CircSipa1l1 modulates melanoma cell differentiation by activating the IGF2BP1-ARHGDIB axis and ERK signaling pathway

Abstract

Background

Clinical evidence demonstrates that induction differentiation therapy is a useful treatment strategy for melanoma. Circular RNAs (circRNAs) plays a crucial role in melanoma cell proliferation, resistance and metastasis. However, the roles of circRNAs during melanoma cell differentiation have not been fully investigated. This study aimed to investigate the role and mechanism of circSipa1l1 in melanoma cell differentiation.

Methods

All-trans-retinoic acid (ATRA) or sodium phenylbutyrate-4 (PB-4) were employed to induce melanoma B16 cells differentiation, and whole transcriptome sequencing was performed to screen for differentially expressed circRNAs. RNA stability assay, quantitative real-time polymerase chain reaction (qRT-PCR), tissue microarray and fluorescence in situ hybridization (FISH) was employed to confirm the existence, expression level and subcellular localization of circSipa1l1. Cell counting kit-8 (CCK-8), colony formation, cell cycle analysis, melanin content, tyrosinase activity assay, RNA pull-down, RNA immunoprecipitation (RIP) and western blotting were used to evaluate the effect of circSipa1l1 on melanoma cell differentiation and explore its regulatory mechanism. Finally, mouse xenograft models were used to assess the effect of circSipa1l1 silencing on tumor growth in vivo.

Results

CircSipa1l1 was significantly downregulated in ATRA- or PB-4-treated B16 cells and highly expressed in melanoma patient tissues. Silencing circSipa1l1 induced cell-cycle arrest and differentiation in melanoma A375 and B16 cells, while its overexpression promoted proliferation. Mechanistically, circSipa1l1 directly interacts with insulin-like growth factor 2 mRNA binding protein 1 (IGF2BP1), a key RNA-binding protein. Silencing circSipa1l1 inhibited the IGF2BP1 and rho GDP-dissociation inhibitor 2 (ARHGDIB) mRNA interaction, destabilizing ARHGDIB mRNA and subsequently inhibiting the extracellular signal-regulated kinase (ERK) signaling pathway—ultimately inducing differentiation and repressing cell cycle progression. Furthermore, silencing circSipa1l1 significantly inhibited tumor growth in both B16 and A375 xenograft models.

Conclusion

Our findings reveal that circSipa1l1 acts as an oncogenic circRNA by regulating the IGF2BP1/ARHGDIB/ERK axis in melanoma, suggesting it could be a potential therapeutic target for melanoma differentiation therapy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07233-4.

Background

Melanoma, a malignancy of the skin, arises from either congenital or acquired benign nevi and is typically found on the skin or mucosal surfaces. Cutaneous melanoma is the most metastatic human cancer, and the current treatment options include surgical excision, immunotherapy, targeted therapy, and radiotherapy [1, 2]. Although surgery can effectively cure melanoma in its early stages, patients deteriorate rapidly as the disease progresses. Moreover, melanoma is highly aggressive and prone to metastasis, resulting in a bleak prognosis [3, 4]. Consequently, it is necessary to acquire a comprehensive understanding of the etiology and progression of melanoma, as well as to devise novel strategies for prevention and treatment.

Cancer is caused by uncontrolled proliferation. Tumor-inducing differentiation therapy stops the proliferation of malignant cells and promotes the differentiation of malignant cells into an epithelial cell phenotype [5–7]. Melanomas usually appear in areas or sites of normal skin that develop along the neural crest, suggesting that melanomas may have a chance to change to normal cells through a process of induced redifferentiation [8–10]. Thus, inducing melanoma cells to redifferentiate into normal cells could be an effective therapeutic intervention.

Circular RNAs (circRNAs) generated by a process referred to as back-splicing have recently attracted considerable attention. Due to the cyclized molecules and lack of 5’ caps and 3’ poly (A) tails, circRNAs are stable, have a long half-life, and are more conserved than linear mRNA [11]. Some circRNAs can serve as competitive endogenous RNAs for microRNAs, facilitate protein-protein interactions by acting as scaffolds, or mediate the recruitment of proteins to distinct subcellular locales [12–14]. Furthermore, circRNA plays an important role in melanoma proliferation, metastasis, and resistance [13, 15]. However, studies on how circRNAs regulate melanoma differentiation remain limited.

