-NBs or Dc-NBs + PTT. O The level of ROS was detected in DU145 cells by CLSM treated with different treatment as indicated. Scale bar = 50 μm. P  In vivo fluorescence images of DU145-tumor-bearing mice at various time points after tail vein injection of Cy5.5 loaded NBs and ex vivo fluorescence imaging of the tumor and the major organs at 48 h. ns, no significance; *P < 0.05; **P < 0.01; ***P < 0.001

Characterization of the photothermal performance of the as-fabricated NBs

The utilization of NBs with photothermal activity primarily aims to integrate photothermal therapy (PTT) for the treatment of prostate cancer. Due to the properties of PDA nanomaterial and their unique concave nanostructure, these NBs offer opportunities for incident light capture and energy conversion. The as-fabricated NBs exhibited broad absorbances ranging from UV to near-infrared (NIR) wavelengths (Fig. 6D), suggesting potential photothermal ability that may be utilized for clinical applications. To explore the in vitro photothermal efficacy of NBs, various power densities and concentration gradients were investigated to elucidate the concentration- and power-dependent attributes of the NBs, thereby highlighting their exceptional photothermal capabilities (Fig. 6E, F). Figure 6G provides an illustration of this phenomenon, indicating that the temperature increased with increasing concentration and prolonged irradiation time. As anticipated, after 4 laser switch irradiation cycles, there was no substantial alteration in temperature, indicating that the NBs demonstrated robust photothermal stability (Fig. 6H). The photothermal conversion efficiency (η) of the NBs was calculated to be 50.2% (Fig. 6I), rendering them conducive to intracellular PTT effects. The NBs can generate significant heat upon near-infrared (NIR) light irradiation, thereby inducing apoptosis in tumor cells.

Preparation of DTX/circBNC2-coloaded OA-Gd/PDA NBs (Dc-NBs)

Docetaxel (DTX) is a taxane that binds tubulin and stabilizes microtubules and, to our knowledge, is the first chemical agent to improve overall survival (OS) in men with mCRPC [50]. The loading of the chemotherapy drug DTX into the NBs primarily aims to leverage the characteristics of the nano-carrier for targeted delivery and controlled release of the drug. The NBs possess a large specific surface area and a unique cavity nanostructure, enabling efficient loading of drug molecules. Additionally, the NBs protect the drug from degradation in the biological environment, enhancing its stability and bioavailability. We tested the DTX loading capacity of the NBs (D-NBs). Owing to the unique morphology and chemical composition of NBs, the hydrophobic drug DTX can be loaded into the hydrophobic cavity of NBs at a drug loading rate of 6%, and the encapsulation efficiency of NBs is 85.7% (Fig. 6J). Next, we sought to load circBNC2 on the surface of NBs. Previous studies reported that polyethylenimine (PEI) can be utilized to mediate the electrostatic adsorption process, enabling the assembly of RNA into nanoparticles [51, 52]. The successful loading of RNA through electrostatic interactions was determined by measuring the zeta potential [53, 54]. As shown in Fig. 6K, the zeta potential of the D-NBs alone is 22.17 mV, which shifts to -18.80 mV after being decorated with circBNC2. These results indicated that DTX and circBNC2 were successfully loaded on NBs. We refer to the DTX/circBNC2-co-loaded NBs as Dc-NBs in this study. By co-loading circBNC2 and the chemotherapy drug DTX into NBs, and passively targeting them to the tumor site, the local heating effect of PTT can be utilized to promote drug release and uptake. This approach enhances therapeutic efficacy while reducing side effects. The tumor-inhibiting action of circBNC2 synergizes with the localized destructive effect of PTT.

Anti-tumor effects of Dc-NBs in vitro

Biocompatibility has always been a prerequisite for the application of nanomedicine in the biomedical field. Therefore, the in vitro biosafety of the NBs in RWPE-1 and DU145 cells was evaluated via CCK-8 assay. As shown in Fig. 6L, there was no significant change in the proliferation capacity of RWPE-1 and DU145 cells treated with different concentrations of NBs, indicating low cytotoxicity. Next, to evaluate whether different synthetic materials and photothermal therapies affect the proliferation of prostate cancer cells, we conducted a CCK-8 assay. Interestingly, we found that Dc-NBs significantly inhibited the proliferation of DU145 cells, particularly under photothermal treatment (Fig. 6M). Moreover, we also observed changes in the levels of proteins associated with ferroptosis, with increased expression of ACSL6, ACSL4, COX2, NOX1 (Fig. 6N), promoted ROS levels (Fig. 6O) and decreased expression of GPX4 and FTH1 (Fig. 6N), upon treatment with Dc-NBs or Dc-NBs combined with PTT. We subsequently verified the in vivo accumulation and penetration of the material in animal models. The accumulation of NBs at the tumor site is primarily due to their size and surface properties, which make them more susceptible to capture by tumor tissue vasculature and enable enrichment at the tumor site through the EPR effect (Enhanced permeability and retention effect, or the high permeability and retention effect of tumor tissues) [55, 56]. Dc-NBs were injected via the tail vein into DU145 tumor-bearing mice, which were then subjected to PTT treatment, and real-time fluorescence imaging was conducted over 48 h. As shown in Fig. 6P, the fluorescence intensity peaked at approximately 24 h and then gradually decreased. Additionally, ex vivo fluorescence imaging of major organs and tumors revealed that fluorescence signals were still detectable in the tumor and kidneys, suggesting that the nanomaterials are metabolized primarily through the kidneys. This does not compromise their therapeutic effect at the tumor site.

