CircABCA1 promotes ccRCC by reprogramming cholesterol metabolism and facilitating M2 macrophage polarization through IGF2BP3-mediated stabilization of SCARB1 mRNA

Abstract

Background

The reliance of clear cell renal cell carcinoma (ccRCC) on exogenous cholesterol import implies a metabolic susceptibility. This susceptibility represents a potential avenue that can be exploited as a novel therapeutic approach for ccRCC. Circular RNAs (circRNAs) are emerging regulators in cancer, yet their roles in ccRCC lipid metabolism and tumor microenvironment remodeling remain unclear. This study investigates the tumor-promoting role of circABCA1 in ccRCC cholesterol homeostasis and M2 macrophage polarization.

Methods

The expression levels of circABCA1, IGF2BP3, SCARB1, autophagy-related proteins, and the IGF1R/PI3K/AKT/mTOR and ABCA1/ABCG1 pathways were measured using RT-qPCR and western blot. Untargeted metabolomics, RNA- sequencing, and MS2 RNA-pulldown were conducted to identify targets. Interaction analyses included RNA immunoprecipitation, RNA pull-down, and RNA fluorescence in situ hybridization (FISH) assays. Lipid raft measurements, cholesterol uptake/efflux assays, and lipophagy assessments were performed. A co-culture system between M2 macrophages and ccRCC cells was established. In vivo tumorigenesis and metastasis were evaluated using xenograft models and a hepatic metastasis model. Statistical analyses involved Student’s t-tests and ANOVA; significance set at P < 0.05.

Results

We identified a novel lipid metabolism-related circRNA, circABCA1, which was upregulated in ccRCC and positively correlated with tumor stage and distant metastasis. Functionally, circABCA1 enhanced the half-life of SCARB1 mRNA by forming a circABCA1-IGF2BP3-SCARB1 mRNA ternary complex, thereby increasing the expression of SCARB1 and consequent cholesterol uptake. Next, elevated cholesterol caused by circABCA1-SCARB1 axis-maintained lipid rafts, initiated IGF1R/PI3K/AKT/mTOR cascade, and protected lipid droplets from being destructed by lipophagy, leading to decreased cholesterol efflux. CircABCA1 facilitated the proliferation and migration of ccRCC in vitro and in vivo in a SCARB1 depended manner. Moreover, we uncovered that circABCA1 facilitated M2 macrophage polarization and subsequent pro-tumor effect by prompting cholesterol uptake of ccRCC from tumor microenvironment in a SCARB1-dependent manner.

Conclusions

CircABCA1 plays a crucial role in promoting ccRCC progression by regulating cholesterol metabolism and facilitating M2 macrophage polarization, representing a potential therapeutic target for ccRCC treatment.

Supplementary information

The online version contains supplementary material available at 10.1186/s12943-025-02398-4.

Background

Clear cell renal cell carcinoma (ccRCC), the most prevalent and lethal subtype of kidney cancer, is distinguished by its hallmark"clear cell"phenotype, driven by aberrant lipid accumulation [1, 2]. This metabolic reprogramming—marked by disrupted cholesterol homeostasis and lipid droplet overload—fuels tumor progression, metastasis, and therapy resistance, positioning ccRCC as a quintessential metabolic malignancy [3, 4]. Despite advances in targeted therapies, the 5-year survival rate for advanced ccRCC remains dismal (< 15%), underscoring the urgent need to unravel the molecular drivers of lipid metabolic dysregulation and their crosstalk with the tumor microenvironment.

Circular RNAs (circRNAs) constitute a subclass of non-coding RNAs distinguished by their unique covalently closed, single-stranded structure devoid of 5'to 3'polarity and poly A tails. These molecules are ubiquitous in diverse cancers and are renowned for their superior stability compared to their linear counterparts. Consequently, deregulated circRNAs impact the hallmarks of cancer by functioning as microRNA (miRNA) sponges, protein decoys, or templates for encoding small peptides [5, 6] and have been proved to participate in lipid metabolism in various cancers. For example, hsa_circ_0086414 suppresses lipid accumulation in ccRCC cells by acting as miR-661 sponge in ccRCC [7]. In addition, circACC1 accelerates fatty acid β-oxidation in colorectal cancer by forming a ternary complex with AMPK [8]. Nuclear p113, encoded by CUX1-derived circRNA, forms a transcriptional regulatory complex that promotes fatty aldehyde conversion into fatty acids and fatty acid β-oxidation in neuroblastoma [9]. However, the involvement of circRNAs in lipid metabolism regulation of ccRCC remains largely unexplored. There is an urgent need to identify additional circRNAs participating in ccRCC lipid metabolism and elucidate their molecular mechanisms.

RNA binding protein (RBP)-insulin-like growth factor 2 mRNA-binding protein 3 (IGF2BP3) is a well-characterized tumor-promoting factor that promotes tumorigenesis by interacting with mRNAs and non-coding RNAs [10–12]. In ccRCC, IGF2BP3 has been shown to regulate downstream mRNAs through non-coding RNA-mediated interactions. For instance, the long non-coding RNA DMDRMR interacts with IGF2BP3 to stabilize CDK4 and extracellular matrix components via m6A-dependent enhancement of IGF2BP3 activity, promoting G1-S transition and ccRCC proliferation [13]. However, the circRNA-binding activity of IGF2BP3 in ccRCC remains uncharacterized, necessitating mechanistic exploration.

Scavenger Receptor Class B Member 1 (SCARB1) plays a pivotal role in cholesterol uptake of cancer cells. It has been established as a crucial indicator of proliferation and migration in various cancers, including prostate cancer, nasopharyngeal carcinoma, and chronic myeloid leukemia [14, 15]. SCARB1 plays a crucial role in the proliferation and survival of ccRCC cells and stands as a promising new target for therapy [16]. Given pronounced sensitivity to cholesterol levels of ccRCC, exploring factors targeting SCARB1 to modulate cholesterol metabolism might offer novel avenues for ccRCC therapy. However, the underlying molecular mechanism responsible for increased expression of SCARB1 remains elusive.

Cholesterol metabolism in ccRCC is reprogrammed to favor substantial cholesteryl ester (CE) accumulation, which constitutes the primary component of lipid droplets abundantly deposited in these cells [17]. This metabolic rewiring is orchestrated by upregulation of cholesterol esterification and transport proteins such as sterol O-acyltransferase (SOAT1), which mediates exogenous cholesterol uptake, catalyzes its esterification, and facilitates CE storage. Pharmacological inhibition of CE accumulation effectively suppresses ccRCC growth and metastasis [18, 19]. Beyond canonical neutral CE hydrolase-dependent pathways, emerging evidence implicates lipophagy as an alternative CE hydrolysis mechanism. Triggered by elevated macroautophagy, lipophagy delivers lipid droplets to lysosomes, where lysosomal acid lipase (LAL) hydrolyzes CE to generate free cholesterol, predominantly destined for cellular efflux [20, 21].

Emerging evidence show that circRNAs could regulate cholesterol metabolism through de novo synthesis or uptake-storage in cancers. For instance, 121-amino acid peptide encoded by circINSIG1 enhances cholesterol biosynthesis in malignancies [22]. CircLDLR maintains SOAT1 expression by competitively binding miR-30a-3p, increasing the contents of low-density lipoprotein cholesterol in colorectal cancer [23]. However, specific mechanisms underlying circRNA-mediated cholesterol dysregulation in ccRCC require further elucidation.

The tumor immune microenvironment plays a key role in cancer progression and response to immunotherapy in ccRCC. Studies demonstrated that tumor-infiltrating immune cells constitute over 90% of stromal components in ccRCC specimens and advanced ccRCC exhibits increased infiltration of immunosuppressive M2-like macrophages, suggesting macrophage polarization modulation as a potential therapeutic strategy [24]. Recent evidences indicate that cholesterol metabolism not only supports cancer cell growth but also plays a vital role in the reprogramming of tumor-associated macrophages (TAMs) [25, 26]. Concurrently, TAMs polarized to an immunosuppressive M2 phenotype are hallmarks of advanced ccRCC, yet how cholesterol flux shapes this process is unclear. CircRNAs, with their ability to stabilize mRNAs and orchestrate signaling cascades, emerge as compelling candidates to bridge these gaps. However, no circRNA has been reported to simultaneously govern cholesterol metabolism and macrophage polarization in ccRCC—a critical niche we address. Here, we identify circABCA1, a novel tumor-promoting circRNA, as a master regulator of cholesterol homeostasis and TAMs polarization in ccRCC, unveiling a previously unrecognized axis with dual therapeutic potential.

In this study, we uncover circABCA1 as a linchpin of ccRCC progression through two interconnected axes: 1. Cholesterol metabolic reprogramming: CircABCA1 stabilizes SCARB1 mRNA via direct interaction with IGF2BP3, enhancing cholesterol uptake while suppressing efflux by shielding lipid droplets from lipophagy, creating a feedforward loop that fuels tumor growth. 2. Immune microenvironment remodeling: By restricting cholesterol efflux, circABCA1 depletes extracellular cholesterol in the tumor microenvironment, thereby skewing macrophages toward pro-tumor M2 polarization. Clinically, circABCA1 expression positively correlates with advanced tumor stage, metastasis, and proliferation, positioning it as a prognostic biomarker. By elucidating this circRNA-centric regulatory network, our study deepens the understanding of ccRCC pathogenesis and opens avenues for precision therapies targeting the circABCA1-IGF2BP3-SCARB1 axis.

Results

Identification of circABCA1, a circular RNA implicated in lipid metabolism of ccRCC

Given that the extensive accumulation of lipid droplet is a distinctive feature of ccRCC, we consistently observed significant lipid accumulation in ccRCC tissue via oil red and H&E staining (Fig. S1A and B). In addition, lipidomics analysis confirmed that triglycerides (TG) and CE were the most prominently altered lipids in ccRCC tissue (Fig. 1A).

Fig. 1.

