The molecular axis hnRNPU/circKCNK2/EDC4/IL-11 aggravates osteolytic bone metastasis of RCC

Abstract

Bone metastasis, which is associated with adverse outcomes, is a serious health concern for renal cell carcinoma (RCC) patients, especially considering the limited therapeutic options. In this study, we investigated the expression profiling of circRNAs in five primary RCC samples and RCC-bone metastases (Bone Met) using high-throughput screening and identified an upregulated circRNA in Bone Met (hsa_circ_0016459, circKCNK2). Notably, overexpression of circKCNK2 could promote osteoclast differentiation and accelerate the destruction of osteolytic bone metastasis by stimulating IL-11 secretion. Additionally, we observed that RCC with a high circKCNK2 expression could benefit from an anti-IL-11 strategy rather than a denosumab-based therapeutic regimen. At the molecular level, circKCNK2 is competitively bound to EDC4 (a scaffold protein of P-bodies). The interaction between circKCNK2 and EDC4α-helical disrupted the combination of DCP1 and DCP2, which weakened the function of P-bodies and resulted in an increased level of IL-11 mRNA and finally activated STAT-3 signaling in osteoclast precursors (OPs). This axis could be blocked with a mutation of EDC4α-helical. Further experiments revealed that increased circKCNK2 production in bone metastases was attributed to decreasing expression of heterogeneous nuclear ribonucleoprotein U (hnRNPU) under an acidic microenvironment. Our findings suggest that circKCNK2 could have a critical role in linking P-bodies to IL-11/STAT-3 signaling. Developing a secure and effective gene delivery system targeted at circKCNK2 is promising for RCC therapy.

Introduction

As a major site of hematogenous tumor cell spread, bones are a commonly targeted metastatic site of renal cell carcinoma (RCC) [1]. Generally, RCC patients with bone metastasis are not only associated with poor prognosis but also have an extremely low quality of life. Osteolytic bone metastasis is typically characterized by skeletal complications, such as fractures, anemia, and nerve compression [2], and this destructive characteristic is attributed to osteoclasts (OCs) [3].

Osteolytic bone metastasis is frequently observed in various cancers, such as breast carcinoma (BCa) and lung cancer [4, 5]. It modulates bone resorption and triggers a “vicious cycle” by releasing growth factors that contribute to the development of bone metastases. Scientists have raised concerns about the severe mortality of RCC-bone metastasis; given the complex condition of the advanced disease stage, the standard treatment for RCC-bone metastasis has not been well defined, and improved treatment strategies are needed [6].

Circular RNA (circRNA), which appears to be highly tissue-specific [7, 8], is identified as a vital regulator in many solid tumor types. However, its role in RCC-bone metastasis remains unclear. Unlike mature messenger RNA (mRNA), circRNAs are commonly described as closed-loop structures lacking 5’-3’ polarity and devoid of a poly-adenylated tail [9]. Regarding the diverse mechanisms it can act through, circRNAs can interact with microRNAs or proteins, regulate transcription and mRNA-splicing, and even translate into peptides [10, 11]. Furthermore, along with the breadth of its regulatory functions, the complexities of circRNA biogenesis are yet to be revealed.

Herein, we identified a circRNA named circKCNK2, a transcript expressed heavily in bone metastases and modulates the secretion of Interleukin-11 (IL-11). As a member of the IL-6 cytokine family, IL-11 distinctly modulates the activity of bone-resorbing OCs [12]. At the molecular level, IL-11 functions by forming a signaling assembly with the glycoprotein 130 (gp130) receptor and then triggering its downstream STAT-3 signaling [13]. Indeed, it has been widely reported that IL-11 can promote tumor growth and invasion [14]. In general, these reactions are mediated by IL-11, which can be produced by either host immune cells or the tumor cells themselves.

In our study, we found that circKCNK2 disrupted the stability of the EDC4/DCP1/DCP2 complex and upregulated IL-11 expression in RCC cells. As an osteolytic factor, IL-11 subsequently promoted bone metastatic lesion development, triggered STAT-3 signaling, and facilitated the differentiation of osteoclast precursors (OPs), which subsequently led to the differentiation of OCs. Moreover, decreasing hnRNPU expression was identified in the acidic microenvironment created by OCs. Downregulation of hnRNPU resulted in circKCNK2 production in RCC. Meanwhile, targeting circKCNK2 with a gene delivery system (MS/PEI-PBA) can significantly alleviate bone lesions in mouse models, indicating that circKCNK2 might function as a clinical diagnostic biomarker and potential therapeutic target for RCC.

Methods

Arraystar human circRNA array analysis

Each sample’s total RNA was measured with a NanoDrop ND-1000 spectrophotometer. The guidelines given by Arraystar dictated the steps for sample preparation and microarray hybridization. Using random primers as per the Super RNA Labeling protocol by Arraystar Inc., the total RNA from each specimen was amplified and then converted into fluorescent cRNA (Arraystar Inc.).

The labeled cRNAs were then hybridized to the Arraystar Human circRNA Array. After the washing steps, an Axon GenePix 4000B microarray scanner was used to scan the arrays. The resulting scanned images were processed using GenePix Pro 6.0 software (Axon) for grid alignment and data extraction. Subsequently, we utilized the R software package for quantile normalization and additional data handling. Hierarchical clustering was used to display the circRNAs that exhibited significant differential expression between the groups.

Patient sample

This research was approval from the Ethics Committees of Renji Hospital and adhered to established ethical standards throughout. The expression levels of circKCNK2 in RCC samples were assessed via qRT-PCR. These samples were procured from 65 patients who underwent surgical resection at Renji Hospital (Renji cohort) in Shanghai. Survival and correlation analyses were performed based on the expression levels within this Renji cohort. Patients were diagnosed independently by three pathologists. Written informed consents were obtained from all patients before collection of tissues, which were used for ISH and IHC. Bone metastasis specimens were typically acquired through either biopsy or surgical excision procedures, the process did not involve the retention of bone tissue.

Cell culture and reagents

The human RCC cell strains 786-O and Caki-1 were sourced from the American Type Culture Collection (Manassas, VA, USA) and propagated following suggested guidelines. All cells were confirmed to be free from mycoplasma contamination and were validated using short tandem repeat (STR) profiling before experimentation. The cells were maintained in MEM (Gibco) supplemented with 10% FBS (Gibco) and incubated at 37 °C in a 5% CO₂ humidified environment.

Conditioned medium collection

In a summarized procedure, RCC cells (either Caki-1 or 786-O) were placed in a 10 cm dish and grown until about 85% confluence was achieved. After washing the cells with phosphate buffer saline (PBS) twice, they were incubated for 24 h in 6 mL Dulbecco’s Modified Eagle Medium (DMEM) which was both serum-free and free of phenol red. The obtained conditioned medium was then cleared of cell debris using a 0.22 μm filter and concentrated via a Millipore protein concentrator (Amicon Ultra-15, 10 kDa). This medium was subjected to centrifugation at 3500 g until its concentration was tripled. To prevent potential damage from multiple freeze-thaw cycles, the medium was portioned out and then frozen at −80 °C and is recommended for use within 3 months.

CRISPR/Cas9 knockout (KO)

To establish IL-11 KOs, we confirmed the essentiality of intronic complementary sequences and their downstream complementary regions. The KO was generated using a standard approach. Generally, single-guide RNAs (sgRNAs) were engineered and integrated into the LentiCRISPRv2 vector (Puro, catalog 52961). Using a mix of LentiCRISPRv2, pVSVg (item 8454), and psPAX2 (item 12260), we generated the CRISPR/Cas9 lentivirus. This lentivirus was subsequently employed to infect RCC cells to induce gene KO.

Western blot

Cultured RCC cells or mouse tumor tissue was subjected to lysis using RIPA buffer (Cell Signaling), complemented with 1 mM PMSF (Beyotime) and a 1× protease/phosphatase inhibitor blend (Roche). To isolate nuclear proteins, the Nuclear and Cytoplasmic Protein Extraction Kit from Beyotime was employed, adhering to the supplied protocol. The bicinchoninic acid (BCA) protein assay kit (Beyotime) was used to assess protein concentrations. Subsequently, each sample, with an amount of 30 μg, was applied to a 4–20% sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gel.

For protein identification, gel-separated proteins were moved to Immuno-blot polyvinylidene fluoride sheets (Roche, with a pore size of 0.22 μm) utilizing the Trans-Blot Turbo Transfer apparatus (Bio-Rad). Afterward, these membranes were subjected to blocking in TBS-T (tris-buffered saline supplemented with 0.1% Tween-20) enriched with 5% skimmed milk powder and incubated at 37 °C for 3 h.

Subsequently, the membranes were exposed to primary antibodies diluted using the primary antibody dilution buffer (Beyotime) and left to incubate overnight at 4 °C. Thereafter, the membranes underwent three wash cycles in TBS-T before being treated with horseradish peroxidase-linked secondary antibodies (prepared in TBS-T with 1% skimmed milk powder) for 60 min at ambient temperature. The chemiluminescent signals were visualized using an enhanced chemiluminescence system (Millipore, Billerica, MA, USA). Detailed information about the specific antibodies employed is provided in Table 1.

Table 1.

Antibodies used in the experiments.

Product	Source	No. of catalogue
Primary antibody:
Western blot:
anti-β-TUBULIN	Cell Signaling Technology	#2146
anti-Histone H3	Abcam	ab1791
anti-H3K27me3	Abcam	ab6002
anti-H3K27ac	Abcam	ab4729
anti-EDC4	Proteintech	17737-1-AP
anti-IL-11	Abcam	ab187167
anti-RBM6	Abcam	ab188318
anti-DDX58	Abcam	ab180675
anti-hnRNPU	Proteintech	14599-1-AP
anti-Lamin B1	Abcam	ab16048
anti-CTSK	Abcam	ab207086
anti-TRAP	Abcam	ab191406
anti-DCP1	Abcam	ab183709
anti-DCP2	Abcam	ab288158
anti-Kla	PTM-BIO	PTM-1401
anti-T-STAT3	Abcam	ab68153
anti-P-STAT3	Abcam	ab76315
anti-E-Cadherin	Abcam	ab212059
anti-N-Cadherin	Abcam	ab76011
anti-KCNK2	Abcam	ab90855
IF:
anti-EDC4	Proteintech	17737-1-AP
anti-GW182	Abcam	ab156173
anti-LSM14A	Proteintech	18336-1-AP
IP:
anti-DCP1	Abcam	ab47811
anti-DCP2	Abcam	ab288158
anti-EDC4	Abcam	ab72408
anti-GST tag	Abcam	ab111947
anti-FLAG tag	Abcam	ab205606
anti-Pol II	Santa Cruz	sc-9001
IHC:
anti-hnRNPU	Proteintech	14599-1-AP
anti-Ki67	Cell Signaling Technology	#9449
anti-QKI	Proteintech	13169-1-AP
anti-SF2	Proteintech	12929-2-AP
anti-SF3A1	Proteintech	15858-1-AP
Secondary antibody:
Western blot:
anti-rabbit IgG-HRP	Cell Signaling Technology	7074
anti-mouse IgG-HRP	Cell Signaling Technology	7076
IHC:
anti-rabbit IgG-HRP	Proteintech	SA00001-2
IF:
anti-rabbit IgG-HRP	Cell Signaling Technology	4413
anti-mouse IgG-HRP	Cell Signaling Technology	4408

Co-immunoprecipitation (co-IP)

For co-IP assays, cells were lysed in IP lysis buffer (0.5% NP-40, 150 mM NaCl, 20 mM HEPES, pH 7.4, 2 mM EDTA, and 1.5 mM MgCl2) supplemented with protease inhibitor cocktail for 40 min on ice, the cell lysates were first treated with EDC4/DCP1/DCP2 antibodies, allowing them to incubate for 4–6 h, followed by interaction with protein A/G agarose beads for an extended period of 12–16 h. After the binding process, the proteins were washed in a lysis buffer. The co-IP samples were further eluted and analyzed by western blot.