Anti-tumor effects of Dc-NBs in a subcutaneous and metastatic tumor model

To investigate the anti-tumor effects of Dc-NBs in vivo, we firstly investigated the in vivo anti-tumor efficacy of Dc-NBs in BALB/c nude mice bearing subcutaneous DU145 tumor-bearing models. The general experimental procedure is illustrated in Figure S9A. Briefly, DU145 tumor-bearing mice were randomly divided into five groups: peritumoral injection of saline (Ctrl), NBs, D-NBs, Dc-NBs, and Dc-NBs + PTT (for PTT treatment). At the end of the intervention, we observed that Dc-NBs + PTT group resulted in a more significant decrease in tumor sizes and weight than those from the Ctrl, NBs, D-NBs, Dc-NBs (Fig. 7A). Furthermore, the tumor size and tumor weight in the Dc-NB + PTT group were the smallest among the five groups, and the inhibitory efficiency reached approximately 90%, as calculated by the tumor weight (Fig. 7B, C). As shown in Figure S9B, treatment with Dc-NBs and Dc-NBs + PTT significantly increased the circBNC2 expression, whereas decreased the miR-4298 expression in subcutaneous tumor (Figure S9C), which compared with treated with Ctrl or NBs group. Moreover, the results of the IHC analysis revealed that treatment with Dc-NBs + PTT significantly increased the expression of ACSL6 and decreased the expression of GPX4 in xenografted tumors (Fig. 7D, S9D and S9E). These results indicated that treatment with Dc-NBs + PTT activated ferroptosis in xenografted tumor cells. In addition, we observed that the body weights of the mice in all the groups were not significantly affected by any treatment measures, indicating no significant toxicity (Figure S9F). Moreover, histological sections of vital organs (heart, liver, lung and kidney) stained with haematoxylin and eosin did not show any apparent injury to the cellular structures after intravenous administration of Ctrl, NBs, D-NBs, Dc-NBs or Dc-NBs + PTT (Figure S9G), which further confirmed that the use of Dc-NBs for delivery is safe.

Fig. 7

Antitumor effect of Dc-NBs in subcutaneous and metastatic tumor model. A Representative images of subcutaneous xenograft tumors (n = 6/group). B The tumor growth curves for xenografted tumor under treatment as indicated. C The tumor weight revealed that suppressed effects of Dc-NBs + PTT in vivo. D The representative images of HE staining (Scale bar = 100 μm) and immunostaining (ACSL6 and GPX4) analysis in subcutaneous xenograft tumors, Scale bar = 50 μm. E Schematic illustration of lung metastasis model establishment and treatment. F Representative images of IVIS imaging, which revealed the inhibitory effect of Ctrl, NBs, D-NBs, Dc-NBs or Dc-NBs + PTT treatment on lung metastasis of DU145-luc cells in vivo.G and H Representative images of HE staining for mice lung sections, which revealed the metastatic nodules upon treatment with Ctrl, NBs, D-NBs, Dc-NBs or Dc-NBs + PTT (G), and numbers of metastatic nodules in nude mice lung were analyzed (H). I The survival of lung metastasis mice model by treated- Ctrl, NBs, D-NBs, Dc-NBs or Dc-NBs + PTT. J and K Analysis for expression of circBNC2 (J) and miR-4298 (K) in lung metastasis tissues, upon treatment with Ctrl, NBs, D-NBs, Dc-NBs or Dc-NBs + PTT. L The representative images of HE staining analysis of lung metastatic nodules in each group. Scale bar = 200 μm. The IHC staining for Ki67 and ACSL6 in lung metastatic nodules from treated- Ctrl, NBs, D-NBs, Dc-NBs or Dc-NBs + PTT, Scale bar = 50 μm. M Blood biochemistry analysis. ns, no significance; *P < 0.05; **P < 0.01; ***P < 0.001