In this study, we clarified a specific circRNA, circSipa1l1, which was found to be highly conserved and significantly upregulated in melanoma. We identified circSipa1l1 as a differentiation therapy target in melanoma. Mechanistically, circSipa1l1 binding to insulin-like growth factor 2 mRNA binding protein 1 (IGF2BP1) enhanced the stability of rho GDP-dissociation inhibitor 2 (ARHGDIB) mRNA, which in turn affected melanoma growth, differentiation, and cell cycle progression through the extracellular signal-regulated kinase (ERK) signaling pathway. Our findings reveal a potential new mechanism of circSipa1l1/IGF2BP1/ARHGDIB axis-mediated melanoma progression and provide a novel therapeutic target for melanoma differentiation therapy.

Methods

Reagents

The Universal Virus Concentration Kit (Cat no. C2901S) were purchased from Beyotime (Shanghai, China). All-trans retinoic acid (ATRA) (Cat no. 55547) and sodium phenylbutyrate-4 (PB-4) (Cat no. 567616) were purchased from Sigma Aldrich (Shanghai, China). The primary antibodies used are as follows: cyclin-dependent kinase 6 (CDK6) (Cat no. sc-796, 1:1000), CyclinD1 (Cat no. sc-8396, 1:1000), and cyclin-dependent kinase 4 (CDK4) (Cat no. sc-23896, 1:1000), purchased from Santa Cruz Biotechnology (Shanghai, China) Co., Ltd.; microphthalmia-associated transcription factor (MITF) (Cat no. 1309I-1-Ap, 1:1000), p-ERK (Cat no. 28733-1-AP, 1:1000), ARHGDIB (Cat no. 16122-1-AP, 1:1000), and IGF2BP1 (Cat no. 22803-1-AP, 1:1000) purchased from Proteintech China (Wuhan, Hubei, China); tyrosinase-related protein 2 (TRP2) (Cat no. ab221144, 1:1000) and Tyrosinase (TYR) (Cat no. ab180753, 1:1000) purchased from Abcam (Shanghai, China); tyrosinase-related protein 1 (TRP1) (Cat no. bs-15510R, 1:1000) purchased from Bioss (Beijing, China); ERK (Cat no. YT1625, 1:1000) purchased from RuiYing Biotechnology Company (SuZhou, JiangSu, China); and β-actin (Cat no. T0022, 1:5000) purchased from Affinity Biosciences (JiangSu, China).

Cell culture

Mouse melanoma B16 cells (Cat no. SCSP-5096), human melanoma A375 cells (Cat no. SCSP-533), and human immortalized keratinocyte cells (HaCaT) (Cat no. SCSP-5091) were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). B16 and A875 cells were cultured in HG-DMEM (Cat no. RNBL6943, Sigma, Shanghai, China) with 10% FBS, and HaCaT cells were cultured in RPMI-1640 (Cat no. RNBL6943, Sigma, Shanghai, China) with 10% FBS. In addition, the A375 cell culture medium was supplemented with 1% sodium pyruvate (100 mM) (Cat no. SP0100, Solarbio, Beijing, China). Cell transfection was performed according to the manufacturer’s instructions. All small-interfering RNAs (siRNAs), small hairpin RNAs (shRNAs), and overexpression (OE) plasmids were purchased from IBSBIO Co., Ltd. (Shanghai, China). Lipofectamine™ 3000 (Cat no. L3000015, Thermo Fisher Scientific, Shanghai, China) was used for transfection.

Animals

The animal experiments conducted in this study were purchased from GemPharmatech (Jiangsu, China) and strictly adhered to the Institutional Guidelines for Animal Care of Binzhou Medical University (BZMU-2021-221; BZMU-2022-252). For the B16 melanoma xenograft model, 6 weeks of C57BL/6 male mice were subcutaneously inoculated with B16 cells, and tumor growth was subsequently measured every 3 days. After the tumor volume reached approximately 100 mm³, the mice were randomly divided into two groups, sh-NC and sh-circSipa1l1 lentivirus were injected in situ into the tumors of the mice by multipoint intratumoral injection every 3 days for 15 days. For the A375 melanoma xenograft model, 6 weeks of BALB/c-nu male mice were subcutaneously inoculated with circSipa1l1-silenced A375 cells and related control A375 cells and then randomly divided into four groups. The tumor volumes and mouse weights were measured and recorded at 3-day intervals. After 15 days, an empty vector or ARHGDIB-overexpressing lentivirus (OE-ARHGDIB) was injected in situ into the tumors of control and treated mice by multipoint intratumoral injection twice a week, respectively. The mice were euthanized 2 weeks after tumor injection, and the tumors were dissected, photographed, and weighed. The tissue and serum were taken for subsequent experiments.