In vitro cyclization

Biotin-labeled or unlabeled RNAs were paired with specific DNA splints in a molar ratio of 1:1.5. This combined solution was then exposed to 90 °C for a brief span of 2 min, followed by a gradual cooling to ambient temperature (between 20–25 °C) within 15 min to encourage the formation of the pre-ligation structure. For the creation of circRNAs, ligation activities were conducted overnight at 16 °C using the T4 DNA ligase. This was followed by a treatment phase of 30 min with RNase R and DNase I at 37 °C. Finally, the RNA underwent purification via phenol-chloroform extraction.

In situ hybridization (ISH) analysis and RNA-FISH

CircKCNK2 probes, conjugated with digoxigenin, were formulated according to the guidelines set by Biosearch Technologies (https://www.biosearchtech.com/). The tissue microarray of the tumor was subjected to dual xylene treatments, 100% ethanol (twice), 90% ethanol, and 75% ethanol, followed by a 15 min incubation in 3% H2O2. Afterward, the array was hybridized with probes under non-denaturing conditions. All procedures were conducted in line with the methodologies proposed by Biosearch Technologies. After these processes, the sectioned specimens were visualized using BCIP/NBT, and nuclear fast red served as a contrasting stain. Imaging of these samples was performed using the Nikon Eclipse 80i instrument (Nikon, Tokyo, Japan).

For RNA fluorescence in situ hybridization (RNA-FISH), RCC cells from the specified group were plated on cover glasses and fixed using a fixation buffer (4% formaldehyde, 10% acetic acid) for 10 min. Subsequently, the cells were permeabilized in 70% ethanol overnight and rehydrated in 2× SSC buffer (300 mM NaCl and 30 mM sodium citrate [pH 7.0]) containing 50% formamide. Hybridization was performed overnight at 37 °C with 30 ng of a circKCNK2-specific probe. Biotin-labeled probes were then incubated with Streptavidin-Cy3™ (Sigma-Aldrich, S6402) in 2× SSC buffer supplemented with 8% formamide, 2 mM vanadyl-ribonucleoside complex, and 0.2% RNase-free BSA at 37 °C for 1 h in the dark. Nuclei were counterstained with DAPI (0.1 μg/ml), followed by two washes with 2× SSC buffer containing 8% formamide at room temperature (RT) for 15 min each. Images were captured using a Carl Zeiss laser confocal microscope, with three images taken per cover glass, and representative images are presented.

RNA isolation and qRT-PCR

RNA was isolated using the TRIzol agent (Takara Bro Inc, 9108) following the recommended guidelines from the producer. Using the Hiscript® III Reverse Transcriptase kit (Vazyme, R223-01), 1 μg of RNA specimens were transcribed in reverse to generate cDNA. qRT-PCR was conducted on the QuantStudio™ 3 Real-Time PCR Detection System with the aid of ChamQ Universal SYBR qPCR Master Mix (Vazyme, Q711-02). Using the 2 − ∆∆CT technique, the comparative expression of each gene was measured and standardized based on the internal reference, glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The distinct primers used in this qRT-PCR investigation can be found in Table 2.

Table 2.

Primers and probes used in the experiments.

Gene	Forward primer	Reverse primer	Application
circKCNK2	CCGTTAGGAAACACCTCCAA	GGCAGATTTAGGATCCAGCA	qRT-PCR
circFANCA	AGCTGGACTGCGAGAGAGAG	GTGGAAGAACTGCTCGCATC	qRT-PCR
circPLCL2	CCATCCCAGTCCAGTTCCTA	AAGGGCCCTAGCTCAAGAAG	qRT-PCR
KCNK2	CTCTTGGCTGGAGTTGGAGA	GCCACAAAGAGTACACAGCC	qRT-PCR
FANCA	GCTGCTTATCTCCAGGCCTT	GTGTGTCCAGAGAGAGAGGG	qRT-PCR
PLCL2	GGTTGCCGGAGTGTTGAATT	TGTGCTGAACCATTACCTTCTG	qRT-PCR
IL-11	ATGAACTGTGTTTGCCGCC	CCGTCAGCTGGGAATTTGTC	qRT-PCR
IL-8	CAGTTTTGCCAAGGAGTGCT	ACTTCTCCACAACCCTCTGC	qRT-PCR
IL-6	GACAGCCAACTACGATGATG	GCAAGTCTCCTCATTGAATCC	qRT-PCR
IL-1	TGGAAAAGCGATTTGTCTTCAA	CAGTTATATCCTGGCCGCCT	qRT-PCR
IL-10	GCCAAGCCTTGTCTGAGATG	TGAGGGTCTTCAGGTTCTCC	qRT-PCR
TNF-alpha	GTCAACCTCCTCTCTGCCAT	CCAAAGTAGACCTGCCCAGA	qRT-PCR
RUNX2	CTGTGGTTACTGTCATGGCG	AGGTAGCTACTTGGGGAGGA	qRT-PCR
PTHrP	TTACGGCGACGATTCTTCC	CAGTCACTCCAGAGTCTAACC	qRT-PCR
CXCL-6	CTGGTCCTGTCTCTGCTGTG	CACCTTGGAGCACTGCGG	qRT-PCR
RANKL	TGGGCCAAGATCTCCAACAT	GGTGCTTCCTCCTTTCATCAG	qRT-PCR
VEGFA	GACGGACAGACAGACAGACA	CGAGAACAGCCCAGAAGTTG	qRT-PCR
ET-1	GAGAAACCCACTCCCAGTCC	CCAGGTGGCAGAAGTAGACA	qRT-PCR
COX-2	TGACCCAGAACTACTTTTCAACA	TCCATGTTCCAGCAATATAGAGT	qRT-PCR
RSPO-2	GCGAATGGGGAACTTGTAGC	TGCCTCATTGTCATCTTGCA	qRT-PCR
MCP-1	CAATCAATGCCCCAGTCACC	GGGACACTTGCTGCTGGT	qRT-PCR
DKK-1	AGGTTCTGTTTGTCTCCGGT	CTCCACAGTAACAACGCTGG	qRT-PCR
IGF-1	CCACAGGGTATGGCTCCAG	CTTCTGGGTCTTGGGCATGT	qRT-PCR
TGF-beta	ATCGACATGGAGCTGGTGAA	CTCCTTGGCGTAGTAGTCGG	qRT-PCR
M-CSF	ACCAAGCCTGATTGCAACTG	CTGCCTGGATCCACTGTGT	qRT-PCR
IFN-alpha	GTCCTCCATGAGCTGATCCA	GTTTCTCCCACCCTCTCCTC	qRT-PCR
EDC4	AGCATCTGCGGGACATACTC	CTCCAGAGAGGCAAATGACC	qRT-PCR
hnRNPU	CTGTTCCCGCATGTTCTCTG	CCCTGGATTTTCTGCTGCAT	qRT-PCR
β-Actin	GGACTTCGAGCAAGAGATGG	AGGAAGGAAGGCTGGAAGAG	qRT-PCR
U1	TGAAGGCGCTTTTCTCATGG	CAGGGGAAAGCACGAACG	qRT-PCR
GAPDH	CATGAGAAGTATGACAACAGCC	AGTCCTTCCACGATACCAAAGT	qRT-PCR
Probe-circKCNK2	Biotin 5’- AGGTGCCGCCACTATGGTTG		ISH/FISH

For quantification of circKCNK2 copy number, 8 × 10⁵ RCC cells (786-O or Caki-1) were collected using TRIzol™ Reagent (Invitrogen™, 15596018) and subjected to RNA isolation. DNase treatment and cDNA synthesis were performed utilizing the Maxima™ H Minus cDNA Synthesis Master Mix with dsDNase (Thermo Fisher Scientific, M1682). For qPCR, cDNA corresponding to 0.033% of the total RNA, along with 2 pmol of each primer, was employed in conjunction with Platinum SYBR Green qPCR SuperMix-UDG (Thermo Fisher Scientific, 11733038), following the manufacturer’s protocol.

RNA immunoprecipitation (RIP) assay

The RIP was conducted with the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore). In brief, cells were harvested and subjected to lysis in a cold buffer containing both protease and RNase inhibitors. Following centrifugation at 13,000 × g for 10 min, the supernatant was saved both for the RIP procedure and as a 10% input volume reference. The supernatant underwent overnight incubation at 4 °C with 3 μg of the EDC4-specific antibody (Abcam) or IgG, combined with RIP buffer and 30 μL A/G protein magnetic beads. Subsequently, these beads underwent five wash cycles. RNA bound to the immune-precipitate was then recovered using the aforementioned RNA extraction technique. The retrieved RNA was transcribed in reverse and subsequently assessed using qRT-PCR. When juxtaposed with the IgG control, the percentage of input indicates the fold enhancement in the immunoprecipitated specimens.

Lentivirus constructs and siRNAs

We generated RCC cell lines expressing luciferase-green fluorescent protein (GFP) by transducing the 786-O cells with a lentivirus based on the pLEX vector and encoded the luciferase-GFP fusion gene. Using the pEGFP-C1 vector, we engineered plasmids to express EGFP-bound NBDY. CircKCNK2 originated from the back-splicing of exon 1 of the KCNK2 gene. The circKCNK2-shRNA was synthesized according to the circularization site, and sequences were cloned into the pLVX-shRNA2 interference vector. Full-length circKCNK2 was cloned in a pLC5-ciR vector with KpnI and BamHI restriction enzymes for over-expression. The vector incorporates complementary Alu sequences adjacent to the two restriction enzyme sites. Through co-transcription with the Alu sequences, the linear circKCNK2 initiates back-splicing to generate circKCNK2. Using the Lipo Plus DNA transfection reagent, the pLC5-circKCNK2 plasmid was transfected into the cells.

To produce lentivirus, the target plasmids RRE, REV, and VSVG were co-transfected into HEK293T cells. To achieve lentivirus-mediated gene delivery, RCC cells were transfected with these tailored lentivirus vectors for 24 h. We then cultivated stable expression cell lines from the RCC cell lines through lentivirus infection and subsequent selection using puromycin.

Specific siRNAs targeting circRNAs, hnRNPU, RBM6, and DDX58 were crafted by Tsingke Biotech and transfected with Lipofectamine RNAiMAX (Invitrogen). For circRNA knockdown, we utilized nuclear and cytoplasmic fractions obtained from cellular fractionation and further examined the distribution of candidate circRNAs between the nucleus and cytoplasm using PCR. The results indicated that the majority of three circRNAs are localized in the cytoplasm.

The siRNA is incorporated into the RNA-induced silencing complex (RISC), RISC recognizes and binds to the back-splice junction of the circRNA, leading to its degradation. Therefore, the siRNAs targeting the back-splice junction of circRNAs were designed to specifically knock down cytoplasmic circRNAs without affecting their linear RNA counterparts. A detailed list of these sequences can be found in Table 3.

Table 3.

The sequences of siRNAs.