Identification of circABCA1, a circular RNA implicated in lipid metabolism of ccRCC. A Volcano plots of the lipid metabolomics with significant changes in ccRCC. n = 3 technical replicates per group. B The workflow for circRNA screening in 4 paired ccRCC tissues was indicated. C Genomic loci were shown, and Sanger sequencing was used to validate head-to-tail splicing of circABCA1. PCR were conducted on both cDNA and gDNA to detect ABCA1 mRNA and circABCA1, employing both divergent and convergent primers. D RT-qPCR analysis for circABCA1 and linearABCA1 in total RNA reverse transcribed by random 6 mers or oligo dT primers. E RT-qPCR was used to detect the relative expression of linear ABCA1 and circABCA1, whose total RNA from Caki-1 and 786-O cells was digested with or without RNase R. F RT-qPCR assay was performed to detect the stabilities of circABCA1 and linearABCA1 after actinomycin D treatment in Caki-1 or 786-O cells. G Subcellular distribution of circABCA1 determined by RT-qPCR assays. β-actin served as the cytoplasmic marker, while U6 transcripts were utilized as the nuclear marker. H RNA fluorescence in situ hybridization (FISH) for circABCA1 (red) and nuclei were stained with DAPI (blue) in Caki-1 or 786-O cells. Scale bar = 10 µm. All Data were presented as mean ± SD. Student’s t-test, n = 3. NS means P ≥ 0.05, ^**P < 0.01, ^***P < 0.001

To further investigate the mechanism of regulating lipid metabolism in ccRCC, we performed ceRNA microarray in four matched ccRCC tissues and adjacent noncancerous tissues, and 3749 differentially expressed circRNAs and 3760 differentially expressed mRNAs were screened. We chose 962 circRNAs implicated in TG and cholesterol (CHO) metabolism by Gene Ontology (GO) analysis. Among the candidate circRNAs, circABCA1 (hsa_circ_0087805) was prioritized for functional validation based on its distinct expression pattern compared to its parental gene, ABCA1, a feature not observed in other candidate circRNAs. Specifically, circABCA1 was upregulated, whereas ABCA1 mRNA was downregulated in ccRCC. This discordant expression trend suggests potential unique role of the circABCA1 transcript (Fig. 1B).

CircABCA1 originates from exons 35 to 41 of the ABCA1 gene, located on chromosome 9 at the q31.1 locus. Sanger sequencing analysis confirmed the predicted splicing junction of circABCA1. Next, we designed convergent and divergent primers, and confirmed that circABCA1 was specifically amplified using cDNA, whereas linear ABCA1 mRNA was amplified from both cDNA and gDNA templates using convergent primers (Fig. 1C). Additionally, random primers rather than oligo dT primers successfully amplified circABCA1, thereby providing further evidence of its circular structure (Fig. 1D). Notably, circABCA1 exhibited superior stability compared with linear ABCA1 after the treatment by RNase R or actinomycin D treatment (Fig. 1E and F). Moreover, subcellular fractionation and FISH assays revealed that circABCA1 was predominantly localized in the cytoplasm (Fig. 1G and H). Collectively, these findings underscore the stability of circABCA1 as a circRNA in ccRCC cells.

Upregulated circABCA1 correlates with clinicopathological characteristics, regulates the viability and migration of ccRCC, and maintains the high CHO level in ccRCC

We delved into the clinical significance of circABCA1 and its effects on ccRCC cells. The expression of circABCA1 in 154 paired ccRCC was upregulated compared with noncancerous tissues. The elevated expression of circABCA1 was observed in a substantial proportion (136 out of 154) of ccRCC patients. Furthermore, the ccRCC sample cohort was stratified based on WHO/ISUP grade, tumor T stage, the presence of distant metastasis. Specifically, higher levels of circABCA1 were associated with a higher WHO/ISUP stage, a more advanced tumor T stage, and the presence of distant metastasis. Moreover, higher levels of circABCA1 were confirmed in the Ki67 ≥ 10% group than in the Ki67 < 10% group (Fig. 2A-F).

Fig. 2.

Upregulated circABCA1 correlates with clinicopathological characteristics, regulates the viability and migration of ccRCC, and maintains the high CHO level in ccRCC. A, B RT-qPCR were utilized to detect circABCA1 expression in adjacent or ccRCC tissue. C-F RT-qPCR were utilized to detect circABCA1 expression in C different WHO/ISUP stages (I + II [n = 114] vs. III + IV [n = 40)]), D Tumor stages (T1a–T1b [n = 130] vs. T2–T4 [n = 24]), E distant metastases (No [n = 145] vs. Yes [n = 9]) and F Ki67 level (Ki67 < 10% [n = 74] vs. Ki67 ≥ 10% [n = 44]). G-I CCK-8 and colony formation rescue assay were performed to detect the cell viability activity; Transwell assays were performed in Caki-1 and 786-O cells to detect the cell migration activity after treatment with si-circABCA1. Scale bar = 50 μm. J RNA-seq of cells with circABCA1 knockdown and Gene Ontology (GO) analysis of differential expression genes was conducted. K Representative photographs of ccRCC cells treated with si-circABCA1 for 24 h. L The total-cholesterol (T-CHO) level was detected in Caki-1 or 786-O cells after the overexpression of circABCA1 or interference of circABCA1. M T-CHO level was detected in Caki-1 or 786-O cells treated with MβCD and ov-circABCA1. N The triglyceride (TG) level was detected in Caki-1 or 786-O cells after the overexpression of circABCA1 or interference of circABCA1. NS means P ≥ 0.05, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001. All Data were presented as mean ± SD. Student’s t-test, n = 3

To assess the functional impact of circABCA1 on ccRCC cells, initial analysis revealed that the expression of circABCA1 was elevated in ccRCC cells compared with HK-2 cells (human kidney cell line) (Fig. S2A). Subsequently, we designed three small interfering RNAs (si-RNAs) targeting the back-splice junction region of circABCA1 based on CircInteractome (circinteractome.nia.nih.gov/). Notably, only si-circABCA1#1 demonstrated efficient knockdown of circABCA1 in both Caki-1 and 786-O cell lines, without affecting ABCA1 mRNA expression (Fig. S2B and C). Next, as shown in Fig. 2G-I and Fig. S2D, CCK-8, Colony formation, Transwell assays revealed that the knockdown of circABCA1 inhibited the proliferation and migration of ccRCC cells.

We next explored the potential role of circABCA1 in lipid metabolism, RNA-seq analyses were performed on ccRCC cells transfected with si-circABCA1. The differentially expressed mRNAs comprised CHO metabolic genes. However, few TG metabolic genes were identified, suggesting that circABCA1 might be primarily involved in the regulation of CHO metabolism (Fig. 2J). Next, si-circABCA1-treated ccRCC cells exhibited a decrease in cellular density accompanied by shortened cellular extensions. Furthermore, some cells displayed a shrunken morphology, which aligning with previous studies that similar cellular morphologies in conditions of low CHO levels [18], implying that the knockdown of circABCA1 may substantially reduce CHO levels (Fig. 2K).

Furthermore, we examined the effect of circABCA1 on total CHO (T-CHO) levels, as shown in Fig. 2L, T-CHO was down-regulated with circABCA1 knockdown and was restored by overexpression of circABCA1 (ov-circABCA1). Next, as shown in Fig. 2M, ov-circABCA1 partly elevated T-CHO level diminished by methyl-β-cyclodextrin (MβCD), a CHO scavenger for cells, indicating that the high expression of circABCA1 can elevate CHO level in ccRCC cells. In contrast, the regulatory effect of circABCA1 on TG did not reach statistical significance (Fig. 2N). Taken together, circABCA1 is positively associated with tumor stage and distant metastasis, and functions as a potential oncogene to enhance CHO levels in ccRCC.

CircABCA1 induces the expression of SCARB1 to maintain HDL uptake and inhibit CHO efflux

To identify the most likely downstream target of circABCA1 in regulating cholesterol levels, we performed RNA-seq analysis after silencing circABCA1 and found that the SCARB1 mRNA exhibited the most profound fold change (|fold change|= 3.92) and statistically significant alteration (|log₁₀P|= 5.71) in the CHO metabolic pathways (Fig. 3A). Furthermore, circABCA1 expression showed significant positive correlation with SCARB1 mRNA levels in ccRCC tissues (n = 137, r = 0.556, P < 0.001). SCARB1 expression can be induced by the overexpression of circABCA1 and down-regulated by si-circABCA1. Moreover, ov-circABCA1 rescued the decreased level of SCARB1 mediated by si-circABCA1 (Fig. 3B-D), indicating that SCARB1 was the downstream target of circABCA1 in ccRCC cells. Next, we investigated whether the reduction in T-CHO levels upon knockdown of circABCA1 was dependent on SCARB1. As shown in Fig. 3E, si-circABCA1 resulted in a decrease in T-CHO levels, whereas overexpression of SCARB1 (ov-SCARB1) had no significant effect on T-CHO levels. However, ov-SCARB1 counteracted the reduction in T-CHO levels caused by si-circABCA1, indicating that circABCA1 regulated T-CHO levels in a manner dependent on SCARB1.

Fig. 3.

CircABCA1 induces the expression of SCARB1 to maintain HDL uptake and inhibit CHO efflux. A The fold change and P value of cholesterol storage relative genes with treatment by si-circABCA1 were detected by RNA-seq. B Correlation analysis was performed between circABCA1 and SCARB1 expression in ccRCC. C, D The protein levels of SCARB1 in Caki-1 or 786-O cells with the overexpression or interference of circABCA1. E T-CHO level was detected in Caki-1 or 786-O cells after the overexpression of SCARB1 or interference of circABCA1. F–H HDL uptake levels were assessed using Dil-HDL and analyzed by flow cytometry and fluorescent microscopy imaging after the overexpression of SCARB1 or interference of circABCA1. Scale bar = 20 µm. I Cholesterol uptake levels were assessed using BODIPY-cholesterol and analyzed by flow cytometry after the overexpression of SCARB1 or interference of circABCA1. J Rate of cholesterol efflux was measured after the overexpression of SCARB1 or interference of circABCA1. NS means P ≥ 0.05, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001. All Data were presented as mean ± SD. Student’s t-test, n = 3

Considering that SCARB1 is known for its functions in selective uptake of HDL and promotion of cholesterol efflux, we firstly confirmed that si-circABCA1 decreased the uptake of HDL and CHO labeled by Dil or BODIPY respectively, but ov-SCARB1 surprisingly had no effect on the HDL or CHO uptake, implying the existence of a saturation phenomenon in HDL or CHO uptake. Notably, ov-SCARB1 rescued the diminishment of HDL or CHO uptake induced by si-circABCA1 (F ig. 3F-I).