We then generated lung metastasis models to explore the role of Dc-NBs in regulating the progression of tumor metastasis in vivo (Fig. 7E). Two weeks after the tail vein injection of DU145-luc cells, Dc-NBs were injected into the mice via the tail vein. We subsequently obtained bioluminescence images (BLIs) of the mice to evaluate tumor progression in vivo. As depicted in Fig. 7F, the mice were imaged to visualize the lung metastases. Meanwhile, no apparent change in body weight was observed during the treatment period (Figure S9H). Furthermore, we clearly observed that the fluorescence intensity was weaker in the Dc-NB + PTT group. Similarly, HE staining revealed that, compared with treated-Ctrl, Dc-NBs + PTT inhibited the formation of metastatic nodules in the lung (Fig. 7G and H). Moreover, treatment with Dc-NBs + PTT prolonged the overall survival (OS) of DU145-luc cell-bearing metastatic mice compared with that of the control groups (Fig. 7I). We next detected the expression levels of circBNC2 and miR-4298 in metastatic lung nodes via RT-qPCR assay. As shown in Fig. 7J, treatment with Dc-NBs and Dc-NBs + PTT significantly increased the circBNC2 expression, whereas decreased the miR-4298 expression in lung metastases nodules (Fig. 7K), which compared with treated with Ctrl or NBs group. Furthermore, the results of the IHC analysis revealed that treatment with Dc-NBs + PTT significantly increased the expression of ACSL6 and decreased the expression of Ki67 in lung metastasis nodules (Fig. 7L, S9I and S9J). Additionally, in further blood biochemical analyses, all groups still showed no significant adverse effects (Fig. 7M). These results suggested that the Dc-NBs + PTT treatment strategy effectively suppressed PCa cell metastasis to the lungs of the nude mice.

In conclusion, we demonstrated that circBNC2 activates ferroptosis to inhibit tumor progression. Our findings emphasized that circBNC2/miR-4298/ACSL6 could be exploited as potential therapeutic targets for CRPC therapy (Fig. 8). Moreover, we developed a nanobowl delivery while co-loading circBNC2 and docetaxel, which combined with PTT. The delivery system has no evident side effects and can be a safe agent for further investigation.

Fig. 8

Schematic diagram depicting the tumor-suppressing effects of circBNC2 in prostate cancer. A Schematic illustration indicates the mechanism of circBNC2 in regulating PCa growth via miR-4298/ACSL6-mediated ferroptosis. B Schematic illustration of step-wise fabrication of co-loading DTX and circBNC2 of bowl-like structured Dc-NBs delivery system. C Dc-NBs nanodelivery is a promising nanotherapeutic strategy for treatment of metastatic PCa

Discussion

During the past two decades, despite significant advancements in prostate cancer treatments, the issue of mCRPC has remained a challenging problem in clinical therapy [57]. Therefore, investigating the mechanisms of mCRPC development and identifying key regulatory factors to reverse chemotherapy resistance are highly clinically and scientifically important.

circRNAs are a newly discovered class of noncoding RNAs. Owing to their unique closed-loop structure, they have garnered increasing attention for their role in cancer biology. Numerous studies have confirmed that circRNAs play a significant role in the occurrence and development of PCa [16]. In this study, by analysing the differential expression of circRNAs in gene chips from PCa patients, we identified a potentially tumor-suppressive circRNA, circBNC2. We also demonstrated in collected clinical samples that circBNC2 is downregulated in PCa tissues compared with adjacent tissues. More intriguingly, previous studies confirmed that circBNC2 promotes glycolysis and progression in hepatocellular carcinoma by sponging miR-217 to increase HMGA2 levels [58]. However, in ovarian cancer, circBNC2 inhibits progression by regulating the miR-223-3p/FBXW7 axis [59]. These results indicate that circBNC2 plays a multifaceted role in different contexts. Intriguingly, for the first time, we demonstrated that the overexpression of circBNC2 significantly inhibits the proliferation, migration, and invasion of PCa cells, especially AIPC cells (DU145 and PC3), whereas the knockdown of circBNC2 has the opposite effect. To further elucidate the molecular mechanism by which circBNC2 exerts its tumor-suppressive effects in PCa cells, we performed RNA-seq on circBNC2-overexpressing DU145 cells. Subsequently, KEGG pathway analysis was used to analyse these DEGs, which revealed that circBNC2 is a regulator of ferroptosis in PCa cells.