ID	sense	antisense
si-circKCNK2#1	UACAACCAUAGUGGCGGCA	UGCCGCCACUAUGGUUGUA
si-circKCNK2#2	AUUACAACCAUAGUGGCGG	CCGCCACUAUGGUUGUAAU
si-circFANCA#1	CAUCUGACCUCAAAUGAUCCU	AGGAUCAUUUGAGGUCAGAUG
si-circFANCA#2	CUGACCUCAAAUGAUCCUCCU	AGGAGGAUCAUUUGAGGUCAG
si-circPLCL2#1	UUGAGCUAGGGCCCUUGCUCA	UGAGCAAGGGCCCUAGCUCAA
si-circPLCL2#2	UCUUGAGCUAGGGCCCUUGCU	AGCAAGGGCCCUAGCUCAAGA
si-DDX58#1	GAAGAUCUUGAGGAUAAGA	UCUUAUCCUCAAGAUCUUC
si-DDX58#2	GUAAUGCUGGUGUAAUUGA	UCAAUUACACCAGCAUUAC
si-hnRNPU#1	CAUACAUGAAGUUCGUAUU	AAUACGAACUUCAUGUAUG
si-hnRNPU#2	CCAAUACCUGAAGAGUAUA	UAUACUCUUCAGGUAUUGG
si-RBM6#1	GGUAGACACCGAUCUAGGA	UCCUAGAUCGGUGUCUACC
si-RBM6#2	GGAGCAUUCUUUCAGCUAU	AUAGCUGAAAGAAUGCUCC
si-KCNK2#1	UAGAAAUCCCUUUUUGCUCGU	GAGCAAAAAGGGAUUUCUACU
si-KCNK2#2	UUUCAUAACAUUAAUGGUCGU	GACCAUUAAUGUUAUGAAAUG

Induction of OC differentiation in mouse OPs

To analyze circKCNK2-induced OC differentiation, the bone marrow cells were isolated from the femur and tibia of 6–8-week-old C57BL/6 mice. To ensure a smooth and sterile procedure, begin by spraying the mouse with 70% ethanol to disinfect the fur. Using sterile scissors and forceps, carefully remove the hind legs at the hip joint. After isolating the legs, meticulously remove the muscle and soft tissue to expose the femur and tibia. Cut both ends of the femur and tibia (proximal and distal) to reveal the bone marrow cavity. Next, insert a 1 mL syringe filled with PBS or HBSS, fitted with a needle, into one end of the bone and gently flush out the bone marrow into a sterile container. Pass the resulting bone marrow suspension through a 70 µm cell strainer to eliminate any clumps or bone fragments. Centrifuge the filtered cell suspension at 300 g for 5 min at 4 °C. Finally, discard the supernatant and resuspend the cell pellet in 96-well plates with α–MEM containing 10% FBS and 1% penicillin–streptomycin for 24 h to generate OPs. The attached cells were identified as OPs, which were then incubated for 48 h in completed media (alpha-MEM enriched with 10% FBS and 1% penicillin/streptomycin) with M-CSF (10 ng/mL; PeproTech, Rocky Hill, NJ, USA) to produce OCs. The culture medium was replaced every 2 days. For osteolysis assessments, these OCs were further activated with M-CSF (20 ng/mL) and RANKL (100 ng/mL, PeproTech). The culturing medium changed every 72 h. Subsequently, the differentiated OCs were fixed and subjected to TRAP staining. Mature osteoclasts were defined as TRAP-positive multinucleated cells with 3 or more nuclei.

TRAP staining and HE staining

For the cultured OPs, the culture medium was discarded, and the cells were rinsed thrice using PBS. Subsequently, they were fixed using 4% paraformaldehyde for a quarter of an hour, followed by staining as per the guidelines provided with the TRAP staining kit (Servicebio, Wuhan, China). After 30 min and removal of the staining solution, TRAP activity was assessed colorimetrically. Concurrently, ImageJ software was utilized to evaluate and compute the quantity and dimensions of multinucleated OCs. For paraffin-embedded tissue slices, a similar TRAP staining protocol was applied.

Bone tissues from mice were preserved in a 4% paraformaldehyde mixture before undergoing a two-week decalcification process at 4 °C in 0.5 M ethylenediaminetetraacetic acid (EDTA). The EDTA mixture was replenished weekly. Following this, paraffin slices of 6-μm thickness were stained using tartrate-resistant acid phosphatase (TRAP; Millipore-Sigma, 387A-1KT) or H&E. The captured visuals of TRAP and H&E were facilitated through a Leica microscope (DM4000b). Quantitative evaluations of the OC’s surface, the eroded region, and the tumor expanse were carried out within a range of 0.2 to 2 mm beneath the growth plate. The measurements were recorded using both the OsteoMeasure Analysis System (Osteometrics) and ImageJ software (NIH) following the guidelines provided by the respective manufacturers.

Recombinant protein purification and pull-down assay

Flag-tagged DCP2 produced in 293 T cells were isolated using A/G-Sepharose beads, which attached to antibodies specific to either the Flag markers. Elution of these immunocomplexes was carried out with corresponding peptides. The GST-EDC4 protein, extracted from Escherichia coli, was purified with glutathione Sepharose beads (provided by Beyotime) and subjected to dialysis against a mixture containing 20 mM Tris-HCl (pH 8.0) and 10% glycerol. Then, it was maintained at 4 °C overnight.

For the biotin-RNA pull-down procedure, cell lysates were generated via ultrasonication in RIP buffer composed of 150 mM KCl, 25 mM Tris (pH 7.4), 0.5 mM dithiothreitol, 0.5% NP-40, and a full set of protease inhibitors and RNase inhibitors. These prepared lysates underwent a pre-clearing step with streptavidin magnetic beads (courtesy of Invitrogen). Next, biotin-tagged RNA probes, synthesized in vitro and bound to streptavidin magnetic beads, were introduced to the cell lysates and kept at 4 °C for about 4 h. After this period, the mixtures underwent five wash cycles using the RIP buffer and were finally eluted using an elution buffer.

Intratibial injection

Six-week-old BALB/C nude mice served as the foundation for a bone metastasis model, in line with the guidelines sanctioned by the Renji Hospital’s Committee for the Care of Experimental Animals. To induce bone metastasis, we administered an injection of 5 × 10⁵ RCC cells directly into the mice’s tibia.

In the experiment comparing the efficacy of IL-11 with Denosumab, these mice received an intravenous injection of IL-11 neutralizing antibody (X203, Aldevron) or denosumab (Xgeva; Amgen) administration (5 mg/kg, 2–3 times per week). For the control group, intravenous infusion of the immunoglobulin IgG was administered (5 mg/kg, 2–3 times per week).

In the experiment confirming the effects of MS-based delivery of si-circKCNK2, after a week post-operation, the bone metastasis mouse model was further randomly divided into three subgroups, receiving 30 μl intra-tibial injections of PEI-PBA/si-NC, PEI-PBA/si-circKCNK2, and MS/PEI-PBA/si-circKCNK2, respectively.

After 3 weeks, the mice underwent euthanasia, with their tibiae collected and preserved in 4% formaldehyde. Before being subjected to H&E staining for morphological assessment, the preserved tibiae specimens underwent high-definition micro-CT scanning (Inveon, Siemens, Germany) to evaluate the extent of bone destruction. For a deeper insight, we performed quantitative evaluations on the scanned depictions, focusing on metrics like the BV/TV ratio. This assisted in drawing comparisons and quantifying bone damage across distinct tibial lesion zones within each group.

Orthotopic xenograft model

BALB/c nude mice (SIPPR-BK Experimental Animal Co., China) were used to establish the orthotopic xenograft model. All surgeries were performed under isoflurane anesthesia. Body weight was measured weekly. Six-week-old mice were randomly allocated into two groups (n = 6, per group): vector vs oe-circKCNK2, pcDNA3.1 vs oe-KCNK2 and sh-NC vs sh-hnRNPU. Subsequently, mice were anesthetized, and approximately 1 × 10^{^6} luciferase-labeled 786-O cells mixed with 40% Matrigel (BD, San Jose, CA, USA) were injected into the left subrenal capsule. Tumor growth and spread of tumor cells from the primary site to other organs were monitored every two weeks using the AniView100 in vivo imaging system (BLT, Guangzhou, China). At six weeks, mice were euthanized, and tumor specimens (e.g., lungs, liver, lymph nodes) were harvested for histopathological analysis to identify metastatic lesions using hematoxylin and eosin (H&E), the results suggested by bioluminescence were finally verified through pathological examination. By combining these approaches, a comprehensive analysis of metastasis in the mouse model can be achieved.

Bioinformatics and statistical analysis

The clinical relevance of KCNK2 and hnRNPU in the RCC specimens compared with standard kidney samples was investigated using the GEPIA online resource (http://gepia.cancer-pku.cn/index.html). For graphical representation of the data, we employed the R software (version 4.0.0, accessible at https://www.R-project.org) with the following packages: ggplot2, Complex Heatmap, and Maftools. Survival rates of RCC patients were evaluated using Kaplan-Meier analysis and log-rank tests, primary endpoints were progression-free survival (PFS) per Response Evaluation Criteria in Solid Tumors, version 1.1 (RECIST v1.1) by blinded independent central radiology review. Data were collated from a minimum of three independent experiments and displayed as the average ± standard deviation. For assessments involving three or more groups, one-way analysis of variance (ANOVA) served as the statistical method. Dunnett’s subsequent test was harnessed when juxtaposing every group to a unified control, whereas, for multiple paired comparisons with uniform group sizes, either Tukey’s subsequent test or the least significant difference method was chosen. Comparison of proportion and correlation analysis was evaluated by Chi-square test. All assays, encompassing IP/IB and Immuno-FISH, were conducted with triplicate biological samples. Statistical evaluations were carried out using the GraphPad Prism 7 program.

Study approval

The mice were housed in the Renji Hospital Animal Facility. Every procedure that involved the animals adhered rigorously to the directives established by the Renji Hospital’s Ethical Board. All animal experiments conducted in our research strictly adhere to the highest standards of animal ethics and welfare regulations. Our research complies with international guidelines, including the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines, as well as local and national regulations governing the use of animals in scientific research. We are committed to the principles of the 3Rs (Replacement, Reduction, and Refinement) to minimize animal use, reduce suffering, and improve experimental design. Animals are housed in environments that meet their physiological and behavioral needs, with access to appropriate nutrition, water, and veterinary care. Experimental designs include balanced gender representation, and housing conditions are standardized to minimize environmental variability. These steps ensure that our findings are not biased by gender or external factors. Samples of human RCC, encompassing tissue microarrays, were procured after receiving written consent from the involved individuals. This research was performed in accordance with the Declaration of Helsinki. The Ethics Committee of Renji Hospital granted authorization for this research.

Results

Increased expression of circKCNK2 was associated with RCC progression and bone metastasis

Utilizing the Arraystar human circRNA (V2.0) platform, we performed a differential analysis between five RCC-primary tumors and five RCC-bone metastases (Bone Met). To identify circRNAs potentially contributing to bone metastasis, we focused on the top 3 upregulated circRNAs in Bone Met (Fig. 1a and Table 4 and Table 5), namely, circFANCA (hsa_circ_0040994), circPLCL2 (hsa_circ_0001275), and circKCNK2 (hsa_circ_0016459). To verify the properties of these three circRNAs, we subjected total RNA from RCC tumor tissue to RNase R treatment, an enzyme known to selectively degrade linear RNA [15], before reverse transcription-polymerase chain reaction (RT-PCR) amplification. As a control, the linear transcript of three candidate genes did not show differences between the Primary tumor and Bone Met (Supplemental Fig. 1a).

Fig. 1. Increased expression of circKCNK2 was associated with RCC progression and bone metastasis.

a Heatmap displaying the results of differential expression analysis using the Arraystar Human Circular RNA Microarray, comparing RCC-primary tumor and bone metastases (Bone Met) samples (n = 5, respectively). b 786-O and Caki-1 cells were subjected to migration and invasion assays with siRNAs targeting circKCNK2, circFANCA and circPLCL2 (left), and the quantified results are presented (right). c Correlation between KCNK2 mRNA and circKCNK2 expression was evaluated using RNA extracted from RCC samples. d Efficiency of oe-circKCNK2 was confirmed by qRT-PCR. e Comparison of cell proliferation rates in 786-O (top) and Caki-1 (bottom) cells transfected with lentivirus-expressing circKCNK2 or control vector. g Expression of EMT markers in vector group and oe-circKCNK2 group were detected by western blot. f 786-O and Caki-1 cells overexpressing circKCNK2 were subjected to Transwell-migration and invasion assays (top), cells per field were counted (bottom). h Nude mice were orthotopically xenografted with luciferase-expressing oe-circKCNK2 or control 786-O cells (n = 6), and bioluminescent imaging was used to observe the progression of tumor at indicated time points (2 weeks, 4 weeks and 6 weeks). i Kaplan-Meier survival curves for oe-circKCNK2 group and control group of mice was plotted. P value was calculated by log-rank test. j Representative of histogram analysis (left) of normal kidney, control RCC group and oe-circKCNK2 RCC group, body weight and tumor volume (right) of mice were assessed at sacrifice. k Northern blot validated the specificity of ISH probe across different samples. l Representive images of ISH highlighted elevated circKCNK2 expression in bone metastases compared to metastases in other sites (left), and the quantified results are presented (right). m Comparison of proportion and correlation analysis in RCC patients with different circKCNK2 expression were evaluated by Chi-square test. n RNA samples extracted from RCC (Renji cohort) were subjected to qRT-PCR assays to detect circKCNK2 level, and Kaplan-Meier survival analysis was conducted to stratify them into circKCNK2-low (n = 32) and circKCNK2-high expression (n = 33) groups. o Radionuclide bone scan imaging of RCC patients in circKCNK2-high group (left, case ID: RBM-11; right, case ID: RBM-06), both exhibited tumor progression despite denosumab therapy. One-way ANOVA followed by Dunnett’s post hoc test was performed in b and l, the 2-tailed Student’s t-test was used in c, d, f and j. Two-way ANOVA was performed in e. Error bars show the standard deviation (SD) from at least three independent experiments. ns, not significant; *P < 0.05; **P < 0.01, ***P < 0.001.