Tissue microarray and in situ hybridization

Tissue microarrays containing a total of 88 tumor tissues from melanoma patients and seven controlled normal tissues were obtained from Zhuoli Biotechnology Company Ltd. (Shanghai, China). The collection of human tissue microarrays was approved by the Zhuoli Biotechnology Company Biomedical Ethics Committee (ZLL-15-01). Briefly, the 1.0-mm-diameter sections obtained from the per-donor block were redirected to a recipient block microarrayer. Sections with a thickness of 2 μm were extracted from the recipient block and subsequently affixed onto glass slides using an adhesive tape transfer system for ultraviolet cross-linkage. The process of in situ hybridization was carried out following the manufacturer’s guidelines (BOSTER Biological Technology Company Ltd, Wuhan, China). Briefly, the slices were dewaxed with xylene and rehydrated with 100%, 75%, and 50% ethanol. Following digestion with proteinase K, the specimens were immobilized, hybridized with the circSipa1l1 probe at 55 °C overnight, incubated with HRP, and stained with diaminobenzidine.

CCK-8 assay

B16 and A375 cells were initially placed in 96-well plates. The next day, Lipofectamine 3000 (Cat no. L3000015, Thermo Fisher Scientific, Shanghai, China) was used to transfect the cells with siRNAs or plasmids. The cells were then cultured for another 72 h. Subsequently, each well received Cell Counting Kit-8 (CCK-8) solution (Cat no. CA1210, Solarbio, Beijing, China), and was incubated at 37 °C for a duration of 2 h. The absorbance at 450 nm was quantified using a micro-ELISA plate reader (Bio-Tek Inc., USA).

Colony formation assay

B16 and A375 cells were seeded in 12-well plates at a density of 200 cells per well. After 3–4 days, the cells were transfected and cultured for another week until cell clumps were visible. The cells were then fixed with 4% PFA (Cat no. BL539A, Biosharp, Shanghai, China), and stained with 1% crystal violet (Cat no. G1062, Solarbio, Beijing, China), and imaged after discarding the staining solution.

Melanin content

siRNA or plasmid was transfected into the B16 and A375 cells, followed by another treatment after 72 h. The cells were then treated with sodium hydroxide solution and absorbance was measured at 400 nm using a micro-ELISA plate reader (Bio-Tek Inc., USA).

Tyrosinase activity assay

B16 and A375 cells were plated and siRNA or plasmid was transfected. After transfection and incubation for 72 h, cell viability was assessed using CCK-8 (10%, Solarbio, Beijing). The cells were rinsed twice with PBS, followed by digestion and centrifugation to collect cells. The cells were treated with 100 µL of a 5% sodium deoxycholate solution (Cat no. IT0370, Solarbio, Beijing, China), and then cooled on ice for 15 min before being placed in a water bath at 37 °C for 10 min. Afterwards, 300 µL of 1% L-DOPA solution (Cat no. ID0360, Solarbio, Beijing, China) was added to each well. The plate was then incubated in a water bath at 37 °C for 2 h, and the solution was transferred to the plate. Lastly, the absorbance at 475 nm was measured using a micro-ELISA plate reader (Bio-Tek Inc., USA).

Cell cycle analysis

B16 and A375 cell lines were seeded and subjected to transfection with either siRNA or plasmid. After 72 h, the treated B16 and A375 cells were digested and fixed in 70% ethanol at 4 °C overnight. To assess the DNA content, the DNA content quantitation assay (Cell Cycle) kit (Cat no. CA1510, Solarbio, Beijing, China) was used. The samples were then extracted and examined using a flow cytometer (BD FACSCalibur, USA).

RNA fluorescence in situ hybridization (FISH) and immunofluorescence (IF)