Intriguingly, contrary to our expectations, though the SCARB1 level was decreased, the CHO efflux was induced in si-circABCA1 group and ov-SCARB1 reversed such tendency. Furthermore, overexpression of SCARB1 failed to promote the CHO efflux in ccRCC cells, showing that circABCA1-SCARB1 axis regulated the CHO efflux by specific mechanisms associated with the low level of SCARB1 (Fig. 3J). In summary, si-circABCA1 could decrease the uptake of HDL by down-regulating SCARB1 thus lead to low CHO uptake and high CHO efflux, whose mechanism need to be further explored.

The ccRCC-suppressive effect of si-circABCA1 is mediated through SCARB1-dependent cholesterol modulation

Previous studies have demonstrated that the expression of SCARB1 is increased in ccRCC. Diminished CHO levels and SCARB1 knockdown suppress ccRCC cells proliferation and migration [18]. Given the observed CHO depletion upon si-circABCA1, we firstly investigated whether inhibition roles of si-circABCA1 on ccRCC cells are cholesterol-dependent. As shown in Fig. S3A-I, CCK-8, colony formation, EdU assays, Transwell as well as wound healing assays demonstrated that both si-circABCA1 and si-SCARB1 significantly impaired ccRCC cell proliferation and migration. Notably, exogenous cholesterol supplementation rescued these inhibitory effects of si-circABCA1 and si-SCARB1 on both proliferation and migration. These results indicate that knockdown of circABCA1 and SCARB1 inhibit ccRCC progression in a CHO-dependent manner.

Additionally, our data demonstrate that circABCA1 positively regulates SCARB1 expression. Functional assays including CCK-8, colony formation, EdU incorporation, Transwell migration, and wound healing consistently showed that circABCA1 knockdown significantly inhibited both proliferation and migration of ccRCC cells. Importantly, SCARB1 overexpression effectively rescued the tumor-suppressive phenotypes induced by si-circABCA1 (Fig. 4A-D, S4A-D).

Fig. 4.

Knockdown of circABCA1 suppresses migration and proliferation of ccRCC in a SCARB1-dependent manner. A-C Proliferation ability was examined by CCK-8, colony formation and EdU assays in the Caki-1 and 786-O cells after the overexpression of SCARB1 or interference of circABCA1. Scale bar = 50 μm. D Transwell assays were performed in Caki-1 and 786-O cells to detect the cell ability of migration and invasion after the overexpression of SCARB1 or interference of circABCA1. Scale bar = 50 μm. E The subcutaneous xenografts of si-NC, ov-NC, ov-SCARB1, si-circABCA1, si-circABCA1 + ovSCARB1 cells in implanted in nude mice. n = 5. F Left: The changes in tumor volume were recorded every 4 days in nude mice. Right: The weight of xenografts at indicated time. G, H Hepatic metastasis model of si-NC, ov-NC, ov-SCARB1, si-circABCA1, si-circABCA1 + ov-SCARB1 cells and H&E staining was established to evaluate the ability of cells to metastasize in vivo. Scale bar = 100 µm. NS means P ≥ 0.05, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001. All Data were presented as mean ± SD. Student’s t-test, n = 3

In vivo, as depicted in Fig. 4E and F, the growth rate and weight of subcutaneous tumor xenografts were suppressed by circABCA1 knockdown and restored by SCARB1 overexpression. Additionally, Ki67 staining indicated that the proliferative capacity was inhibited by circABCA1 knockdown and SCARB1 overexpression rescued the proliferation ability inhibited by si-circABCA1. Consistent with this, si-circABCA1 reduced the expression of SCARB1 (Fig. S4E). Furthermore, in models of hepatic metastasis, the knockdown of circABCA1 resulted in a decrease in both the number and size of metastatic nodules. Rescue experiments shown that the overexpression of SCARB1 restored the migration ability that was inhibited by si-circABCA1 (Fig. 4G and H, Fig. S4F and G). Collectively, these findings suggest that knockdown of circABCA1 inhibits the proliferation and migration of ccRCC both in vitro and in vivo through the regulation of SCARB1 (Fig. S13).

CircABCA1-IGF2BP3 interaction is required for regulating SCARB1 expression and HDL uptake

CircRNAs play pivotal roles typically by acting as miRNA sponges, interacting with proteins, or being translated into functional peptides [27]. As shown in Fig. S5A, circABCA1-flag was synthesized and inserted into the pLC5-ciR vector, and there was no significant blot compared to the positive control ABCA1-flag, indicating that circABCA1 failed to translate any peptides in ccRCC.

Next, to identify proteins that may bind to circABCA1, we performed MS2-RNA pull down assay and analyzed the product by LC–MS/MS (Fig. S5B and C). Among the potential interacting proteins of circABCA1, RBP-IGF2BP3 exhibited the highest abundance (unique peptide = 21, coverage = 32%) and score (Protein score = 584) [28] (Fig. 5A and Supplementary Table S8). Notably, AGO2, a core component of the RNA-induced silencing complex (RISC), was not detected in the analysis (Fig. S5D). These results indicated that circABCA1 mainly exerted its function through interacting with IGF2BP3 rather than acting as a miRNA sponge. Besides, RIP assay with anti-IGF2BP3 antibody corroborated the interaction between IGF2BP3 and circABCA1. Notably, ABCA1 mRNA was absent in anti-IGF2BP3 group (Fig. S5E), indicating that IGF2BP3 specifically binds to circABCA1, rather than its host gene transcript. Furthermore, the cytoplasmic colocalization of endogenous circABCA1 and IGF2BP3 was confirmed in Caki-1 (r = 0.836) or 786-O (r = 0.8591) cells (Fig. 5B).

Fig. 5.

CircABCA1 binds to IGF2BP3 to regulate SCARB1 mRNA stability and subsequently modulate cholesterol uptake. A RNA pull-down assays were used to confirm the directly binding relationship between IGF2BP3 and circABCA1. B RNA-FISH combined with IF to visualize the location of circABCA1 and IGF2BP3 in Caki-1 and 786-O cells. Scale bar = 10 µm. r means Pearson Correlation Coefficient between red and green fluorescence. C The protein level of SCARB1 were measured after the treatment of ov-IGF2BP3, si-circABCA1 and si-IGF2BP3 in Caki-1 and 786-O cells. D HDL uptake levels were assessed using Dil-HDL and analyzed by flow cytometry and fluorescent microscopy imaging after the treatment of ov-IGF2BP3, si-circABCA1 and si-IGF2BP3 in Caki-1 and 786-O cells. Scale bar = 20 µm. E Schematic representation of IGF2BP3 and circABCA1 binding sites on the SCARB1 mRNA 3'UTR. F Rative luciferase activity of luciferase reporter gene with 3'UTR WT or 3'UTR Mut in control and circABCA1-overexpression ccRCC cells. G Relative luciferase activity of luciferase reporter gene with 3'UTR WT or 3'UTR Mut in control and IGF2BP3-overexpression ccRCC cells. H RT-qPCR was employed to measure the half-life of SCARB1 mRNA in ccRCC cells treated with actinomycin D (ActD) for specified durations. I The binding relationship between circABCA1, SCARB1 mRNA and IGF2BP3 was measured via RIP assays. β-actin served as the negative controls. J Schematic diagrams illustrating the RNA-binding domains in IGF2BP3 and listing different IGF2BP3 truncation mutants. K RNA pull-down assays with targeting circABCA1 or SCARB1mRNA in cells transfected with plasmids encoding truncated IGF2BP3 mutants. L The half-life of SCARB1 mRNA in Caki-1 and 786-O cells which were transfected with plasmids encoding truncated IGF2BP3 mutants for indicated times. M RT-qPCR was used to detect the relative enrichment of SCARB1 mRNA of RIP product in ccRCC cells after the overexpression of circABCA1 with anti-IGF2BP3. N Diagram of the binding relationship of circABCA1/IGF2BP3/SCARB1 mRNA. NS means P ≥ 0.05, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001. All Data were presented as mean ± SD. Student’s t-test, n = 3

We subsequently investigated whether circABCA1 regulates SCARB1 in conjunction with IGF2BP3. Notably, IGF2BP3 expression showed significant positive correlation with SCARB1 mRNA levels in ccRCC tissues (n = 137, r = 0.7018, P < 0.001) (Fig. S5F). As shown in Fig. 5C and Fig. S5G, in ccRCC cells, protein level of SCARB1 was decreased in both knockdown of IGF2BP3 (si-IGF2BP3) and si-cirABCA1 groups. Notably, overexpression of IGF2BP3 (ov-IGF2BP3) restored the decreased level of SCARB1 caused by si-circABCA1. Moreover, si-circABCA1 or si-IGF2BP3 could result in decreased HDL and CHO uptake, while overexpression of IGF2BP3 neutralized the reduction of HDL and CHO uptake caused by si-circABCA1 in Caki-1 and 786-O cells (Fig. 5D, Fig. S5H-J). Above results suggest that circABCA1 combines with IGF2BP3 to regulate the expression of SCARB1 and subsequent HDL uptake.

The stability of SCARB1 mRNA is enhanced by forming a circABCA1-IGF2BP3-SCARB1 mRNA ternary complex

Next, we aimed to clarify the regulatory mechanisms of circABCA1 and IGF2BP2 on the expression of SCARB1. Initially, circABCA1 and IGF2BP3 exhibited limited impact on the expression of SCARB1 pre-mRNA (Fig. S6A). This observation implies that the modulation of SCARB1 expression by the circABCA1-IGF2BP3 axis is independent of its transcriptional regulation. Binding sites for circABCA1 and IGF2BP3 within the SCARB1 mRNA 3'UTR were predicted using IntaRNA 2.0 (rna.informatik.uni-freiburg.de/IntaRNA) and RBPmap (rbpmap.technion.ac.il) (Fig. 5E).