Table 4.

Top 10 differential-expressed circRNAs (Primary tumor vs Bone Met) based on the detection of Arraystar human circRNA V2.0 microarray.

circRNA	Alias	t	AveExpr	log2FC	adjP
hsa_circRNA_101911	hsa_circ_0040994	−9.37997	10.017	−3.83608	8.12E-05
hsa_circRNA_000361	hsa_circ_0001275	−8.17524	9.975	−3.3536	0.00014
hsa_circRNA_016459	hsa_circ_0016459	−11.9118142	6.681	−3.02537	2.80E-05
hsa_circRNA_104511	hsa_circ_0082708	−8.61703732	7.442	−3.02458	0.000117
hsa_circRNA_103461	hsa_circ_0067185	−4.085038367	10.010	−2.90581	0.006823
hsa_circRNA_406888	NA	−12.08136344	6.233	−2.86982	2.60E-05
hsa_circRNA_091722	hsa_circ_0091722	−6.326445463	7.186	−2.86135	0.000598
hsa_circRNA_102485	hsa_circ_0050102	−9.134036598	10.873	−2.85821	9.03E-05
hsa_circRNA_104510	hsa_circ_0004069	−7.500209223	7.760	−2.82509	0.000226
hsa_circRNA_406510	NA	−7.215139137	6.084	−2.8026	0.000279

Table 5.

Information of RCC samples in circRNA microarray.

Sample Type	Age	Sex	Sample source	Sample collection	Histological type	Sample collection
Primary Tumor	56	M	Kidney (left)	Partial nephrectomy	ccRCC	High
Primary Tumor	78	F	Kidney (left)	Radical nephrectomy	ccRCC	Low
Primary Tumor	66	M	Kidney (left)	Radical nephrectomy	ccRCC	Low
Primary Tumor	76	M	Kidney (left)	Partial nephrectomy	ccRCC	High
Primary Tumor	61	F	Kidney (left)	Radical nephrectomy	ccRCC	High
Bone metastases	77	M	Rib (right)	Biopsy	ccRCC	High
Bone metastases	70	M	Pelvis	Surgical resection	ccRCC	Low
Bone metastases	80	F	Thoracic vertebra	Surgical resection	ccRCC	Low
Bone metastases	81	F	Rib (right)	Biopsy	ccRCC	Low
Bone metastases	65	M	Pelvis	Surgical resection	ccRCC	Low

F Female, M Male, RBM RCC bone metastasis, PD Progression disease, PR Partial Response, SD Stable disease.

The electrophoresis results revealed that all of these circRNAs were resistant to RNase R digestion (Supplemental Fig. 1b). Sanger sequencing was performed to identify the head-to-tail splice junction of these circRNAs (Supplemental Fig. 1c). Moreover, Northern blot analysis revealed the presence of these three circRNAs in RCC samples. In contrast, their linear transcripts, susceptible to RNase R degradation, were also detected (Supplemental Fig. 1d, e). In addition, the three circular transcripts remained stable under the treatment of actinomycin D (Supplemental Fig. 1f).

Next, we utilized small interfering RNA (siRNA) to examine whether suppressing the expression of each circRNA could alter the malignant phenotype of two RCC cell lines (786-O and Caki-1). After verifying the efficiency of target gene knockdown by northern blot and qRT-PCR (Supplemental Fig. 1g, h), we performed Transwell assays, revealing that the silence of circKCNK2, but not the others, significantly reduced the migratory and invasive capacity of RCC cells (Fig. 1b). We performed an assay to silence the linear KCNK2 mRNA (Supplemental Fig. 1i, j) and subsequently reassessed the impact on migration and invasion of RCC cells. The findings indicated that knocking down KCNK2 did not induce significant alterations in the malignant phenotype of RCC (Supplemental Fig. 1k). Through analyzing the expression levels of circKCNK2 and linear KCNK2 in samples from RCC patients, we found no significant association between linear KCNK2 and circKCNK2 (Fig. 1c).

Overexpression vectors of circKCNK2 and KCNK2 were designed (Supplemental Fig. 1l, m). CircKCNK2 (hsa_circ_0016459) originates from the KCNK2 gene situated on the human chromosome (chr) 1 (Supplemental Fig. 1n), and the length of a total of 429 bp was annotated by CircBase (http://www.circbase.org/). As anticipated, in RCC cells, circKCNK2 exhibited greater resistance to RNase R digestion in comparison to KCNK2 mRNA (Supplemental Fig. 1o). Quantitative real-time PCR (qRT-PCR) assays were performed to assess circKCNK2 expression in cell lines. Relative to 293 T cells (normal control), RCC cells exhibited elevated levels of circKCNK2 and an increased circular/linear KCNK2 ratio, particularly in Caki-1 cells (Supplemental Fig. 1p).

We subsequently established stable KCNK2 (Supplemental Fig. 2a) and circKCNK2 overexpression (oe-circKCNK2) cell lines (Fig. 1d). In vitro experiments revealed that overexpression of circKCNK2 markedly enhanced the malignant phenotype of RCC (Fig. 1e) compared to KCNK2 overexpression alone (Supplemental Fig. 2b). the Transwell assay suggested that overexpression of circKCNK2, rather than KCNK2, could promote the EMT process in RCC, this conclusion that was further validated in Western blot (Fig. 1f, g).

We further evaluated the significance of circKCNK2 in tumorigenesis. Orthotopic xenografts of luciferase-expressing RCC cells were established (Fig. 1h) in nude mice (6 weeks old, n = 6 per group). Survival analysis showed that the oe-circKCNK2 group was closely associated with poor survival (Fig. 1i). After 6 weeks, the left kidneys of mice were enucleated, tumors within the left kidneys were subsequently examined pathologically, we finally observed a lower body weight and a higher tumor volume in the oe-circKCNK2 group (Fig. 1j). Analyzing transcriptome data from the GEPIA database, we observed that the expression level of KCNK2 seems to have no significant impact on the prognosis of RCC patients (Supplemental Fig. 2c). Not surprisingly, in vivo assays suggested that linear KCNK2 did not affect the malignant phenotype of RCC cells (Supplemental Fig. 2d, e).

In situ hybridization (ISH) assay was conducted by using probes with validated specificity (Fig. 1k), and the results indicated an increased circKCNK2 expression in bone metastases compared with other metastatic sites of RCC patients (Fig. 1l). To investigate the correlation between circKCNK2 expression and progression-free survival (PFS) in RCC patients, qRT-PCR assays were conducted using primary RCC tissue. Patients in the circKCNK2-high group had a higher rate of bone metastasis than those in the circKCNK2-low group (Fig. 1m and Table 6). The survival curves implied a notable link between elevated circKCNK2 expression level and poor PFS (Fig. 1n). In particular, drawing on clinical cohorts assessing circKCNK2, we discerned patients suffering from bone metastasis in the high-circKCNK2-expression group exhibited a higher pronounced tumor progression rate and a heightened likelihood of bone-associated complications even after receiving denosumab treatment (a recommended agent for bone metastasis), in contrast to their counterparts with lower circKCNK2 expression (Fig. 1o and Table 7).

Table 6.

Correlation between circKCNK2 expression and clinicopathologic characteristics in RCC patients.

Characteristics	Cases	circKCNK2 expression
		Low	High	p-value
Gender				0.1811
Male	48	26	22
Female	17	6	11
Age				0.4149
<65	50	26	24
≥65	15	6	9
T stage				0.8658
T1-2	42	20	22
T3-4	23	12	11
Bone metastasis				0.0098
Negative	50	29	21
Positive	15	3	12
Pathological type				0.3041
ccRCC	56	29	27
Non-ccRCC	9	3	6

Comparison of proportion and correlation analysis was evaluated by Chi-square test. Bold value indicates statistically significant differences.

Table 7.

RCC patients with different circKCNK2 expression underwent denosumab treatment.

Patient ID	Age	Sex	Bone metastatic site	Denosumab Therapy (Months)	Histological Type	Response	CircKCNK2 expression
RBM-1	70	F	Rib (left)	6	ccRCC	PD	High
RBM-2	58	M	Pelvis	6	ccRCC	PR	Low
RBM-3	59	M	Pelvis	5	ccRCC	SD	Low
RBM-4	72	M	Rib (right)	8	ccRCC	SD	High
RBM-5	65	M	Multiple sites	6	ccRCC	SD	High
RBM-6	66	M	Pelvis	5	ccRCC	PD	High
RBM-7	59	F	Pelvis	8	ccRCC	SD	Low
RBM-8	79	M	Thoracic vertebra	10	ccRCC	SD	Low
RBM-9	81	M	Rib (right)	4	ccRCC	PD	Low
RBM-10	60	F	Rib (leift)	6	ccRCC	PR	Low
RBM-11	59	M	Lumber vertebra	15	ccRCC	PD	High
RBM-12	66	F	Lumber vertebra	12	ccRCC	SD	High

F Female, M Male, RBM RCC bone metastasis, PD Progression disease, PR Partial Response, SD Stable disease.

CircKCNK2 contributed to RCC bone metastasis by promoting OC differentiation

To evaluate the potential influence of circKCNK2 on bone formation or osteolysis in an in vivo setting, we designed and performed experiments with BALB/C nude mice that were intra-tibially injected with either sh-NC-786-O or sh-circKCNK2-786-O cells (Supplemental Fig. 2f). By synthesizing the images of computed tomography (CT) scanning and micro-computed tomography (micro-CT), we observed a notable lower tumor burden and less severe destruction of trabecular bone structure in the mice injected with sh-circKCNK2-786-O cells (Fig. 2a, b). In-depth analysis using micro-CT unveiled an evident rise in Bone Volume to Total Volume (BV/TV) ratio and Trabecular Thickness (Tb.Th) in the sh-circKCNK2 group compared with the sh-NC group, whereas there was a notable reduction in Bone Surface to Bone Volume (BS/BV) ratio and Trabecular Separation (Tb.Sp) in the sh-circKCNK2 group (Fig. 2c). Using hematoxylin and eosin (H&E) staining methods, a marked reduction of tumor size of bone metastases in the sh-circKCNK2 group were confirmed (Fig. 2d, e). Additionally, tartrate-resistant acid phosphatase (TRAP) staining showed a decreased activity of OCs in the sh-circKCNK2 group (Fig. 2f, g). Correspondingly, we performed CT scanning for mice that were intra-tibially injected with oe-circKCNK2 and vector (control) 786-O cells (Supplemental Fig. 2g). H&E staining revealed exacerbated osteolytic lesions in the oe-circKCNK2 group (Supplemental Fig. 2h, i), and TRAP staining indicated increased tumor burden compared with the control group (Supplemental Fig. 2j, k). These data verified that circKCNK2 contributed to the osteolytic bone metastasis of RCC.

Fig. 2. CircKCNK2 contributed to RCC bone metastasis by promoting OC differentiation.