The m-circSipa1l1 probe was synthesized by Shanghai Integrated Biotech Solutions Co., Ltd. (Shanghai, China) and the h-circSipa1l1 probe was designed by RiboBio Co., Ltd. (Guangzhou, China). The cells were fixed and treated with 0.5% Triton X-100. Prehybridization was performed, followed by the addition of denatured probes. The cells were then subjected to hybridization at 37 °C overnight. To detect IGF2BP1 (Cat no. 22803-1-AP, Proteintech, USA), the primary antibody was added and incubated overnight at 4 °C. Subsequently, the cells were treated with the secondary antibody Alexa Fluor 488 (anti-Rabbit) (Cat no. A-21206, Thermo Fisher Scientific, USA) for 1 h at room temperature. DAPI staining was used to visualize the cell nuclei, and a fluorescence confocal microscope (Zeiss LSM880, Germany) was employed for sample visualization. For quantification of nuclear vs. cytoplasmic circSipa1l1 level via RNA fluorescence in situ hybridization (FISH), confocal images were processed in ImageJ. The nuclear region was delineated based on DAPI staining, and the cytoplasmic region was defined as the entire cell area excluding the nuclear region. The integrated density of Cy3 fluorescence (representing circSipa1l1) was measured in both the nuclear and cytoplasmic regions for each cell. The percentage of circSipa1l1 fluorescence in the nucleus or cytoplasm was calculated as: (Integrated density of Cy3 fluorescence in the nucleus/cytoplasm) / (Total integrated density of Cy3 fluorescence in the entire cell) × 100%. For circSipa1l1/IGF2BP1 co-localization analysis, confocal images were analyzed using ImageJ software with the JACoP plugin. Alexa Fluor 488 (IGF2BP1) and Cy3 (circSipa1l1) channels were separated from the merged image. Region-of-interest (ROI) line scanning was performed to obtain grayscale intensity profiles, and data were plotted using GraphPad Prism.The Pearson’s correlation coefficient (r) was calculated to quantify co-localization; a Pearson’s r > 0.5 was considered indicative of significant co-localization.

RNA stability assay

For the actinomycin D assay, 2 µg/mL actinomycin D (Cat no. HY-17559, MCE, USA) was added to B16 or A375 cells or A375 cells transfected with siRNA or plasmids at the corresponding time points. For the RNase R treatment assay, total RNA was subjected to incubation with 5 U/µg RNase R (Cat no. R7092S, Beyotime, Shanghai, China) at 37 °C for 10 min. Afterward, the total RNA was extracted, and subsequent processes including reverse transcription and quantitative real-time polymerase chain reaction (qRT‒PCR) analyses were carried out as previously described.

Nuclear and cytoplasmic fraction assay

RNA extraction of the nucleus and cytoplasm was performed separately using a Nuclear and Cytoplasmic Protein Extraction kit (Cat no. P0028, Beyotime, Shanghai, China). B16 and A375 cells were gathered and treated with cytosolic protein extractant A, which included RNase inhibitors. The resulting mixture was incubated in an ice bath for 15 min, after which cytosolic protein extractant B was added and the solution was centrifuged. TRIzol (Cat no. 15596026, Invitrogen, Shanghai, China) was added to the supernatant, which was precipitated for subsequent cytosolic and nuclear RNA extraction. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed to detect circSipa1l1 expression, where U6 small nuclear RNA (snRNA) served as the nuclear internal control, and GAPDH mRNA served as the cytoplasmic internal control. The relative expression of circSipa1l1 in each fraction was calculated using the 2 − ΔΔCt method. Data were analyzed with GraphPad Prism (Prism 10) software, and the nuclear/cytoplasmic ratio of circSipa1l1 was presented as mean ± S.E.M. from three independent experiments.

Biotin-labeled RNA pull-down

RNA pull-down experiments were performed as per the RNA pull-down kit manufacturer’s instructions (Cat no. Bes5102, BersinBio, Guangzhou, China). Briefly, the denatured circSipa1l1-labeled biotin probe was incubated with streptavidin beads for 30 min at room temperature in a rotating incubator. Nucleic acid was removed, and prewashed protein samples were added to the magnetic beads and centrifuged at room temperature for 2 h. The final supernatant was used for mass spectrometry, silver staining, and western blotting analyses.

RNA Immunoprecipitation

The RNA immunoprecipitation (RIP) assay was executed utilizing an RIP Kit (Cat no. Bes5101, BersinBio, Guangzhou, China). Briefly, the cells were lysed, DNA was removed from the lysates, equal amounts of antibody IGF2BP1 and antibody IgG were added separately, and the samples were subjected to overnight incubation at 4 °C. Subsequently, magnetically charged beads were added and the samples were further incubated at 4 °C for a duration of 1 h. Following this, RNA was extracted and eluted for subsequent analysis using qRT‒PCR.

RNA and genomic DNA (gDNA) extraction, and nucleic acid electrophoresis

RNAs were extracted from B16 and A375 cells using TRIzol method by the manufacturer’s guidelines. For the extraction of gDNA from B16 and A375 cells, the SteadyPure Universal Genomic DNA Extraction Kit (Cat no. AG21010, Accurate Biology, Changsha, China) was employed following the manufacturer’s instructions. The gel was placed into an electrophoresis tank for PCR product separation (including a DNA ladder). The gel was visualized using a UV lamp and an imaging system (BioSpectrum 510 imaging system, motorized platform, Germany).