Existing studies have demonstrated that circRNA can directly bind to mRNA and enhance mRNA stability [29, 30]. Firstly, we identified a continuous binding motif between circABCA1 and SCARB1 mRNA 3'UTR (−31.9 kcal/mol). Next, the interaction between circABCA1 and SCARB1 mRNA was confirmed by RNA-RNA pull-down assays (Fig. S6B and C). We subsequently constructed luciferase reporter minigenes containing 3'UTR wild-type (3'UTR-WT) or 3'UTR mutant (3'UTR-Mut) of SCARB1 mRNA. For the SCARB1 mRNA 3'UTR-Mut luciferase reporter, the GAGCUGCUGCUGAAG motif, which is required for the interaction with circABCA1, was replaced. CircABCA1 overexpression significantly enhanced enzymatic activity in SCARB1 mRNA 3'UTR-WT transfectants, while showing minimal effects on SCARB1 mRNA 3'UTR-Mut constructs (Fig. 5F). These results confirm that circABCA1 directly binds to 3'UTR to enhance SCARB1 mRNA stability.

As a sequence rich in AU-rich elements and m6A recognition sites, the SCARB1 mRNA 3'UTR possesses the potential to be bound by the IGF2BP family and enhance its stability. With anti-IGF2BP3 antibody, SCARB1 mRNA was amplified by RIP-RT-qPCR assay and anti-IgG served as the negative control, indicating that IGF2BP3 could bind to SCARB1 mRNA 3'UTR (Fig. S6D). In luciferase reporter assays, IGF2BP3 overexpression enhanced the luciferase activity of SCARB1 mRNA 3'UTR-WT. However, in the SCARB1 mRNA 3'UTR-Mut (mutant deficient in circABCA1 binding) group, the ability of enhancing luciferase activity upon overexpressing IGF2BP3 was significantly reduced compared with 3'UTR-WT (Fig. 5G). These results demonstrate that IGF2BP3 could bind to SCARB1 mRNA 3'UTR, enhancing SCARB1 mRNA levels and the GAGCUGCUGCUGAAG site, which represents the binding site of circABCA1 in the 3'UTR, is also crucial for IGF2BP3 to enhance SCARB1 mRNA levels. Given the non-overlapping binding sites of circABCA1 and IGF2BP3 on the SCARB1 mRNA 3'UTR, they might simultaneously bind to and cooperatively enhance SCARB1 mRNA translation without competitive interference.

Furthermore, based on analysis of RBP-binding potential, we identified three non-overlapping IGF2BP3-binding candidate motifs (Motif1-Motif3) within the circABCA1 sequence. Notably, these motifs reside outside the region implicated in SCARB1 mRNA 3'UTR binding (Fig. S7A). Subsequent overexpression plasmids harboring mutations in these motifs (circABCA1-Motif1-mut to Motif3-mut) were transfected into ccRCC cells. RNA pulldown assays employing specific probes revealed that only Motif1-mut abrogated IGF2BP3 capture by circABCA1 (Fig. S7B and C). RIP assays further confirmed that anti-IGF2BP3 antibody immunoprecipitated circABCA1-WT but not circABCA1-Motif1-mut in ccRCC cells overexpressing Motif1-mut, demonstrating that Motif1 is essential for IGF2BP3 binding (Fig. S7D). Interestingly, luciferase reporter and Actinomycin D assays showed that overexpression of circABCA1-Motif1-mut (mutant deficient in IGF2BP3 binding) still promoted SCARB1 mRNA translation and stability, indicating that circABCA1 can upregulate SCARB1 independently of IGF2BP3 binding. However, this enhancement of SCARB1 mRNA translation and stability was attenuated compared to overexpression of wild-type circABCA1 (ov-circABCA1), suggesting that IGF2BP3 binding augments the function of circABCA1 in ccRCC cells (Fig. S7E-G). Moreover, as shown in Fig. S8A, we found that circABCA1 had no effect on IGF2BP3 abundance. Based on these findings, we propose the formation of a ternary IGF2BP3-circABCA1-SCARB1 complex.

Subsequently, we investigated the binding interactions among the three components within the circABCA1-IGF2BP3-SCARB1 mRNA ternary complex. Firstly, circABCA1 overexpression substantially extended SCARB1 mRNA stability, while IGF2BP3 knockdown partly abolished this effect, indicating that sufficient IGF2BP3 levels are important for circABCA1-mediated stabilization of SCARB1 mRNA (Fig. 5H). Secondly, with anti-IGF2BP3 antibody, both circABCA1 and SCARB1 mRNA were amplified by RIP-PCR, demonstrating the binding relationship of IGF2BP3-circABCA1-SCARB1 mRNA (Fig. 5I).

Given the RNA-binding KH domains of IGF2BP3 were required for binding to its RNA cargos, we engineered truncated IGF2BP3 mutants for each KH domain to elucidate the binding sites of IGF2BP3 with circABCA1 and SCARB1 mRNA (Fig. 5J). RNA pull-down assays revealed KH1 domain mediates circABCA1 binding, while KH4 domain specifically associates with SCARB1 mRNA in ccRCC cells (Fig. 5K). These domain-specific interactions were further validated by RIP assays (Fig. S8B and C). These demonstrate non-competitive binding between circABCA1 and SCARB1 mRNA on IGF2BP3, meaning that both can coexist on IGF2BP3 simultaneously.

Next, as shown in Fig. 5L, only full-length IGF2BP3, but not truncated mutants, effectively stabilized SCARB1 mRNA. Anti-IGF2BP3 RIP assays demonstrated circABCA1 overexpression increased IGF2BP3-bound circABCA1 levels (Fig. S8D), suggesting IGF2BP3 acts as a “sponge” for circABCA1. Additionally, as shown in Fig. 5M, the overexpression of circABCA1 resulted in higher enrichment level of SCARB1 mRNA, suggesting that high expression of circABCA1 in ccRCC cells facilitates the binding of more SCARB1 mRNA to IGF2BP3. Collectively, these data establish that in ccRCC cells, high-expressed circABCA1 induce more SCARB1 mRNA bind to IGF2BP3, enriching circABCA1 and SCARB1 mRNA on IGF2BP3 and spatially increasing the probability of their mutual binding, thus more SCARB1 mRNA 3'UTR is bound by circABCA1 and IGF2BP3, efficiently promoting an increase in SCARB1 levels (Fig. 5N).

Si-circABCA1 down-regulates SCARB1 leading to high-autophagy levels, which contribute to lipophagy, and result in elevated CHO efflux

As previously mentioned, we interestingly found that si-circABCA1 was able to prominently increased CHO efflux by downregulating SCARB1 (Fig. 3J). Further research is crucial to elucidate the mechanisms underlying this efflux. Although CHO efflux is primarily mediated by ABCA1 and ABCG1 [31], knockdown of circABCA1 did not affect the expression of these proteins (Fig. S9A). Several studies suggest that autophagy may facilitate CHO efflux [32, 33]. To investigate this, we treated ccRCC cells with rapamycin, an autophagic activator. As shown in Fig. S9B, rapamycin-treated ccRCC cells exhibited increased CHO efflux compared with controls, indicating that high autophagy can induce CHO efflux in ccRCC. Additionally, the stimulatory effect of si-circABCA1 on CHO efflux was inhibited by chloroquine (Chlo), an autophagy inhibitor (Fig. 6A). Furthermore, Chlo partially reversed the reduction in T-CHO levels caused by si-circABCA1, increasing them to 50% of negative control (NC) levels. Conversely, overexpression of SCARB1 fully rescued the decrease in T-CHO levels caused by si-circABCA1, but this rescue was reduced to 50% of NC levels after rapamycin treatment (Fig. 6B). These findings suggest that autophagy-induced CHO efflux might contribute to the low T-CHO levels observed in si-circABCA1 treated cells.

Fig. 6.

Si-circABCA1 down-regulates SCARB1 leading to high-autophagy levels, which contribute to lipophagy, resulting in elevated CHO efflux. A Rate of CHO efflux was measured after the treatment with or without si-circABCA1 and chloroquine (Chlo). B T-CHO level was detected in Caki-1 or 786-O cells after the treatment of overexpression of SCARB1, interference of circABCA1, Chlo and Rapamycin. C Representative images of the co-localization between LC3B (red) and lipid droplets (green) in ccRCC cells after the overexpression of SCARB1 or interference of circABCA1. nuclei were stained with DAPI (blue). Scale bar = 10 µm. Pearson Correlation Coefficient were calculated to quantified the co-localization between LC3B and lipid droplets in ccRCC cells. D Transmission electron microscopy (TEM) of NC and si-circABCA1 in ccRCC cells were shown. Red arrows target autolysosome, yellow stars target lipid droplets, blue arrows target mitochondrion. N means nucleus. Scale bar = 1 μm. E, F The content of Free cholesterol (FC), Cholesteryl ester (CE) and T-CHO in ccRCC cells were measured for indicated times after the overexpression of SCARB1 or interference of circABCA1. NS means P ≥ 0.05, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001. All Data were presented as mean ± SD. Student’s t-test, n = 3

The hydrolysis and liberation of CE from lipid droplets to generate free cholesterol (FC) are crucial for CHO efflux [19, 20]. As shown in Fig. S9C, si-circABCA1 significantly disrupted lipid droplets in ccRCC cells, reducing their number. However, co-treatment with si-circABCA1 and chloroquine partially rescued the number of lipid droplets, while ov-SCARB1 fully restored their number. Moreover, phagocytosis of lipid droplets by lysosomes was observed (Fig. S9D), suggesting that autophagy plays a role in modulating lipid droplets, a process known as lipophagy. Lipophagy involves the recruitment of autophagosomes by LC3-bound lipid droplets, ultimately leading to their degradation [20]. As shown in Fig. 6C, the percentage of lipid droplets co-localized with LC3B increased in si-circABCA1-treated groups. Rescue experiments demonstrated that the percentage of co-localization in si-circABCA1-ov-SCARB1-treated cells was restored compared with si-circABCA1-only groups. Furthermore, electron microscopy confirmed that lipid droplets were engulfed by autolysosomes in si-circABCA1-treated cells (Fig. 6D).

Next, we assessed FC, CE, and T-CHO levels in ccRCC cells treated with different conditions at various time points. As shown in Fig. 6E and F, while CE and T-CHO levels consistently decreased, FC levels increased within the first 12 h in si-circABCA1-treated groups. Interestingly, cotreatment with si-circABCA1 and ov-SCARB1 restored FC levels to a statistically insignificant level compared with the si-NC + ov-NC control. In summary, si-circABCA1-induced downregulation of SCARB1 promotes lipophagy, leading to increased FC levels and subsequently high CHO efflux, which contributes to decreased T-CHO levels in ccRCC cells.