Representative images of CT (a) and micro-CT (b) comparing sh-NC and sh-circKCNK2 group (n = 6). c In-depth analysis of micro-CT was performed to identify BV/TV, Tb.Th, BS/BV and Tb.Sp in the harvested tibia from sh-NC and sh-circKCNK2 group. Representative HE staining (d) from sh-NC and sh-circKCNK2 group were shown, analysis of tumor size in sh-NC and sh-circKCNK2 group were measured (e). Scale bar = 50 μm.Representative TRAP staining (f) from sh-NC and sh-circKCNK2 group were shown, analysis of relative osteolytic lesion area in sh-NC and sh-circKCNK2 group were measured (g). Scale bar = 50 μm. Representative images of osteoclast differentiation and IF staining on mouse OPs were presented (h) as M-CSF only, M-CSF + RANKL, combined M-CSF + RANKL and conditioned media (CM) from si-NC/si-circKCNK2 786-O cells (i, quantification). Representative images of RAW264.7 cells (top chamber) subjected to the Transwell migration assay with CM from 786-O or Caki-1 cells treated with si-NC, si-circKCNK2#1 and circKCNK2#2 (j) with quantification of migrated cell (k). qRT-PCR analysis was performed to detect the cytokines involved in circKCNK2-silenced or overexpression circKCNK2 786-O cells (l), IL-11 was identified as the candidate positively correlated with circKCNK2 expression (m). n Pearson’s correlation analysis was conducted between circKCNK2 and transcription level of IL-11 from RCC patients (Renji cohort). o Analysis of ELISA for si-circKCNK2 or oe-circKCNK2-regulating IL-11 secretion in 786-O and Caki-1 cells. p Evaluation of osteolytic markers in RAW264.7 cells through western blot following co-culture with conditioned medium from si-circKCNK2 cells. The statistical difference was assessed through 2-tailed Student’s t- test in c, e, g, l and n, One-way ANOVA with Tukey’s post hoc test was performed in i, One-way ANOVA followed by Dunnett’s post hoc test in k and o. Error bars show the standard deviation (SD) from at least three independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001.

To unravel the impact of circKCNK2 on bone homeostasis, its effect on OC differentiation was comprehensively investigated based on in vitro models. As a positive control, OC differentiation was induced in both mouse OPs and human peripheral blood mononuclear cells (PBMCs) by priming with 100 ng/ml RANKL. This effect was further enhanced following co-culture with conditioned medium (CM) derived from 786-O cells. However, a reversal effect was observed when co-culturing with CM derived from si-circKCNK2 786-O cells, indicating that the CM from si-circKCNK2 786-O cells significantly inhibited OC differentiation both in mouse OPs and human PBMCs (Fig. 2h, i and Supplemental Fig. 2l, m). Instructively, by utilizing RAW264.7 cells as an alternative OP model, we ascertained that CM from si-circKCNK2 RCC cells diminished OP recruitment more potently than the control (Fig. 2j, k).

Anticipating the involvement of potential regulators in CM from RCC cells, we concentrated our investigation on cytokines related to osteolytic formation [16] (Table 8). The qRT-PCR assessments disclosed that only IL-11 mRNA manifested the corresponding alteration following the modulation of circKCNK2 expression in 786-O cells (Fig. 2l, m). We have further identified a positive correlation between IL-11 and circKCNK2 by extracting RNA from RCC samples and conducting correlation analysis (Fig. 2n). Through enzyme-linked immunosorbent assay (ELISA) and western blot assays, we found that the protein level of IL-11 fluctuated in tandem with changes in circKCNK2 expression (Fig. 2o and Supplemental Fig. 2n). Moreover, upon incubation with the CM from RCC cells transfected with si-circKCNK2, we noticed a corresponding phenotype characterized by a marked reduction of osteolytic markers level (CTSK, TRAP), which may be mediated by the downregulation of IL-11/STAT3 (Fig. 2p and Supplemental Fig. 2o). Conversely, osteolytic markers were significant upregulated in oe-circKCNK2 RCC cells in comparison to control cells (Supplemental Fig. 2p).

Table 8.

OC-associated cytokine transcripts refer to Fig. 2L.

Number	Symbol	Full name
1	CXCL-6	C-X-C motif chemokine ligand 6
2	M-CSF	Macrophagecolony stimulating factor
3	IL-1	Interleukin-1
4	IL-10	Interleukin-10
5	IL-11	Interleukin-11
6	IL-8	Interleukin-8
7	IL-6	Interleukin-6
8	TGF-beta	Transforming growth factor beta
9	TNF-beta	Lymphotoxin alpha
10	ET-1	Endothelin 1
11	COX-2	Cytochrome c oxidase subunit II
12	IGF-1	Insulin like growth factor 1
13	DKK-1	Dickkopf WNT signaling pathway inhibitor 1
14	IFN-alpha	Interferon-alpha 14
15	RSPO2	R-spondin 2
16	MCP-1	Mast cell protease 1
17	RANKL	Tumor necrosis factor superfamily member 11
18	PTHrP	Parathyroid hormone like hormone
19	RUNX2	RUNX family transcription factor 2
20	VEGFA	Vascular endothelial growth factor A

CircKCNK2 was physically bound with EDC4

Given that circRNAs could interact with proteins and regulate gene transcription, we next prepared probes specifically targeting the junction region of circKCNK2 (both sense and antisense), which were incubated with lysates derived from 786-O cells. The resulting eluate was analyzed via silver staining (Fig. 3a). Utilizing mass spectrometry, we identified the most differentially expressed protein corresponding to circKCNK2 as EDC4 (Fig. 3b and Table 9). Interaction between circKCNK2 and EDC4 was reaffirmed through western blot (Fig. 3c).

Fig. 3. CircKCNK2 was physically bound with EDC4.

a, b ChIRP assays were performed by incubating 786-O cell lysates with probeset-circKCNK2 (sense) and probeset-ctrl (antisense), as showen by silver staining (a). Eluents were detected by LC-MS/MS to identify the RNA binding proteins (RBP) of circKCNK2, the annotated MS/MS spectrum assigned to EDC4 peptide, at 150 kDa (b). c ChIRP-western blot assays confirmed the interaction between circKCNK2 and EDC4 proteins. d Secondary structure of circKCNK2 was predicted based on minimum free energy (MFE). Schematic diagram (e) of in vitro RNA/Protein assays were conducted with several biotin-labelled truncations (Sense, Antisense, nt 1-52, nt 52-208, nt 208-312, nt 312-416) to detect the interaction region (f). g RNA pull-down assay was performed between to identify the interaction between nt 52-208 and EDC4 protein. h RIP and qPCR detected the binding of circKCNK2-nt 52-208. i In vitro RNA-protein binding assays using biotinylated circKCNK2 sense or antisense transcripts were conducted with indicated recombinant EDC4 protein. j Schematic illustrating cyclization of linear transcript was generated in vitro (left), both linear and cyclized RNAs were further identified through RT-PCR (right). k The cyclized RNA pull-down assays were performed. l Subcellular fractionation of RCC cells were verified by western blot (top) and the level of circKCNK2 were measured in both nucleus and cytoplasm by qRT-PCR (bottom). U1 and GAPDH were employed to serve as the control of nuclues and cytoplasm, respectively. m FISH and IF detection using probes targeting circKCNK2 (green) and antibody targeting EDC4 (red) in 786-O and Caki-1 cells (left), quantitative analysis was shown (right). Nuclei were stained with DAPI (blue). Scale bar: 60 μm. The 2-tailed Student’s t-test was used to compare the expression level of two groups in h, and the χ2 test in l. Error bars show the standard deviation (SD) from at least three independent experiments. *P < 0.05; **P < 0.01;***P < 0.001.

Table 9.

Results from ChIRP experiments. Top 5 differential proteins were identified by Mass spectrometry.

NO.	Reference	Coverage	Unique peptides	Peptides	Fold change
1	EDC4	151.6	3	3	146.4385178
2	TPR	267.1	53	53	99.71737584
3	EPHA2	108.2	13	12	79.65548689
4	QARS1	87.7	26	26	58.03863458
5	MCM7	81.3	2	2	57.45741533

To characterize this interaction in detail, we obtained the predicted secondary RNA structure of the circKCNK2 sequence. With thorough consideration of its stem-loop structure, several truncated variants were designed (Fig. 3d, e). Through sequential deletion and RNA pull-down assays, as previously described [17], we pinpointed the crucial region (Fragment 2: 52-208 nt) of circKCNK2 for EDC4 interaction (Fig. 3f). Conversely, both RNA pull-down and RNA immunoprecipitation (RIP) experiments confirmed that the interaction could be blocked by the mutation of the binding site (Fragment 2: 52-208 nt) from circKCNK2 (Fig. 3g, h). Referring to the result of the mass spectrometry analysis (predicted EDC4 protein sequence: GGQLQEQLTQQLSQALSSAVAGR, 1016-1038 aa), we performed in vitro binding assays and identified the exact amino acids that comprised the circKCNK2 binding region within the α-helical subunit of EDC4 (Fig. 3i). Likewise, we showed that immunopurified EDC4 was capable of co-precipitating with cyclized biotin-labeled circKCNK2 in vitro (Fig. 3j, k).

Based on subcellular fractionation assays, we noticed that circKCNK2 was preferentially located in the cytoplasm (Fig. 3l). Using probes with validated specificity (Supplemental Fig. 3a), fluorescent in situ hybridization (FISH) and immunofluorescence (IF) tests illustrated that circKCNK2 and EDC4 were predominately co-localized within the cytoplasm of RCC cells (Fig. 3m). These data collectively underscored the association between circKCNK2 and EDC4.

On this basis, we determined to investigate whether circKCNK2 could regulate EDC4. We evaluated EDC4 expression in RCC cells through western blotting. However, neither circKCNK2 overexpression nor knockdown significantly affected transcription or translation EDC4 in RCC cells (Supplemental Fig. 3b–d). Furthermore, we investigated the protein-coding potential of circKCNK2. However, based on its base sequences, no strong evidence of potential internal ribosome entry sites (IRES) was found, indicating that circKCNK2 might not be translated like some previously reported circRNAs [18] (Supplemental Fig. 3e).

As a scaffolding protein of processing bodies (commonly called P-bodies), EDC4 facilitates complex assembly and bolsters interactions between DCP1 and DCP2 by offering binding sites [19]. DCP1 and DCP2, identified as functional enzymes associated with mRNA turnover, are involved in mRNA decay and translational repression. To shed light on the relationship between circKCNK2 and DCP1 or DCP2, we again carried out a ChIRP assay, but no direct evidence emerged to support that circKCNK2 could directly bind with DCP1 or DCP2 (Supplemental Fig. 3f). We tried to ascertain if there are some differences between the RCC samples from the high- and low-circKCNK2-expression groups (Supplemental Fig. 3g), unsurprisingly, no definite trend was displayed. Additionally, as a control, we found three candidate circular transcripts (circKCNK2, circPLCL2 and circFANCA) were preferentially located in the cytoplasm (Supplemental Fig. 3h). Collectively, these results indicated that circKCNK2 could regulate the stability of IL-11, yet this effect might not be dependent on modulating the expression of the EDC4/DCP1/DCP2 complex.

CircKCNK2 activated IL-11 expression by interacting with EDC4^α-helical

To elucidate the influence of circKCNK2 on the EDC4/DCP1/DCP2 complex, we transfected circKCNK2-siRNA into RCC cells and analyzed the relative degree of co-association between endogenous EDC4 and DCP1/DCP2. Co-immunoprecipitation (Co-IP) assays verified that the interaction between DCP2 and EDC4 was intensified in the si-circKCNK2 group, which suggested that the binding between circKCNK2 and the α-helical subunit of EDC4 (EDC4^α-helical) may impinge on the interaction between DCP2 and EDC4. Concomitantly, there was an impairment in the capacity of DCP2 to bind with DCP1, yet the presence of this circular RNA did not have a significant impact on the direct interaction between DCP1 and EDC4 (Fig. 4a–c and Supplemental Fig. 4a–c). Furthermore, the GST pull-down assay showed that in vitro transcribed circKCNK2 disrupted binding between the EDC4 and DCP1/DCP2 complex, but this effect was not observed with circKCNK2 antisense (Fig. 4d). Meanwhile, the FLAG-DCP2 pull-down experiments demonstrated that the circKCNK2 sense strand disrupted the stability of the DCP1/DCP2 complex, while mutation of the α-helical subunit (recombinant EDC4^Δα-helical protein) distinctly restored the interaction between DCP1 and DCP2 (Fig. 4e). These findings supported that circKCNK2 could interact with the α-helical subunit of EDC4 protein and disrupt DCP1/DCP2 complex formation.

Fig. 4. CircKCNK2 activated IL-11 expression by interacting with EDC4^α-helical.