Ferroptosis is a form of cell death characterized by iron-dependent lipid peroxidation. Extensive studies suggest that ferroptosis plays a pivotal role in tumor suppression, thus providing new opportunities for cancer therapy [60]. Several studies have reported that PCa cells can increase sensitivity to ferroptosis by altering iron metabolism pathways, subsequently increasing iron uptake and storage [61, 62]. Ferroptosis inducers then amplify dysregulated iron metabolism to induce the death of PCa cells. Furthermore, studies have demonstrated that PCa cells can resist lipid peroxidation by upregulating the expression of antioxidant enzymes such as GPX4, thereby reducing the occurrence of ferroptosis [63]. Therefore, the development of novel ferroptosis inducers provides new therapeutic options for PCa. To further explore the mechanism by which circBNC2 promotes ferroptosis in PCa cells, we analyzed the ferroptosis-related DEGs on the basis of the RNA-seq results. We found that ACSL6 was the most significantly upregulated gene among the upregulated genes induced by circBNC2 overexpression in DU145 cells.

Among the ACSL family members, ACSL4 is well known as a ferroptosis activator [64]. ACSL4 promotes the synthesis of acyl-CoA, subsequently increasing long-chain unsaturated fatty acids (such as phosphatidylethanolamine in membrane phospholipids) in cancer cells [65]. These unsaturated fatty acids are more susceptible to iron-induced free radical reactions, thereby promoting ferroptosis. ACSL6 is involved in lipid synthesis and is most abundant in the brain [66]. Further studies revealed that ACSL6 is associated with diseases such as Alzheimer’s disease and schizophrenia [67]. However, few studies have investigated the function of ACSL6 in cancers. Chen et al. demonstrated that the expression of ACSL6 was significantly increased in colorectal cancer (CRC) and that ACSL6 participated in tumor initiation and early tumorigenesis [68]. In CRC cells, the overexpression of ACSL6 promotes fatty acids synthesis to provide intermediate metabolites and energy for cell proliferation [69]. However, a recent study also revealed that ACSL6 was downregulated in other cancers, according to data mining results [68]. In this study, our data demonstrate that the overexpression of ACSL6 significantly increased MDA and Fe²⁺ levels and elevated ROS levels but decreased GSH levels in DU145 and PC3 cells. Conversely, knocking down ACSL6 in DU145 and PC3 cells led to decreased ferroptosis. These data indicate that ACSL6 may be an activator of ferroptosis in PCa cells. More importantly, we also demonstrated that circBNC2 induced ACSL6 expression along with the activation of ferroptosis, suggesting that circBNC2 is a potential upstream regulator of ACSL6 in DU145 and PC3 cells. How does circBNC2 regulate ACSL6 expression? Studies have confirmed that circRNAs act as competing endogenous RNAs (ceRNAs), functioning as miRNA sponges that bind to miRNAs and prevent them from inhibiting their target mRNAs [70]. Through prediction and validation, circBNC2 was shown to act as a molecular sponge for miR-4298, thereby promoting ACSL6 expression. In summary, this study is the first to identify a tumor-suppressive circRNA, circBNC2, in PCa. CircBNC2 drives ferroptosis through the miR-4298/ACSL6 axis, thereby exerting its tumor-suppressive effects in PCa cells.

PTT is a widely used minimally invasive local treatment for cancer [71]. PTT uses mainly photothermal agents to increase the temperature of the tumor site under NIR light irradiation, subsequently inducing cancer cell death. To our knowledge, this is the first study to develop a PTT treatment strategy using a delivery carrier coloaded with DTX and circBNC2 (Dc-NBs) for PCa. We demonstrated, at both the cellular level and the animal experimental level, that the developed Dc-NBs exhibited adequate antitumor effects, high drug loading capacity, good photothermal conversion efficiency, strong NIR absorption, and NIR-responsive drug release capabilities, making them suitable for chemo-photothermal combined cancer therapy. Additionally, animal experiments confirmed that treatment with Dc-NBs did not cause detectable toxicity to the major organs of the animals, indicating that this nanocarrier has good biocompatibility. However, the detailed regulatory mechanism of Dc-NBs in cell proliferation, migration and invasion remains to be investigated in further work. Nonetheless, the therapeutic strategy designed in this study provides a promising direction for addressing the significant clinical problems associated with prostate cancer.

In conclusion, this study revealed the molecular mechanism by which circBNC2 exerts its tumor-suppressive effects by driving the circBNC2/miR-4298/ACSL6 axis to promote ferroptosis in PCa cells. Additionally, this study developed an efficient and safe PTT treatment strategy based on a nanodelivery system that coloads circRNA (circBNC2) and DTX. This research provides new targets for treating CRPC and offers new ideas for developing chemo-photothermal combination therapies targeting CRPC.