CircABCA1 regulates autophagy through IGF1R/PI3K/AKT/mTOR cascade in a CHO-dependent manner involving lipid rafts

Given that si-circABCA1 promotes lipophagy, a type of autophagy, we assessed the activity of autophagy by measuring the ratio of LC3B/LC3A and P62 levels. As illustrated in Fig. 7A, knockdown of circABCA1 increased the LC3B/LC3A ratio and decreased P62 levels, while overexpression of SCARB1 rescued these changes induced by si-circABCA1. These findings implied that the elevated autophagy levels caused by si-circABCA1 are dependent on SCARB1.

Fig. 7.

CircABCA1 regulates autophagy through IGF1R/PI3K/AKT/mTOR cascade in a CHO-dependent manner involving lipid rafts. A Protein level of LC3B/LC3A, P62 was detected by western blot in Caki-1 or 786-O cells after the overexpression of SCARB1 or interference of circABCA1. B Protein level of LC3B/LC3A, P62 was detected by western blot in Caki-1 or 786-O cells after the overexpression of circABCA1 or MβCD. C, D The contents of lipid rafts in Caki-1 or 786-O cells were measured by CT-B staining after the overexpression of SCARB1 or interference of circABCA1. nuclei were stained with DAPI (blue). Scale bar = 50 μm, and were quantified by flow cytometry. E Western blot analysis showing the expression of p-IGF1R, p-PI3K, p-AKT, p-mTOR in Caki-1 or 786-O cells after the overexpression of SCARB1 or interference of circABCA1. NS means P ≥ 0.05, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001. All Data were presented as mean ± SD. Student’s t-test, n = 3

Considering SCARB1's role as a CHO regulator, the autophagy-regulating function of the si-circABCA1-SCARB1 axis might be CHO-dependent. To further investigate this, we downregulated CHO levels using MβCD. MβCD induced high autophagy levels in ccRCC cells, and overexpression of circABCA1 reversed this effect (Fig. 7B). Additionally, exogenous CHO reduced the autophagy level induced by si-circABCA1, indicating that circABCA1 regulates autophagy in accordance with CHO levels (Fig. S10A and B). Next, we examined the role of SCARB1 in autophagy in ccRCC. Treatment with si-SCARB1 induced high autophagy level in ccRCC cells, and exogenous CHO rescued this effect, suggesting that si-SCARB1 promotes autophagy by decreasing CHO levels (Fig. S10C and D). Taken together, both circABCA1 and SCARB1 regulate autophagy through CHO modulation, and down-regulated CHO levels lead to higher autophagy in ccRCC cells.

Notably, MβCD also disrupts lipid rafts by depleting CHO, thereby increasing autophagy levels [34, 35]. Overexpression of circABCA1 rescued autophagy induced by MβCD through increasing CHO levels in ccRCC cells. Therefore, we hypothesized that si-circABCA1 may contribute to high autophagy levels through reducing lipid rafts in ccRCC. As shown in Fig. S10E and S10F, MβCD led to a decline in both the integrity and content of lipid rafts, while overexpression of circABCA1 significantly restored these levels, indicating that high circABCA1 protects lipid rafts from being disrupted by low CHO levels. Furthermore, knockdown of circABCA1 significantly decreased the integrity and content of lipid raft, and rescue experiments showed that overexpression of SCARB1 counteracted this trend, suggesting that si-circABCA1-SCARB1 disrupts lipid rafts by downregulating CHO levels (Fig. 7C and D).

The phosphorylation of IGF1R and PI3K is strongly associated with lipid rafts [36]. Our previous studies have shown that ccRCC tissue exhibits high levels of PI3K-AKT- mTOR phosphorylation and low autophagy levels [37–39]. Additionally, down-regulated SCARB1 in ccRCC significantly reduce PI3K/AKT phosphorylation [18]. Furthermore, we found that ccRCC cells treated with si-circABCA1 exhibited increased sensitivity to Temsirolimus (Fig. S11A). Temsirolimus is an mTOR inhibitor and first-line treatment drug for ccRCC [40]. Thus, we investigated whether the phosphorylation of IGF1R/PI3K/AKT/mTOR is the underlying cause of the elevated autophagy levels induced by si-circABCA1-SCARB1. As illustrated in Fig. S11B, MβCD significantly inhibited the phosphorylation of IGF1R/PI3K/AKT/mTOR, and this effect was reversed by overexpression of circABCA1, indicating that low lipid raft levels lead to reduced phosphorylation of IGF1R/PI3K/AKT/mTOR, and high circABCA1 promotes phosphorylation by maintaining lipid rafts. Moreover, we examined the role of SCARB1 on the phosphorylation of IGF1R/PI3K/AKT/mTOR. Si-SCARB1 decreased IGF1R/PI3K/AKT/mTOR phosphorylation, which was rescued by CHO, suggesting that SCARB1 mediates the regulation of IGF1R/PI3K/AKT/mTOR phosphorylation by CHO in ccRCC cells (Fig. S11C and D). Moreover, si-circABCA1 significantly decreased IGF1R/PI3K/AKT/mTOR phosphorylation, and rescue experiments showed that overexpression of SCARB1 counteracted this trend (Fig. 7E). In conclusion, our findings suggest that the destruction of lipid rafts in ccRCC cells, which suppressed the phosphorylation of IGF1R/PI3K/AKT/mTOR, might be the underlying mechanism of si-circABCA1-SCARB1-induced lipophagy.

Si-circABCA1 suppressed M2 macrophage polarization and exerted its anti-tumor effect in a SCARB1-dependent manner

CHO have also been reported to play an essential role in macrophages polarization [25, 26]. As shown in Fig. S12A-B, we found that the exogenous CHO resulted in a significant upregulation of M1 macrophage-associated genes expression (IFNG, TNFα, CD86, IL6) and a profound diminution of M2 macrophage-associated genes expression (IL4, CD206, CD163, IL10, TGFB). Next, the addition of exogenous CHO to the culture medium (CM) of IL-4-induced M2 macrophages resulted in a decrease in the expression of M2 macrophage-associated genes (Fig. S12C and S12D). These results suggesting that CHO may be instrumental in remodeling macrophages polarization by promoting M1-like polarization but suppressing M2-like polarization. Notably, IHC staining shown that T1 stage ccRCC exhibited extremely low levels of CD206, a marker of M2-like polarization, compared with T3 stage (Fig. 8A), indicating that reducing M2 macrophage infiltration may improve the prognosis of ccRCC.

Fig. 8.

Si-circABCA1 suppressed M2 macrophage polarization and exerted its anti-tumor effect in a SCARB1-dependent manner. A The expression of CD206 was detected by IHC in three T1 stage patients and three T3 stage patients. Scale bar = 100 µm. B, C Schematic representation of the co-culture system of M2 macrophages and ccRCC cells with Transwell assay. D, E Images (D) and quantification (E) of BODIPY staining in macrophages cultured for 24 h in different treated ccRCC culture medium (CM). Scale bar = 50 µm. F RT-qPCR analysis of macrophages polarization-associated gene expression in macrophages cultured for 24 h in different treated ccRCC CM. G Cell Apoptosis rate of normal ccRCC cells cultured for 24 h in different treated ccRCC CM and were analyzed by flow cytometry. H ELISA analysis was performed to measure the concentration of TNF-α in different treated ccRCC CM. I The expression of CD206 was detected by IHC in subcutaneous tumor xenografts. Scale bar = 100 µm. NS means P ≥ 0.05, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001. All Data were presented as mean ± SD. Student’s t-test, n = 3

Given that si-circABCA1 significantly enhances CHO efflux, we observed a notable increase in CHO content in the supernatant CM of ccRCC cells following si-circABCA1 treatment and a decrease following ov-circABCA1, indicating that high levels of circABCA1 enable ccRCC to competitively uptake CHO from the microenvironment (Fig. S12E). Next, we established a co-culture system of ccRCC cells and M2-like macrophages using a Transwell assay (Fig. 8B and C). As shown in (Fig. S12F-I), co-culturing si-circABCA1 ccRCC cells and M2-like macrophages resulted in an increase in M1 markers and a decrease in M2 markers, confirming that si-circABCA1 can remodel M2 macrophages in the tumor microenvironment toward M1-like polarization.

Considering that circABCA1 and SCARB1 jointly regulate CHO efflux, our co-culture system revealed that si-circABCA1-treated ccRCC cells co-cultured with M2 macrophages increased the neutral lipid content of the M2-like macrophages, indicating that the uptake of CHO was elevated in macrophages. Furthermore, ov-SCARB1 in ccRCC cells mitigated the increase in neutral lipid levels observed in M2 macrophages caused by si-circABCA1-treated ccRCC cells (Fig. 8D and E). This CHO uptake induced the expression of M1 macrophage-associated genes, while inhibited M2-associated markers. Next, ov-SCARB1 exerted a restorative effect, reversing the trends established by si-circABCA1 (Fig. 8F).

To investigate the plausible feedback effect of circABCA1-SCARB1-regulated CHO efflux-induced macrophage polarization on ccRCC proliferation, we collected CM from the co-culture system of differently treated ccRCC cells after 24 h, then introduced CM to normal ccRCC cells (Fig. 8C). Our results showed that CM collected from si-circABCA1-treated groups significantly inhibited proliferation and promoted apoptosis of ccRCC cells. Conversely, CM generated from si-circABCA1-treated groups with overexpression of SCARB1 reversed this trend (Fig. S12J, Fig. 8G). ELISA analysis demonstrated a significant increase in TNF-α production by M1-like macrophages co-cultured with the si-circABCA1 CM, indicating an increased proportion of tumor-suppressive M1 macrophages. Moreover, by introducing CM form si-circABCA1 + ov-SCARB1 group, the high TNF-α level initiated by si-circABCA1 group were significantly blocked (Fig. 8H). Furthermore, subcutaneous tumor xenografts showed that tumors from si-circABCA1 treatment displayed decreased M2-like polarization based on the detection of CD206 protein, which was reversed by overexpression of SCARB1 (Fig. 8I). Together, above data suggest that circABCA1 promote M2 macrophage polarization in a SCARB1-dependent manner (Fig. S13).