Immunoprecipitation analysis depicted the interaction between EDC4 (a), DCP1 (b), and DCP2 (c) in 786-O cells transfected with si-NC and si-circKCNK2. d An in vitro GST pull-down assay was conducted. Recombinant GST-EDC4 protein-conjugated beads were incubated with cell lysates in the presence of circKCNK2 or antisense RNA. e In vitro FLAG pull-down assay was performed by incubating recombinant DCP2 conjugated beads with cell lysates and indicated recombinant proteins (GST-EDC4-WT or GST- EDC4^Δα-helical) and RNAs (AS-circKCNK2/Sense-circKCNK2). DCP2-associated proteins were subsequently eluted and detected via immunoprecipitation assay using indicated antibodies. f 786-O cells (left) and Caki-1 cells (right) were transfected with GFP-tagged NBDY to identify the role of circKCNK2 in P-bodies assembly, the IL-11 expression was measured by qRT-PCR. g Images generated through confocal microscopy of IF staining of 786-O and Caki-1 cells with indicated treatments. Cell transfected with NBDY was set as a positive control. White arrows indicate P-bodies as defined by red staining (GW182). scale bars represent 25 μm. The stability of IL-11 was assessed by qRT-PCR at the indicated time points (2 h, 4 h, 6 h and 8 h) by using actinomycin D on EDC4-WT or EDC4^Δα-Helical cells (h), reverse transcribed IL-11 RNA was shown by gel electrophoresis (i). j Stability of IL-11 mRNA evaluated by qRT-PCR at the indicated time points (2 h, 4 h, 6 h and 8 h) by using actinomycin D on indicated groups. Protein level of IL-11 was analyzed by ELISA (k) and western blotting (l) in indicated groups. One-way ANOVA with Tukey’s multiple comparison post-test was used in f and k. Two-way ANOVA was used in h and j. Error bars show the standard deviation (SD) from at least three independent experiments.*P < 0.05;**P < 0.01; ***P < 0.001.

P-bodies are ribonucleoprotein granules enriched with mRNAs, RNA-binding proteins, and mediators of mRNA degradation [20]. We sought to ascertain whether circKCNK2-mediated regulation of P-bodies participates in the decay of IL-11 mRNA. Serving as a positive control, RCC cells were transfected with a plasmid (EGFP-N1) carrying NBDY, which was recognized as a microprotein that impedes the assembly of P-bodies [21, 22]. Upon assessing the transcription and protein expression levels of IL-11, we discovered that the dissolution of P-bodies due to NBDY counteracted the transcriptional inhibition of IL-11 induced by circKCNK2 knockdown (Fig. 4f and Supplemental Fig. 4d). Based on the results of nuclear run-on assays, we have ruled out the possibility of circKCNK2 regulating IL-11 expression at the transcriptional level (Supplemental Fig. 4e).

IF staining showed that the oe-circKCNK2 RCC cells displayed diffused cytoplasmic staining (GW182 and LSM14A, markers of P-body), while the EDC4 mutants (EDC4^Δα-helical) reversed this inhibition effect on P-body formation (Fig. 4g and Supplemental Fig. 4f).

Notably, we observed that the IL-11 mRNA level was less stable in RCC cells transfected with EDC4^Δα-helical compared with the control group (EDC4-WT) when treated with Actinomycin D (Fig. 4h, i). It is also worth mentioning that the transfection of EDC^Δα-helical in 786-O and Caki-1 appeared to disrupt the oe-circKCNK2-stimulated stabilization of IL-11 mRNA (Fig. 4j). We further assessed the IL-11 protein level via both ELISA and western blot assays, and the results confirmed consistently that circKCNK2 upregulated IL-11 expression through its interaction with EDC4^α-helical (Fig. 4k, l). These data implied that increased circKCNK2 levels could enhance IL-11 expression by binding with EDC4^α-helical.

Neutralizing IL-11 significantly alleviated osteolysis and reduced the tumor burden in RCC-bone metastasis with high circKCNK2 expression

To determine whether IL-11 is necessary for the progression of RCC-bone metastasis, IL-11 knockout RCC cells were first generated using the CRISPR/Cas9 approach (Supplemental Fig. 5a). Correspondingly, in vitro OC differentiation assays revealed that oe-circKCNK2 notably promoted OC differentiation in human PBMCs; impairment by IL-11 knockout in oe-circKCNK2 RCC cell-induced OC differentiation was also observed (Supplemental Fig. 5b–c). Similarly, these results were obtained in mouse OPs (Supplemental Fig. 5d, e). OC biomarkers, assessed by western blot assays, consistently underscored the importance of the circKCNK2/IL-11/STAT-3 signaling pathway in both mouse OPs and RAW264.7 (Supplemental Fig. 5f, g). These findings suggested that circKCNK2 regulated OC differentiation by altering IL-11 levels.

To demonstrate our prior clinical observation, we injected vector (control) and oe-circKCNK2 786-O cells into the tibia of BALB/C nude mice with IgG, IL-11 neutralizing antibody, or denosumab (6 weeks old, n = 6 per group) and eventually harvested the tibia after 3 weeks. At the designated time point, we detected the sample using both CT scanning and micro-CT (Fig. 5a, b). For quantification, osteolysis was further determined by the bone parameter (BV/TV, Tb.Th, BS/BV, and Tb.Sp) between the aforementioned groups (Fig. 5c), which indicated that the use of an IL-11 antibody for neutralization may be more effective in suppressing osteolytic bone metastasis in the oe-circKCNK2 group compared to denosumab. HE staining showed that employing an anti-IL-11 antibody in vivo, in contrast to denosumab, significantly mitigated the excessive bone metastasis burden induced by oe-circKCNK2 (Fig. 5d, e). TRAP staining verified that neutralizing IL-11 intensively improved the conditions of osteolytic bone lesions in the mice in the oe-circKCNK2 group (Fig. 5f, g). Immunohistochemistry (IHC) analysis of Ki67 revealed that oe-circKCNK2 markedly escalated the malignancy of bone metastases, while administration of the IL-11 neutralizing antibody, in contrast to denosumab, notably attenuated the impact of oe-circKCNK2 (Supplemental Fig. 5h, i). Accordingly, we designed the Transwell assays and cell proliferation assays to demonstrate that IL-11 knockout could mitigate the promoting effect of circKCNK2 overexpression on the malignant phenotype of RCC (Supplemental Fig. 5j, l). These findings collectively validated that the therapeutic efficacy of IL-11 inhibition surpassed that of denosumab in cases of bone metastasis characterized by high circKCNK2 expression.

Fig. 5. Neutralizing IL-11 significantly alleviated osteolysis and reduced the tumor burden in RCC-bone metastasis with high circKCNK2 expression.

Representative images of CT (a) and micro-CT (b) were shown for tibiae from mice. Control or oe-circKCNK2 786-O cells inoculated mice were treated with IgG, IL-11 neutralizing antibody (anti-IL-11) and denosumab, respectively (n = 6). c Quantitative analysis of bone parameters determined by micro-CT (presented as BV/TV, Tb.Th, BS/BV and Tb.Sp). d, e Representative image of HE staining of vector+IgG, oe-circKCNK2+IgG, vector+anti-IL-11, oe-circKCNK2+anti-IL-11, vector+denosumab and oe-circKCNK2+denosumab group (d). Scale bar: 50 μm. Quantitative analysis of tumor area (e). f, g Representative image of TRAP staining of vector+IgG, oe-circKCNK2+IgG, vector+anti-IL-11, oe-circKCNK2+anti-IL-11, vector+denosumab and oe-circKCNK2+denosumab (f). Scale bar: 50 μm. Quantitative analysis of osteolytic area (g). One-way ANOVA with Tukey’s post hoc test was used in c, e and g. Error bars show the standard deviation (SD) from at least three independent experiments. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001.

The hnRNPU was involved in circKCNK2 biogenesis

Transcription factors play pivotal roles in the production of circular transcripts [23, 24]. To identify potential regulators responsible for the upregulation of circKCNK2, we utilized catRAPID (http://s.tartaglialab.com/page/catrapid_group), an algorithm designed to assess the binding affinity of protein-RNA pairs. After analysis, we selected the top three candidates (RBM6, DDX58, and hnRNPU) with the highest comprehensive scores as the potential RBP of circKCNK2 (Supplemental Fig. 6a). To determine if any of these three candidates could indeed influence the biogenesis of circKCNK2, especially given binding of RBPs to circRNAs is not a prerequisite for RBPs to regulate circRNA biogenesis, we designed small interfering RNAs (si-RBM6, si-DDX58 and si-hnRNPU) and transfected them into the RCC cell lines (Supplemental Fig. 6b–d). Silencing of hnRNPU, but not other candidates, led to an increase in circKCNK2 expression levels in both 786-O (Fig. 6a) and Caki-1 cells (Fig. 6b). Notably, silencing of hnRNPU did not influence the expression of linear KCNK2 mRNA (Fig. 6c). The hnRNPU was involved in various cellular operations, ranging from RNA splicing and gene regulation to RNA stability. Its pivotal role in these processes signifies that any dysregulation of hnRNPU potentially leads to various diseases, notably cancer and neurological disorders [25, 26].

Fig. 6. The hnRNPU was involved in circKCNK2 biogenesis.

a, b Expressions of circKCNK2 were analyzed by qRT-PCR when transfected with si-DDX58, si-hnRNPU and si-RBM6. c Expression of KCNK2 mRNA were detected by qRT-PCR in 786-O and Caki-1 transfected with si-hnRNPU. The hnRNPU expression level in RCC and bone metastases were measured by IHC staining (d) and western blot (e). f Expression level of hnRNPU was detected by western blot between circKCNK2-high expression and circKCNK2-low expression group. Survival analysis of OS (g) or DFS (h) between low and high hnRNPU expression group were plotted by the Kaplan-Meier methods with two-tailed log-rank test. Date was obtained from GEPIA dataset. i Survival analysis of PFS between low and high hnRNPU expression RCC patients stratified by qRT-PCR (Renji cohort). Kaplan-Meier methods with two-tailed log-rank test. j Pearson’s correlation analysis was conducted between circKCNK2 and transcription level of hnRNPU from RCC patients (Renji cohort). k, l The hnRNPU binding sites (blue) detected by eCLIP-seq data are shown. The binding site was verified by RIP assay. The transcript abundance of amplicons a–d relative to input was detected by qRT-PCR (l). m Nude mice were orthotopically xenografted with luciferase-expressing sh-hnRNPU or control 786-O cells, bioluminescent imaging was used to observe tumor progression at indicated time points (2 weeks, 4 weeks and 6 weeks). n Kaplan-Meier survival curves for two groups were plotted. o Metastases (top) and body weight of mice (middle) were recorded from sh-NC and sh-hnRNPU groups. Data in a–d and o are mean ± SD; at least three independent per experiment. One-way ANOVA with Dunnett’s post hoc test was used in a-c. Unpaired 2-tailed Student’s t test was used in d and o. ns, not significant;*P < 0.05; **P < 0.01, ***P < 0.001.

Immunohistochemical analysis of RCC tissue microarrays unveiled that the hnRNPU level was markedly diminished in bone metastases (Bone Met) compared with primary tumor (Fig. 6d). This distinction was further corroborated by western blot analysis (Fig. 6e). In evaluating tumor tissue samples from both high- and low-circKCNK2-expression groups, we scrutinized hnRNPU expression, affirming an inverse correlation with circKCNK2 (Fig. 6f). Survival analysis generated from the GEPIA database (http://gepia.cancer-pku.cn) was carried out to assess the prognostic value of hnRNPU. We found that low hnRNPU expression was significantly associated with shorter overall survival (OS) in RCC patients [P < 0.001] (Fig. 6g). Kaplan-Meier analysis also hinted at a shorter disease-free survival (DFS) in the low-hnRNPU expression group [P = 0.061] (Fig. 6h). These findings were confirmed by analyzing the qRT-PCR data from our clinical cohort, which clearly showed that low expression of hnRNPU correlated with a poor progression-free survival (PFS) in RCC patients (Fig. 6i). Simultaneously, correlation analysis revealed that hnRNPU mRNA is inversely related to the circKCNK2 level (Fig. 6j). By evaluating eCLIP-seq data from ENCODE (ENCSR368GER, eCLIP experiment on human adrenal gland tissue) [27], we exhibited potential binding sites for hnRNPU within the exons of the unspliced KCNK2 precursor-mRNA (Fig. 6k). Subsequently, RIP-qRT-PCR assays validated the interaction between hnRNPU and exon 1 of KCNK2 (Fig. 6l), which was near the genomic loci of circKCNK2. Furthermore, OC differentiation assays demonstrated that the silencing of circKCNK2 partially reversed the OC differentiation effect stimulated by si-hnRNPU (Supplemental Fig. 6e, f).