Discussion

The clear cell phenotype observed in ccRCC is closely associated with the intracellular accumulation of lipids, which is a prominent characteristic and correlates with malignancy, prognosis, and resistance to antiangiogenic drugs [3, 41]. Romain et al. showed that maintaining cholesterol homeostasis inhibits apoptosis in ccRCC [42]. Liu et al. indicated that high lipid deposition facilitates ccRCC metastasis [43]. Interfering with lipid deposition has the potential to suppress ccRCC proliferation and migration. For instance, inhibiting ccRCC lipid deposition restrains tumor growth and enhances the anticancer efficacy of everolimus [44]. Similarly, Zhang et al. found that reduced cholesterol levels inhibit cell migration [45]. Here, we found that downregulation of cholesterol significantly inhibited ccRCC proliferation, migration, and invasion.

In the current study, we sought to characterize circRNAs involved in the regulation of lipid metabolism in ccRCC. Our findings revealed that upregulated circABCA1 in ccRCC elevated the CHO levels by increasing the CHO uptake and inhibiting efflux. Many other circRNAs have also been identified to participate in the regulation of lipid metabolism in various cancers. For example, circACC1 modulates both fatty acid β-oxidation and glycolysis, resulting in profound changes in cellular lipid storage in colorectal cancers [8]. Hsa_circ_0086414 contributes to the elimination of lipid deposition in ccRCC [7]. CircLDLR in colorectal cancer increases the cholesterol level by the mediation of SOAT1 [23]. Though circRNAs have been proved to be crucial regulators in lipid metabolism of cancers, there is a lack of research investigating roles in ccRCC, the discovery of circABCA1 might further emphasize the significant role of circRNAs in lipid metabolism in ccRCC.

Our study revealed that circABCA1 contributed to an increased accumulation of cholesterol through the regulation to SCARB1 expression in ccRCC. Studies have demonstrated that inhibition of SCARB1 is sufficient to induce cell-cycle arrest, apoptosis, increased intracellular reactive oxygen species levels, and decreased PI3K/AKT signaling in ccRCC cells [18, 42, 46]. We here found that circABCA1 positively regulated the SCARB1 protein level. CircABCA1 maintained migration and proliferation properties of ccRCC by relying on SCARB1. As an exogenous cholesterol-dependent cancer, previous studies confirmed that SCARB1 is essential for the transport of HDL in ccRCC [18]. Herein, we discovered, for the first time, that decreasing the expression of SCARB1 in ccRCC enhanced cholesterol efflux, a process stimulated by the reduction of cholesterol uptake, suggesting that SCARB1 plays a complex role in regulating lipid metabolism in ccRCC by modulating cholesterol levels.

Lipid droplets are dynamic cytoplasmic organelles that consist of a phospholipid monolayer enclosing a neutral lipid core, primarily composed of cholesterol esters and triacylglycerols [32, 47, 48]. Esterified cholesterol are one of the most abundant lipids stored in lipid droplets of ccRCC [49]. The breakdown of lipid droplets results in the hydrolysis and liberation of cholesterol esters to generate free cholesterol, which can promote cholesterol efflux [20]. Here, we reported the lipophagy, a type of selective autophagy that targets lipid droplets and maintains the homeostasis of lipid droplets [50, 51], was the cause of the efflux induced by the knockdown of circABCA1 in a SCARB1 depended manner. We observed that lipid droplets were delivered to lysosomes via autophagy when silencing circABCA1 in ccRCC, where CE was hydrolyzed to generate free cholesterol for efflux. The decreased number of lipid droplets and disruption of the morphology of the remaining droplets resulted in a peak in free cholesterol concentration observed at approximately 12 h, leading to a high CHO efflux. We also discovered that lipid rafts were disrupted by the knockdown of circABCA1 in a SCARB1 dependent manner in ccRCC, and confirmed that IGF1R acts as an upstream activator of the PI3K/Akt/mTOR signaling axis, which is dependent on the functionality of lipid raft and is inhibited by the knockdown of circABCA1, ultimately inducing autophagy. Similarly, Zhang et al. found that Celastrol could promote cholesterol efflux by inducing lipophagy in ccRCC [45]. Our findings further confirm the pivotal role of lipid droplet homeostasis in ccRCC cells, indicating that circABCA1, which regulates the number and morphology of lipid droplets, may represent a promising target for the lipid droplet homeostasis of ccRCC.

Previous research has demonstrated that cancer cells can interact with and influence resident or infiltrated macrophages, leading to their transformation into pro-tumorigenic phenotypes via CHO [52]. Tao et al. have shown that high cholesterol uptake by IDHwt glioma leads to the depletion of membrane cholesterol in microglial cells, ultimately resulting in M2-like polarization [53]. Goossens et al. and Hoppstädter et al. have indicated that ovarian and lung cancer cells induce M2-like polarization in TAMs by sequestering their cholesterol, which activates the IL4/PI3K/STAT6 signaling axis in these cells [54, 55]. Our findings indicate that the elevated levels of CHO released from ccRCC cells into the microenvironment, due to si-circABCA1-induced downregulation of SCARB1, significantly inhibit the polarization of M2 macrophage. Consequently, it induces the expression of pro-inflammatory cytokines, particularly TNF-α, thereby impairing the proliferation of ccRCC cells.

Recent studies have increasingly documented the role of circRNAs in regulating gene expression through their interactions with RNA-binding proteins (RBPs), thereby mediating diverse biological functions. Here, we pinpointed IGF2BP3 as a crucial interactor of circABCA1, whose disruption abrogated the up-regulatory influence of circABCA1 on SCARB1 expression. Furthermore, the circABCA1/IGF2BP3 complex enhances the stability of SCARB1 mRNA by binding to 3'UTR of SCARB1 mRNA, collaboratively modulating cholesterol uptake phenotypes. Importantly, we have uncovered distinct IGF2BP3 KH domains that engage independently with circABCA1 and SCARB1 mRNA, thereby preventing competition between these two entities for IGF2BP3 binding. This implies that circABCA1 binding may induce conformational changes in IGF2BP3, facilitating its subsequent interactions with SCARB1 mRNA. Similarly, CircITGB6 enhances the stability of PDPN mRNA, a gene that promotes epithelial-mesenchymal transition, through direct interaction with IGF2BP3 [56]. CircSIPA1L3 stabilizes SLC16A1 and RAB11A mRNAs by enhancing their binding to IGF2BP3, thereby promoting lactate export and glucose uptake in [6]. CircARID1A functions as a molecular scaffold that enhances IGF2BP3-SLC7A5 mRNA binding, thereby stabilizing SLC7A5 transcripts [57]. Collectively, the interplay between circRNA and IGF2BP3 may constitute a fundamental regulatory mechanism governing the expression of various mRNAs.

Although our results establish that the circABCA1-IGF2BP3 complex upregulates SCARB1 expression through mRNA stabilization, IGF2BP3 has dual regulatory roles in both transcript stability and translational control, particularly through its m6A reader function in cooperation with METTL family proteins [58]. Furthermore, IGF2BP3 has been shown to promote ribosome biogenesis and translation initiation [59]. These collective findings suggest that circABCA1 may serve as a molecular scaffold that enhances IGF2BP3-mediated translational activation of SCARB1, potentially through facilitating ribosome assembly or improving mRNA accessibility. Further mechanistic investigations are warranted to determine whethercircABCA1-IGF2BP3 axis modulates SCARB1 translational efficiency.

The impact of cholesterol on tumor cells exhibits dynamic and balanced characteristics. Though knockdown of circABCA1 inhibited the proliferation and migration of ccRCC cells, we interestingly discovered that circABCA1 overexpression failed to significantly enhance these phenotypes (data not shown). This was consistent with the effect of circABCA1 overexpression on cholesterol levels in ccRCC cells, as shown in Fig. 2L. This might occur because baseline hyper-cholesterol in ccRCC cells likely saturated cholesterol storage capacity, preventing further accumulation upon circABCA1 overexpression. Crucially, we demonstrated that circABCA1-mediated regulation of ccRCC progression was cholesterol-dependent (Fig. S3). Thus, the absence of overt tumor-promoting effects under overexpression conditions may be attributed to cholesterol saturation in the experimental model.

Conclusions

Our study demonstrates that elevated circABCA1 expression in ccRCC stabilizes SCARB1 mRNA through formation of a circABCA1-IGF2BP3-SCARB1 mRNA ternary complex, thereby enhancing cholesterol uptake and preserving lipid droplets against degradation of lipophagy via maintenance of lipid raft integrity and activating the IGF1R/PI3K/AKT/mTOR signaling pathway. The resultant elevated cholesterol uptake coupled with reduced efflux in ccRCC cells establishes a cholesterol-depleted tumor microenvironment that drives M2 macrophage polarization. Functional analyses confirmed the critical role of the circABCA1-SCARB1 axis in accelerating ccRCC proliferation and metastatic progression. These findings might deepen our understanding of the molecular mechanisms underlying cholesterol homeostasis in ccRCC (Fig. S13).

Materials and methods

Patients and tissue samples

158 paired human clear cell renal cell carcinoma (ccRCC) tissues and adjacent nontumor tissues were collected, with a strategic allocation of 4 paired tissues for the screening phase and 154 paired tissues for the subsequent validation phase. These samples were sourced from the Southwest Hospital of the Army Medical University (Chongqing, China, 2019–2024). The diagnosis of ccRCC was confirmed through histopathological examination of each tissue specimen. This study adhered to the highest ethical standards, as evidenced by its approval granted by the Ethics Committee of the Southwest Hospital of Army Medical University (approval number KY2020121). Prior to surgical intervention, we obtained written informed consent from each participant, upholding their autonomy and rights. All the tissues were promptly immersed in RNAlater (Thermo Fisher, USA) upon collection. Clinical characteristics are listed in Supplementary Table S1.

Chemicals

DMSO was purchased from Solarbio (D8370, 362 Beijing, China). Chloroquine (HY-17589A), Methyl-β-cyclodextrin (HY-101461), Temsirolimus (HY-50910), cholesterol (HY-N0322) and IL4 (HY-P70445), Phorbol 12-myristate 13-acetate (PMA) (HY-18739) were obtained from Med Chem Express (New Jersey, USA). Rapamycin (T1537) were purchased from TargetMol (Shanghai, China).

Microarray analysis

We employed microarray to identify and characterize differentially expressed transcriptome within four matched pairs of ccRCC tissues and their adjacent noncancerous counterparts (Shanghai Genomics Corporation, China) as we previous reported [37]. Data analysis adhered to Agilent's protocols at Sinotech Genomics, applying a fold change cutoff of two. Clinical characteristics are listed in Supplementary Table S2.

Untargeted metabolomics

Lipids were extracted from thawed tissue samples using ice-cold 80% methanol (1:10 w/v), followed by centrifugation (15 min, 4 °C), supernatant collection, vacuum drying, and reconstitution in 80% methanol. Pooled quality control (QC) samples were prepared by combining equal aliquots of all test samples and were analyzed at regular intervals throughout the entire analytical sequence. Pearson's correlation coefficients (PCC) were calculated using the abundance values of QC samples and visualized in a correlation heatmap, where higher correlation values were represented by more intense red coloration.

Chromatographic separation was performed on an ACQUITY UPLC T3 column (40 °C) with a binary gradient of 5 mM ammonium acetate/acetic acid in water (A) and acetonitrile (B) at 0.3 mL/min: 0–0.8 min (2% B), 0.8–2.8 min (2–70% B), 2.8–5.6 min (70–90% B), 5.6–8.0 min (90–100% B).

Lipids were detected using a TripleTOF 6600 mass spectrometer (SCIEX, Framingham, MA, USA) in positive/negative ionization modes (± 4500-5000 V ionspray voltage, 30 PSI curtain gas, 60 PSI source gases, 500 °C interface). Full-scan MS data (60–1200 m/z) were acquired in IDA mode with dynamic exclusion (4 s).

Raw data were converted to mzXML format and processed via XCMS with CAMERA and metaX toolboxes for peak alignment, retention time correction, and lipid annotation using LIPID MAPS/HMDB databases (mass tolerance ≤ 10 ppm), validated by in-house spectral libraries. Differential lipids were identified using R-based statistical pipelines (P < 0.05, fold change > 1.2, variable importance in projection > 1.2) and pathway enrichment analyzed via hypergeometric tests and GSEA (|NES|> 1, FDR < 0.25). Method reproducibility was ensured through integrated QC checks, including isotopic validation and system suitability criteria specific to lipidomic profiling. Clinical characteristics, quality control and analysis results are listed in Supplementary Table S3.

Cell lines and cell culture

Caki-1, 786-O, HK-2 and THP-1 were procured from Meisen CTCC (Zhejiang, China) and verified through STR profiling. All cell lines were cultured as we previous reported [37]. 100 ng/mL of PMA was introduced into the culture medium of THP-1 cells to induce their differentiation into M0 macrophages.

RNA/gDNA extraction and RT-qPCR assays

RNA/gDNA extraction and RT-qPCR assays were performed as we previous reported [37] and β-actin as the internal control. All primers were custom-synthesized by Tsingke (Beijing, China) and are detailed in Supplementary Table S4.

Plasmids, siRNAs and cell transfection

The siRNA targeting circABCA1 (si-circABCA1) and IGF2BP3 (si-IGF2BP3) were produced by GenePharma, located in Shanghai, China. The IRES sequences were synthesized, cloned, and inserted into the pCMV vector supplied by Geneseed. Similarly, full-length IGF2BP3 constructs with various domains (FL-3 × flag, KH-4–3 × flag, KH-2–3 × flag, KH1-3 × flag, and RRMs-3 × lag) were synthesized, cloned, and inserted into the same pCMV vector. The 3 × flag sequence was positioned after the start codon ATG in the coding sequence predicted by Unibio (Chongqing, China). Full-length circABCA1 were inserted into the pLC5-ciR vector (RRID: Addgene_169139). The pCDH-SCARB1 was synthesized by Tsingke. The specific sequences of the siRNAs are provided in Supplementary Table S5.

RNA-pulldown assay

Biotinylated RNA probes targeting specific sequences were purchased from GENESEED (Guangzhou, China). Prior to hybridization, ​RNA folding was induced​ by denaturing probes at 95 °C for 2 min, followed by immediate snap-cooling on ice for 3 min and refolding in Structure Buffer (10 mM Tris–HCl pH 7.0, 100 mM KCl, 10 mM MgCl₂) at 25 °C for 30 min.

For the protein-RNA pulldown, cell lysates were prepared by lysing cultured cells in RIPA Lysis Buffer (Beyotime, China). Hybridization was performed by incubating biotinylated RNA probes (200 pmol) with clarified cell lysates in Binding Buffer (20 mM Tris–HCl pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 0.1% Tween-20, 1 U/μl RNase inhibitor) at 37 °C for 2 h with gentle rotation. RNA-bound complexes were captured using streptavidin-conjugated magnetic beads (Invitrogen, USA), which were pre-blocked with 1 μg/μl yeast tRNA (Sigma-Aldrich, Germany) and washed twice with Wash Buffer A (20 mM Tris–HCl pH 7.5, 500 mM NaCl, 0.1% SDS). To minimize nonspecific binding, bead-bound complexes were sequentially washed under stringent conditions: twice with Wash Buffer A, twice with Wash Buffer B (250 mM NaCl, 0.05% SDS), and once with Wash Buffer C (50 mM NaCl). Bound proteins were eluted in 2 × SDS Loading Buffer (Thermo Fisher Scientific, USA) at 95 °C for 10 min and analyzed by Western blot.

For the RNA-RNA pulldown, to eliminate potential protein contamination, protein-free total RNA was extracted from ccRCC cells follow the protocol of MagMAX™ mirVana™ Total RNA Isolation Kit (Thermo Fisher, USA). Hybridization was performed by incubating biotinylated circABCA1 probes (200 pmol) with protein-free total RNA (20 µg) in Binding Buffer at 37 °C for 2 h with gentle rotation. RNA-bound complexes were captured by streptavidin-conjugated magnetic beads pre-washed with Wash Buffer A and blocked with yeast tRNA (1 µg/µl). After hybridization, bead-probe complexes were washed sequentially under stringent conditions: twice with Wash Buffer A, twice with Wash Buffer B, and once with Wash Buffer C. Bound RNAs were eluted in 100 µl Elution Buffer (10 mM Tris–HCl pH 7.0, 1 mM EDTA, 0.1% SDS) at 65 °C for 10 min, followed by ethanol precipitation and quantified by RT-qPCR.

Negative controls included parallel assays with scramble-sequence probes to confirm binding specificity. The RNA pulldown probe sequences are provided in Supplementary Table S4.

Western blot analysis

Western blot assays were performed as we previous reported [37]. Antibodies we used were listed in Supplementary Table S6. Unedited blot and gel images were in supplementary file.

Actinomycin D (ActD) assay

ccRCC cells were seeded in 12-well plates a day before being treated with 10 μg/mL of actinomycin D (Genview, China) to inhibit RNA transcription. Subsequently, RNA was isolated and RT-qPCR was conducted, using β-actin as the reference gene for normalization. Relative expression was quantified by the 2^−ΔΔCt method.

RNase R treatment

RNase R treatment was conducted to detect the stable circular structure of circABCA1 in ccRCC cells, as previously reported in detail by our previous study [37].

Subcellular fractionation assay

Subcellular fractionation assays were employed to determine the subcellular localization of circRNA, as previously reported by us [37].

Cell counting kit-8 assay and 5‑ethynyl‑2’‑deoxyuridine (EdU) assay

Caki-1 and 786-O cells were harvested and seeded into 96-well plates at a density of 3000 cells/well. After cell attachment, 90 µL of complete medium was supplemented with 10 µL of CCK-8 solution (TargetMol, USA) in each well at time points 0 h, 24 h, 48 h, 72 h, and 96 h. The absorbance at 450 nm was quantified using a Multiskan SkyHigh microplate reader (Thermo Fisher, USA). EdU assays were performed based on the instruction of The Click-iT EdU assay Kit (Bioscience, China).

RNA fluorescence in situ hybridization (FISH) assay

Cy3-labeled circABCA1 probes and the FISH kit were synthesized by GenePharma (Shanghai, China) and performed as previously reported by us [37].

Intracellular total-cholesterol (T-CHO) quantification

Intracellular T-CHO was quantified in ccRCC cells (5 × 10⁵) after 24 h treatment and lysed in ice-cold RIPA buffer (Beyotime, China) with phosphatase inhibitors (Roche, Germany). Supernatants were analyzed for protein concentration using BCA (Thermo Fisher, USA) and cholesterol levels measured with the TTCA Kit (Applygen, China), following manufacturers'protocols. The T-CHO in the culture medium was measured following manufacturers'protocols.

RNA sequencing

To identify the downstream targets and pathways controlled by circABCA1, RNA was isolated from 786-O cells that had been transfected with either circABCA1 siRNA or a control siRNA/vector. Following this, RNA-seq and KEGG pathway analysis were performed by Shanghai Genomics Corp. (Shanghai, China), involving thorough library preparation and detailed computational analysis. Results are listed in Supplementary Table S7.

Cholesterol uptake assay for imaging or flow cytometry

The cholesterol uptake assay for ccRCC cells was conducted using DiI-HDL (Solarbio, China) and BODIPY-cholesterol (TargetMol, China). Cells (3 × 10⁴) were seeded in u-Slide 8-well plates, incubated overnight hen starved in serum-free DMEM for 6 h. DiI-HDL (25 μg/mL) or BODIPY-cholesterol (1 μM) was added, and after 6 h incubation, Hoechst 33,342-stained, and imaged via confocal microscopy (Zeiss LSM 880). For flow cytometry, cells were (3 × 10⁴) seeded in 6-well plates, resuspended with PBS, and analyzed (Cytoflex LX, Beckman). All experiments were replicated three times.

Cholesterol efflux assay

Caki-1 and 786-O cells were seeded in 96-well plates at 100 μL media/well and incubated at 37 °C with 5% CO2 for 2 h. Once attached, 100 μL of labeling reagent-equilibration buffer mix (Abcam, UK) was added per well. Following overnight incubation, cells were washed, and cholesterol acceptors applied to remove excess label. Fluorescence intensities of media and cell lysates (Ex/Em = 482/515 nm) were quantified. Cholesterol efflux was calculated as the percentage of media fluorescence intensity over the sum of cell lysate and media fluorescence intensities, multiplied by 100.