As regulatory factors are closely associated with circRNA biogenesis, we also aimed to investigate the correlation of QKI, SF2, and SF3a1 with circKCNK2 in RCC samples. We performed immunohistochemistry to assess the staining intensity of these three factors in tissue samples. However, the analysis indicated there was no significant differences in the expression of QKI, SF2, and SF3a1 among samples with varying levels of circKCNK2 expression (Supplemental Fig. 6g, h).

Next, we extended our research to investigate the significance of hnRNPU knockdown in tumor progression. Initially, hnRNPU knockdown (sh-hnRNPU) RCC cells were established (Supplemental Fig. 6i). Additionally, we conducted nuclear run-on assays to confirm that hnRNPU did regulate circKCNK2 at the transcriptional level (Supplemental Fig. 6j). Cell proliferation assays revealed an enhanced proliferation rate in the sh-hnRNPU group (Supplemental Fig. 6k). Wound healing assays noted that cell migration was facilitated in the sh-hnRNPU group (Supplemental Fig. 6l, m). Also, sh-hnRNPU significantly promoted cell invasion in Transwell-invasion experiments (Supplemental Fig. 6n, o). In vivo assays were performed by xenografted tumors into the subrenal capsules of the mice. Bioluminescence imaging at different time points (2-week, 4-week, and 6-week) revealed that sh-hnRNPU accelerated tumor growth in vivo (Fig. 6m). More importantly, the sh-hnRNPU group showed shorter survival (Fig. 6n), higher rate of metastasis and lower body weight than the sh-NC group (Fig. 6o). These findings imply that downregulated hnRNPU leads to an increase in circKCNK2 production.

OCs inhibited hnRNPU transcription in RCC by creating an acidic microenvironment

The microenvironment in target organs provides a biological foundation for the progression of metastatic tumors. To determine whether OPs or OCs impact tumor progression, we designed Transwell assays by co-culturing RCC cells with CM from OPs or OCs (Fig. 7a and Supplemental Fig. 7a, b). Results suggest that co-culturing with CM from OCs, but not OPs, significantly promoted both migration (Fig. 7b, c) and invasion of RCC (Supplemental Fig. 7c, d). Notably, we found a dramatic decrease of hnRNPU protein expression in RCC cells treated with CM derived from OCs (Fig. 7d). The notion that the acidic microenvironment generated by excess OC activity leads to irreversible dissolution of bone mineral and organic degradation during osteoclastic resorption is widely recognized [28]. To determine whether the expression pattern of hnRNPU could be altered as a result of low pH conditions, we cultured 786-O and Caki-1 cells under both acidic conditions, tested pH values accordingly (Supplemental Fig. 7e), and conducted the BCECF-AM assay (Fig. 7e and Supplemental Fig. 7f). After co-culturing with OCs or under conditions of decreased intracellular pH (pH = 6.8) for 48 h, the protein level of hnRNPU was significantly downregulated (Fig. 7f). The qRT-PCR results revealed that the mRNA level of hnRNPU was also decreased (Fig. 7g).

Fig. 7. OCs inhibited hnRNPU transcription in RCC by creating an acidic microenvironment.

a Sketch diagram for Transwell assay was designed to test the effect of conditioned medium derived from OPs and OCs on RCC cells. Representative images of Transwell-migration assay, 786-O and Caki-1 cells treated with indicated conditioned medium were shown (b), cells per field was counted (c). d Protein level of hnRNPU in 786-O (top) and Caki-1 cells (bottom) were measured under CM derived from OCs. CM derived from OPs was employed as control. e Images of the fluorescence intensity detected by BCECF-AM assay in 786-O cells co-culture with conditioned medium derived from OPs or OCs. 786-O cells cultured under completed medium (pH = 6.8 or pH = 7.4) were employed as control. f Western blot assays were performed to examine the expression of hnRNPU in 786-O (top) and Caki-1(bottom) cells with control medium (pH = 7.4), acid medium (pH = 6.8) or conditioned medium derived from OCs, respectively. g Effect of low pH and OCs on hnRNPU mRNA were examined in 786-O (left) and Caki-1 cells (right) by qRT-PCR at indicated time points (0 h, 12 h, 24 h and 48 h). h IGV tracks of ATAC-seq enriched from upstream of the TSSs throughout the whole ranges of the hnRNPU gene in cells cultured under control group (Red, pH = 7.4) and acid-treated group (Blue,pH = 6.8;Green,pH = 6.6.i Expression of hnRNPU and H3K27ac of indicated groups were detected in 786-O (top) and Caki-1 (bottom) cells by western blot assays. ITSA-1 was set as a positive control group. H3K27me3 was analyzed as the negative control. H3 was measured as internal reference protein. j ChIP-qPCR detection of Pol II occupancy on the promoter of hnRNPU in cells cultured under control group (Red, pH = 7.4) and acid-treated group (Blue, pH = 6.8; Green, pH = 6.6). Data in c, g, i and j are mean ± SD; at least three independent per experiment. One-way ANOVA with Tukey’s post hoc test was used in c. One-way ANOVA with Dunnett’s post hoc test was used in g and j. ns, not significant; *P < 0.05; **P < 0.01, ***P < 0.001.

Various stressors linked to tumors, including hypoxia, acidification due to lactic acid, and even nutrient deficiencies, can influence gene expression at both transcriptional and post-transcriptional stages [29–31]. Interestingly, no alteration in hnRNPU expression in lactate-treated RCC cells was detected, suggesting that lactate modification might not be a critical factor in regulating hnRNPU gene expression (Supplemental Fig. 7g).

To elucidate how acid exposure affects hnRNPU expression, we employed ATAC-seq to determine if acid-base imbalances alter chromatin accessibility in RCC (Supplemental Fig. 7h–j). Under non-acidic conditions, open chromatin regions within the hnRNPU gene were predominantly located upstream of the transcription start site (TSS); however, in a relatively acidic environment, chromatin accessibility decreased (Fig. 7h). This phenomenon could be attributed to enhanced histone deacetylation prompted by low extracellular pH level, as previously described [32]. To substantiate our hypothesis, we treated the control and acid-exposed cells with Trichostatin A (TSA), an inhibitor for histone deacetylase, or inhibitors of TSA (ITSA-1). Western blot assays indicated that augmenting histone acetylation through the administration of an HDAC inhibitor could counteract the transcriptional suppression of hnRNPU in RCC cells induced by the acidic environment (Fig. 7i). The impact of TSA or ITSA-1 on in vitro histone deacetylase (HDAC) activity was evaluated concurrently (Supplemental Fig. 7k). Furthermore, we conducted a Pol II-ChIP experiment, which corroborated that the transcriptional activity of hnRNPU is modulated by the acidic environment (Fig. 7j). It was remarkable that the peak levels of control genes (DESI2, COX20 and KIF26B, upstream and downstream of hnRNPU) indicated that in a relatively acidic environment, there were slight alterations in the overall chromatin accessibility of genomic loci, however, none of them was significant compared to hnRNPU (Supplemental Fig. 7l). These findings suggest that the acidic environment generated by OCs may reduce chromatin accessibility in RCC by diminishing histone acetylation.

MS-based delivery of si-circKCNK2 diminished tumor burden in an RCC-bone metastasis mouse model

Emerging therapeutic strategies in oncology research highlight the potential of advanced biomaterial-aided vector transport. However, its capability for precise RNA delivery to address metastatic tumors remains largely unexplored.

In our previous work, we demonstrated that psh-circSTC2-lipo@MS, a secure and modifiable targeted gene transport system, exhibited great potential in maintaining extracellular matrix (ECM) metabolism equilibrium within aberrant microenvironments in mouse nucleus pulposus (NP) cells [33]. In the current study, we synthesized an efficient cationic polymer siRNA vector by coupling phenylboronic acid (PBA) to branch polyethyleneimine (PEI) with a 25-kDa molecular (PEI-PBA). The transfecting polyplexes were prepared by mixing circKCNK2 siRNA (si-circKCNK2).

The RNA protective capacity of the PEI-PBA/si-circKCNK2 poly-complexes was examined under treatment with RNase A. Our findings indicate that naked si-circKCNK2, serving as the control group, was highly susceptible to RNase A degradation, having been completely degraded within a 2 h incubation period. By contrast, PEI-PBA encapsulated si-circKCNK2 remained undegraded even after incubation with higher concentrations of RNase A (Fig. 8a), demonstrating the robust efficacy of PEI-PBA/si-circKCNK2 polycomplexes in protecting RNA from degradation. To ascertain whether RNA could be taken up during transfection, we incubated RCC cells with either free FAM-si-circKCNK2 or PEI-PBA/FAM-si-circKCNK2 poly-complexes. In line with previous research [34], confocal microscopy revealed detectable cellular internalization of PEI-PBA/FAM-si-circKCNK2 after 2 h of incubation. By contrast, no uptake was observed for free FAM-si-circKCNK2 (Fig. 8b). Next, RCC cells were incubated with PEI-PBA/si-circKCNK2 polyplexes for 48 h to evaluate the transfection efficiency. Contrary to the low uptake seen in the free si-circKCNK2 group, PEI-PBA/si-circKCNK2 polyplexes significantly reduced both circKCNK2 mRNA level and downstream IL-11 protein level (Fig. 8c, d). We then evaluated the cytotoxicity of PEI-PBA on RCC cells through the CCK-8 assay. Results indicated that RCC cell viability remained largely unaffected within the PEI concentration range of 20–80 μg/mL (Supplemental Fig. 8a).

Fig. 8. MS-based delivery of si-circKCNK2 diminished tumor burden in an RCC-bone metastasis mouse model.

a The stability of PEI-PBA/si-circKCNK2 polycomplexes were measured under RNase A treatment by agarose gel electrophoresis. b Confocal images of 786-O and Caki-1 cells incubated with FAM-si-circKCNK2 or polycomplexes for 2 h. Cell nuclei were identified by DAPI (blue). Gene silencing efficacy of PEI-PBA/si-circKCNK2 were detected by qRT-PCR (c) and western blot (d). e Morphology features of MS/PEI-PBA/si-circKCNK2 were shown as following: dispersibility (left) and microscopy images (center) of MS, fluorescence microscopy images (right) of MS/PEI-PBA/si-circKCNK2. f SEM image of MS. g Detection of IL-11 expression in RCC cells treated with Lipo 3000/si-circKCNK2, PEI-PBA/si-circKCNK2 and MS/PEI-PBA/si-circKCNK2. h The mechanism of MS/PEI-PBA/si-circKCNK2 in the treatment of RCC-bone metastasis. i Schematic diagram of the experimental procedure was designed: Established tumor-bearing mice were injected with PEI-PBA/si-NC, PEI-PBA/si-circKCNK2 and MS/PEI-PBA/si-circKCNK2 (n = 6) and then being evaluated in three weeks (i). j Representative images of CT scan for indicated groups. Representative images of micro-CT (k) for the tibiae of mice in indicated groups, and bone parameters (BV/TV, Tb.Th, BS/BV and Tb.Sp) were measured as presented (l). Representive images of HE staining for indicated groups (m), tumor area for each group was quantified (n). Representive images of TRAP staining for indicated groups (o), osteolytic lesion area for each group was quantified (p). One-way ANOVA with Dunnett’s post hoc test was used in c, one-way ANOVA with Tukey’s post hoc test in l, n and p. ns, not significant; *P < 0.05; **P < 0.01, ***P < 0.001.

However, as previous research has suggested, introducing formulated siRNA polycomplexes systemically can lead to systemic toxicity and immunotoxic reactions [35]. To address this constraint, we constructed a “circKCNK2 silencing-microsphere” by loading PEI-PBA/si-circKCNK2 polyplexes into injectable gelatin methacrylate (GelMa) microspheres (MS). Being expected to exhibit low cytotoxicity, local drug delivery, well-controlled release, and minimized adverse reactions, MS also showed good dispersion in aqueous solution and was visible under the microscope, and assembled MS/PEI-PBA/si-circKCNK2 was visualized under fluorescence and confocal microscopes (Fig. 8e and Supplemental Fig. 8b). Scanning electron microscopy (SEM) also revealed a porous and spherical structure of MS (Fig. 8f).