MS2 RNA-pulldown assay

A circRNA pulldown assay utilizing MS2-capturing protein was conducted to identify RNA-binding proteins interacting with circABCA1 (Geneseed, China). 786-O cells were transfected with circABCA1-MS2 and MS2-CP-GFP vectors. MS2-CP-GFP antibodies bound to the capture protein, enriching for circRNA-associated complexes. Subsequently, circABCA1 pulldown samples were subjected to mass spectrometry analysis (Thermo Fisher, USA) for comprehensive protein profiling. Briefly, for in-gel tryptic digestion, gel pieces were destained in 50 mM NH₄HCO₃/50% acetonitrile, reduced with 10 mM dithiothreitol (56 °C, 60 min), alkylated with 55 mM iodoacetamide, and digested with 10 ng/μl trypsin. Peptides were extracted with 50% acetonitrile/5% formic acid and 100% acetonitrile, dried, and resuspended in 2% acetonitrile/0.1% formic acid. LC–MS/MS analysis was performed using an EASY-nLC 1200 UPLC system with a homemade C18 column (25 cm × 100 μm), employing a 30 min gradient (6–80% solvent B; 0.1% formic acid/90% acetonitrile) at 550 nl/min. Peptides were ionized via nano-electrospray (2100 V) and analyzed on an Orbitrap Exploris 480 mass spectrometer in data-dependent acquisition mode (full MS: 60,000 resolutions, 350–1800 m/z; MS/MS: 15,000 resolution, NCE 28%, 20 s dynamic exclusion). MS/MS data were processed using PD software (v2.4) against the UniProt human database (20,429 entries), with carbamidomethylation as a fixed modification and acetylation/oxidation as variable modifications. Peptide identifications required a score > 20 and high confidence, with mass tolerances of 10 ppm (precursor) and 0.02 Da (fragment) (listed in Supplementary Table S8).

RNA immunoprecipitation (RIP) assay

RIP experiments were carried out using the Magna RNA-binding protein immunoprecipitation kit from Millipore (Burlington, USA). Cells were processed at a concentration of 1 × 10⁷ cells per reaction and lysed in a buffer containing protease and RNase inhibitors for 5 min. Magnetic beads were then combined with 5 μg of antibody or control IgG at room temperature and incubated with the cell lysates overnight at 4 °C. Following this, RNA purification and protein extraction were performed. The immunoprecipitated RNA from Caki-1 and 786-O cells was validated using RT-qPCR with specific primers (listed in Supplementary Table S4).

Cholesteryl ester/free-cholesterol quantitation assay

Based on the instruction of Cholesterol/Cholesteryl Ester Quantitation Assay kit (Abcam, UK), The amount of Caki-1 and 786-O cells (1 × 10⁶ cells) were harvested and washed with cold PBS. Next, lipids were extracted by resuspending the sample in 200 µL of Chloroform: Isopropanol: NP-40 (7:11:0.1) in a micro-homogenizer and spin the extract 5–10 min at 15,000 × g in a centrifuge. the extract should under air dry at 50 °C to remove chloroform. Dissolved the dried lipids with 200 µL of Assay Buffer II/Assay Buffer. Relative reaction Mix were produced and added in to each sample according to the protocol and incubate at 37 °C for 60 min protected for light, measuring with Ex/Em = 535/587 nm for fluorometric assay.

Colony formation assay

Caki-1 and 786-O cells (3 × 10³ cells/well) were seeded onto 6-well plates to evaluate proliferation. Following fifteen-days incubation in complete medium, cells were washed with PBS, fixed with 4% paraformaldehyde (Biosharp, China), and subsequently stained with 1% crystal violet (Beyotime, China). Images were acquired utilizing a scanner (EPSON, Japan).

Migration and invasion assay

As previously reported in our study [37], we investigated the migration and invasion abilities of cells using Transwell migration and invasion assays, adhering to established protocols from prior research.

Luciferase reporter assay

The SCARB1 sequence was inserted downstream of the p-GLO Dual-Luciferase vector (Vigenebio, USA). Mutations were introduced into the binding sites. CcRCC cells were plated in 24-well plates at a confluence of 30% per well 24 h prior to transfection and then co-transfected with 800 ng of the p-GLO Dual-Luciferase reporter plasmid. 48 h later, the relative luciferase activity was determined by calculating the ratio of Firefly luciferase activity to Renilla luciferase activity using a dual luciferase reporter assay kit (Promega, USA). Renilla luciferase activity served as an internal standard. For the circABCA1 and IGF2BP3 overexpression group, the Luc/Rluc ratio was additionally normalized to that of the control sample.

The establishment co-culture system between ccRCC cells and M2 macrophage

THP-1 cells with a density of 80–90% were selected, digested, and counted before being resuspended in RPMI-1640 complete medium at a concentration of 5 × 10⁵ cells/mL. Subsequently, 100 ng/mL of PMA was added. After 24 h, the supernatant was discarded, and the cells were washed three times with PBS. The cells were then cultured in RPMI-1640 complete medium until they reached approximately 80% confluence. At this stage, M0 macrophages were collected and treated with IL-4 (20 ng/mL) for 48 h to obtain M2 macrophages. A Transwell system was set up on a 24-well plate, with 1 × 10⁶ ccRCC cells, which was been transfected with si-circABCA1 and/or SCARB1 plasmid as well as NC were seeded in the upper chamber and M2 macrophages in the lower chamber. Both cell types were evenly distributed and cultured for 48 h.

In vivo tumorigenesis and metastasis assays

For the assessment of in vivo tumorigenesis and metastasis, NSG mice (4 weeks of age, 12–14 g) bearing subcutaneous tumors of 786-O and Caki-1 cells were sourced from the Shanghai Model Organisms Center. For the si-circABCA1 group, cholesterol-modified circABCA1-siRNA (10 nmol, GenePharma) was administered intratumorally every two days for 30 days. Tumor volumes were recorded every 5 days, and post-euthanasia, tissues were excised, weighed, fixed, and subjected to immunohistochemical staining to measure the proliferative markers using anti-Ki67 (Bioss, China) and the expression of SCARB1 using anti-SCARB1 (ABclonal, China). Images were obtained using a fluorescence microscopy (Leica, Germany). In the overexpression group, a lentiviral vector carrying cloned inserts (Oligobio, China) was injected intratumorally for 30 days, with tumor monitoring and processing as above. For the hepatic metastasis model, it was generated by intrasplenic injection of 2.5 × 10⁴ of Caki-1 and 786-O cells and splenectomy. After 60 days, excised hepatic tissues were photographed and evaluated for metastatic nodules then stained with hematoxylin and eosin (H&E). All animal procedures adhered to the ethical guidelines approved by the Chongqing Medical University Ethics Committee (approval number: 2022013). Our study examined male and female animals, and similar findings are reported for both sexes. Once the tumors reached a suitable volume, the mice were randomly assigned to groups, ensuring that the initial tumor volume was similar in each group.

Statistical analysis

Data were represented as mean ± SD from triplicate independent experiments. Statistical comparisons employed Student's t-test for pairwise analysis and one-way ANOVA with Tukey's correction for multiple groups, with significance set at P < 0.05. Analyses were conducted using SPSS 20 (RRID:SCR_002865) or GraphPad Prism 9.0, confirming normal distribution (P > 0.05, Shapiro–Wilk test), justifying the application of parametric tests.

Abbreviations

ccRCC

Clear cell renal cell carcinoma

circRNA

Circular RNA

LD

Lipid droplet

TAM

Tumor-associated macrophage

SCARB1

Scavenger Receptor Class B Member 1

CE

Cholesteryl ester

TG

Triglycerides

CHO

Cholesterol

GO analysis

Gene Ontology analysis

FISH

Fluorescence in situ hybridisation

T-CHO

Total CHO

MβCD

Methyl-β-cyclodextrin

HDL

High-density lipoprotein

RIP

RNA Immunoprecipitation

CCK-8

Cell Counting Kit-8

Chlo

Chloroquine

NC

Negative control

FC

Free cholesterol

IGF1R

Insulin Like Growth Factor 1 Receptor

AKT

Protein Kinase B

PI3K

Phosphatidylinositol 3-Kinase

mTOR

Mammalian Target of Rapamycin

CM

Supernatant culture medium

Authors' contributions

H.N. Y.J. and B.X. conceived and directed the project. Y.J., H.N., B.B.L., J.W.L., A.R. J.L., Z.J.L. performed the experiments. H.N. performed the bioinformatic and statistical analysis. Y.Q.W., L.F.L., L.W., C.W. performed the sample staining and pathological diagnosis. H.N., B.X., W.L. wrote, reviewed, and/or revised the manuscript. All authors discussed and approved the final manuscript.

Funding

This study was supported by the National Natural Science Foundation of China (82373001), Chongqing Talents-Exceptional Young Talents Project (CQYC202005044, CSTC2021YCJH-BGZXM0094), Chongqing Natural Science Foundation Innovation and Development Joint Fund (CSTB2022NSCQ-LZX0043), Science and Technology Research Project of Chongqing Municipal Education Commission (KJZD-K202100405), Future Medical Youth Innovation Team Project of Chongqing Medical University (W0042), Graduate Tutor Team Construction Project of Chongqing (CQMUDSTD202210), Top Graduate Talent Cultivation Program of Chongqing Medical University (BJRC202322), Natural Science Foundation of Chongqing (CSTC2022JXJL120012, 2024NSCQ-KJFZMSX0048).

Data availability

All data necessary to evaluate the conclusions of the paper are included in the paper and/or supplementary materials. Additional data related to this paper may be requested from the authors.

Declarations

Ethics approval and consent to participate

All procedures were performed in compliance with relevant laws and institutional guidelines and have been approved by the appropriate institutional committee. Patient tissue collection was approved by the Ethics Committee of the Southwest Hospital of Army Medical University, Chongqing (Approval No.: KY2020121) in 2020. The privacy rights of human subjects have been observed and that informed consent was obtained for experimentation with human subjects. All animal experiments were conducted in accordance with the protocol approved by the Institutional Animal Care and Use Committee of Chongqing Medical University, Chongqing (Approval No.:2022013) in 2022.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.