After verifying the transfection efficiency in vitro (Fig. 8g), we expanded our investigation in vivo by injecting MS/PEI-PBA/si-circKCNK2 (Fig. 8h), PEI-PBA/si-circKCNK2, or PEI-PBA/si-NC (control group) directly into the established mice model (6 weeks old, n = 6 per group), hoping that circKCNK2 silencing system would suppress osteolytic bone metastasis via the sustained release of si-circKCNK2 (Fig. 8i). At the evaluated time points (after 3 weeks), we observed that the injection of MS/PEI-PBA/si-circKCNK2 postponed the progression of bone metastasis and suppressed bone resorption in both CT and micro-CT scanning (Fig. 8j, k). An in-depth analysis of micro-CT revealed significant differences in the measured parameters in the diaphysis of the right tibia. Specifically, BV/TV and Tb.Th were markedly higher in the MS/PEI-PBA/si-circKCNK2 group compared with the PEI-PBA/si-NC group. Conversely, BS/BV and Tb.Sp of the diaphysis were notably increased in the MS/PEI-PBA/si-circKCNK2 group (Fig. 8l). The effect of the MS-based delivery system on the tibia was completely confirmed by H&E staining (Fig. 8m). As evidenced by quantitative analysis, the tumor burden was much more reduced in the MS/PEI-PBA/si-circKCNK2 group (Fig. 8n). TRAP staining revealed less bone destruction in the MS/PEI-PBA/si-circKCNK2 group (Fig. 8o), with a decreased number of OCs observed in the quantification (Fig. 8p). Moreover, the administration of MS/PEI-PBA/si-circKCNK2 in mice demonstrated remarkable biosafety. No significant weight loss was observed during the entire treatment period (Supplemental Fig. 8c). H&E analysis of major organs, such as the heart, liver, lung, and kidney, indicated no apparent damage to viscera organs (Supplemental Fig. 8d).

Overall, MS/PEI-PBA/si-circKCNK2 demonstrated a promising therapeutic effect. Given the potential adverse reactions caused by systemic therapy, this safe and stable MS-based treatment may be recommended, especially for RCC bone metastasis with high circKCNK2 expression.

CircKCNK2 does not affects OC differentiation via extracellular secretion

CircKCNK2 was likely to regulate OC function by indirectly influencing IL-11, however, whether it directly affects OC differentiation via extracellular secretion remains to be further investigated.

To elucidate the direct effects of circKCNK2 on OPs or OCs, we subjected OPs to oe-circKCNK2 plasmid (Supplemental Fig. 9a) and further conducted OC differentiation assays (Supplemental Fig. 9b) and examined OC-related protein markers (Supplemental Fig. 9c). These results indicated that overexpression of circKCNK2 does not exert a potential influence on OC differentiation. Considering the heterogeneity among cells, circKCNK2 may not exert a direct effect in OPs as in RCC cells but could indirectly influence OC differentiation through the IL-11 signaling axis in tumor cells.

After incubating OPs with the supernatant from RCC culture overexpressing circKCNK2, no significant increase in circKCNK2 expression was detected in OC (Supplemental Fig. 9d). This also implies that circKCNK2 may not be extracellularly secreted by RCC, or alternatively, it might be secreted but is incapable of being internalized by OPs.

Discussion

Bone metastases are considered incurable and pose a significant challenge to researchers. Current treatment strategies primarily focus on blocking bone destruction and targeting OCs. Interleukin-11 (IL-11), a cytokine pivotal to the mechanism of bone metastases, plays an essential role in natural bone turnover. Beyond its role in bone reshaping through the JAK-STAT3 pathway, IL-11 also partakes in a destructive feedback loop involving both bone degradation and solid tumor expansion [36, 37]. Approaches aimed at human IL-11 or its associated signaling routes have been explored in diverse tumor types using pre-clinical models [38, 39]. IL-11 neutralizing antibodies can directly target IL-11, theoretically inhibiting bone metastasis by blocking the IL-11/STAT-3 signaling axis. It also has a certain research foundation and has been applied in many tumors. However, as a systemic therapy, the potential adverse reactions it may cause cannot be ignored. The adverse reactions that IL-11 neutralizing antibodies may cause include but are not limited to: headache, nausea, vomiting, diarrhea, fever, fatigue, decreased appetite, rash, allergic reactions, and possible thrombocytopenia [40].

Bisphosphonates and denosumab (the RANK ligand inhibitor) are both clinically available treatment options for RCC bone metastases [41]. They have both demonstrated considerable potential in reducing skeletal-related events (SREs) and enhancing the quality of life of affected individuals. Denosumab, a fully human monoclonal antibody targeting RANKL, successfully obstructs the interaction between RANKL and RANK, thereby thwarting OC activity, bone resorption, and even tumor progression. An additional advantage is its subcutaneous administration route, which may enhance patient adherence. However, denosumab is not devoid of side effects akin to those seen with bisphosphonates, such as hypocalcemia and osteonecrosis of the jaw.

Conventional radiotherapy stands as a cornerstone in the treatment of bone metastasis. It is primarily employed for palliative purposes, aiding in pain management and the prevention of impending fractures or spinal cord compressions [42]. In particular, the application of stereotactic body radiation therapy (SBRT) in treating non-spine bone metastases (NSBM) is witnessing a rising adoption in clinical practice [43]. However, the implementation of these advanced radiotherapy techniques hinges on the availability of specialized equipment and skilled personnel. The inherent complexity and heterogeneity of bone metastases, along with the diversity in tumor origins, often contribute to less-than-optimal outcomes with routine radiotherapy. Particularly in the context of RCC, which exhibits radiation-resistant histology, escalated dosage is frequently considered [44, 45]. This approach, however, may be accompanied by adverse events, including dermatitis, mucositis, pneumonitis, and fibrosis.

Moreover, Systemic therapeutic approaches (TKIs and immunotherapy) recommended by EAU guidelines [46–48] often combine bone-modifying agents (such as denosumab) for treating RCC patients with bone metastases.

In our study, we carried out a comprehensive comparison of these approaches, suggesting that MSs loaded with PEI-PBA/circKCNK2-siRNA polyparticles could reduce osteolytic reactions and suppress tumor progression. This local delivery method demonstrated a significant, sustained release effect, suggesting that it could be a viable therapeutic option for metastatic tumors, especially when targeting oligometastatic disease. The concept of oligometastatic disease hypothesizes an intermediate state of metastatic progression [49]. Overly aggressive treatment regimens might lead to adverse drug reactions, which are unwarranted and ought to be prevented. Consequently, it is advisable to pursue localized treatment strategies of high efficacy. This strategy can offer significant clinical advantages and extend survival compared with systemic therapies.

In addition, understanding the mechanisms of gene transcription in cancer cells is vital for advancing therapeutic approaches. Our study unveiled a hitherto unappreciated molecular connection between circular RNA (circRNA)-related signaling and the regulation of mRNA stability by processing bodies (P-bodies). P-bodies, recognized as membrane-less organelles, can store or degrade mRNAs in a translationally regulated manner [50]. This signaling, in turn, governs the stability of interleukin 11 (IL-11) mRNA and the bone metastasis microenvironment in RCC.

While much attention has been concentrated on the mRNA housed within P-bodies, the protein components therein are highly likely to play a pivotal role in initiating their intrinsic biological activities. When the 5’-cap structure is eliminated by the DCP2 decapping enzyme and its collaborator DCP1, the translation mechanism ceases, leading to 5’-to-3’ exonucleolytic degradation of messenger RNA [51]. While DCP1 and DCP2 have a direct interaction, the accessory protein EDC4 can amplify the binding between DCP1 and DCP2 [52]. The acquisition of EDC4, along with the presence of oligo-metazoan domains in both DCP1 and EDC4, ultimately activates the decapping efficiency of DCP2.

The molecular underpinnings of EDC4/DCP1/DCP2 triple complexes have not been completely elucidated, especially concerning their assembly and activation. In the research carried out by She et al., a sextuple DCP1 mutant was generated, featuring mutations in both the NR-loop and the EVH1 domain. In yeast cells, these mutations disrupted DCP2 binding and decapping, but the specific contributions of these effects were not further investigated [53]. Aglietti et al. affirmed the essential role of Hs DCP2 W44 in the catalytic mechanism [54]. There are also other aspects warranting investigation. Our findings illuminated how circKCNK2, a circular transcript, regulates the IL-11-containing P-bodies subunit. This circular transcript interrupts the stability of EDC4/DCP1/DCP2 triple complexes, with the binding to EDC4-α-helical resulting in the disruption of DCP1/DCP2. Eventually, this unstable spatial structure reduced the degradation efficiency of P-bodies.

The hnRNPs family is known to control the normal splicing process of various transcripts [55], and multiple hnRNPs have been reported to inhibit the production of circRNAs in Drosophila [56]. Given the evidence suggesting that reduced hnRNPU expression was inversely related to the survival of RCC patients, we believe that anomalous splicing stemming from the depletion of hnRNPU may indeed trigger tumor progression by generating specific transcripts with biological functions.

Stressors associated with tumors, like oxygen deprivation and nutrient scarcity, are known to modulate gene expression at both transcriptional and post-transcriptional stages [57]. In our research, we focused on the crucially acidic niche developed from OCs and discovered that an acidic environment downregulates the expression level of hnRNPU through the deacetylation of histones, further influencing chromatin accessibility. Acidity is identified as a fundamental characteristic of the tumor microenvironment, and it facilitates tumor progression by providing an energy source [58]. Indeed, the comparatively low pH reflects the microenvironment that tumor cells would experience in bone metastatic regions and perpetuates a “vicious cycle” of hnRNPU-associated circKCNK2 production in RCC.

In summary, our study underscored that circKCNK2, as a key regulator of IL-11, links it to the progression of bone metastasis. It represents a potential therapeutic target for clinical drug discovery aimed at combating RCC-induced bone metastasis. However, the translation of these findings into clinical practice necessitates extensive clinical trials to thoroughly assess the safety and efficacy of nanotechnology-based approaches for circKCNK2 inhibition in RCC patients with bone metastasis.

Author contributions

YW executed the experiments, wrote the manuscript and helped designed this study. DZ helped execute the experiments. JL collected the human specimens. SZ helped collect the human specimens. QW maintained the mice. JX helped maintain the mice. NH analyzed data. WC helped analyze data. FW provided resources. JZ supervised the experiments. WX helped analyze data. WZ designed the study and helped write the article.

Funding

This work was supported by the National Natural Science Foundation of China [grant numbers, 82173214 and 82204440]; The Program of Disciplines’ Construction and Talents’ Cultivation in Renji Hospital [2024-Wei Zhai]; The Innovative Research Team of High-level Local Universities in Shanghai, Hospital-pharma Integration Project on Innovative Achievement Translation [grant numbers, SHDC2022CRD022]; Standardized Management Research Project of Shenkang Hospital Development Center [grant numbers, SHDC22021202]; Joint Funds for the innovation of Science and Technology, Fujian province [grant numbers, 2024Y9142]; Clinical Research Incubation Project of Renji Hospital [grant numbers, PNO-0036 and RJPY-DZX-004]; Advanced Program of National Natural Science Foundation in Renji Hospital [grant numbers, RJTJ22-ZD-005]; Program from Institute of Hospital Management, National Health Commission [grant numbers, gjwjwyyglyjs231853]; Project from Shanghai Science and Technology Commission [grant numbers, 23Y21900400]; Project from Shanghai Municipal Health Commission [grant numbers, JKKPZX-2023-A05]; National Key Research and Development Program-Vertical Collaboration Project [grant numbers, 2022YFC2505301]; Startup Fund for Scientific Research of Fujian Medical University (2023QH1065, 2024QH1071); Shanghai Municipal Science and Technology Commission Computational Biology Project (23JS1400803).

Data availability

All data are included in the manuscript. Raw data would be available upon request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41388-025-03476-z.