RNA sequencing and qRT–PCR analysis

RNA-seq analysis was conducted by Novogene Corporation Inc. (Beijing, China). The RNA was reverse-transcribed into cDNA using ABScript III RT Master Mix for qPCR (Cat no. RK20428, ABclonal, China), and qRT-PCR was performed using 2× Universal SYBR Green Fast qPCR Mix (Cat no. RK21203, ABclonal, China). Internal references such as GAPDH, β-actin, and U6 were used, and the mRNA expression levels were determined using the 2^−ΔΔCt method. The primer sequences could be found in Supplementary Table 1.

Western blotting

B16 and A375 cells were collected after 72 h treated with siRNA or plasmid and lysed using RIPA buffer (Cat no. R0010, Solarbio, Beijing, China) containing PMSF (Cat no. P8340, Solarbio, Beijing, China) and protease inhibitor mixture (Cat no. P6730, Solarbio, Beijing, China). Subsequently, 5 × loading buffer (Cat no. P1040, Solarbio, Beijing, China) was added, and the samples were denatured at 100 °C in a metal bath. The protein samples were then separated on SDS-PAGE gels and transferred onto PVDF membranes (Cat no. ISEQ00010, Merck Millipore, Ireland). Following blocking at room temperature, the membranes were incubated with specific antibodies overnight at 4 °C. After washing with TBST, the membranes were incubated with either anti-mouse IgG (1:10000, Cat no. ZB-2305, ZSGB-BIO, Beijing, China) or anti-rabbit IgG (1:10000, Cat no. SA00001-2, Proteintech, Wuhan, China). Finally, the membranes were visualized using ECL Western Blotting Substrate (Biosharp, China) and an imaging system (BioSpectrum 510, motorized platform, Germany).

Histology

The fixed tissues were dehydrated, immersed in paraffin for embedding, sectioned, and then stained with hematoxylin and eosin (H&E). For immunohistochemical (IHC) staining, the samples were dewaxed and hydrated, antigens were repaired using citrate buffer, incubated with primary antibody ARHGDIB (Cat no. S16122-1-AP, Proteintech, USA) and anti-rabbit IgG, and then stained with DAB chromogenic solution and hematoxylin.

Statistical analysis

All data were taken from three independent and representative trials. The experimental data were then visualized and analyzed using GraphPad Prism (Prism 10) software and expressed as the mean ± standard error of mean (mean ± S.E.M.). A two-sided unpaired t-test was used for comparisons between two groups with normal distribution, ANOVA used to calculate the statistical significance for multiple comparisons of more than two groups. Statistical significance was determined at a p-value of P < 0.05.

Results

CircSipa1l1 expression was significantly downregulated in ATRA- or PB-4-treated B16 cells and upregulated in tumor tissues from melanoma patients

To investigate the role of circRNAs in the differentiation of melanoma cells, ATRA and PB-4 were used to induce melanoma differentiation. Cell morphological observation, CCK-8, and colony formation assays showed that ATRA and PB-4 significantly suppressed B16 cell proliferation and colony formation ability (Supplementary Fig. 1A-F). Moreover, when compared with the control group, the proportion of G0/G1 phase B16 cells was enhanced significantly after ATRA or PB-4 treatment, indicating that the cell cycle was arrested at the G0/G1 phase (Supplementary Fig. 1G-H). In addition, melanin content is a significant factor in identifying the differentiation of melanoma cells [16, 17]. ATRA or PB-4 treatment markedly increased B16 cell melanin production (Supplementary Fig. 1I-J). The activity of tyrosinase, a key enzyme in melanin synthesis [18], was also significantly enhanced in ATRA- and PB-4-treated B16 cells (Supplementary Fig. 1K-I). These results indicate that ATRA and PB-4 induce melanoma cell cycle arrest and differentiation.

Next, we performed RNA-sequencing analyses on melanoma B16 cells treated with ATRA and PB-4 to identify the potential circRNAs contributing to cell differentiation. Only three co-differentially expressed circRNAs were identified in the ATRA- and PB-4-treated groups in comparison with the control group; circAkap7 and circDennd1b were significantly upregulated, while circSipa1l1 was significantly downregulated (Fig. 1A). The expression of these three circRNAs was further examined by qRT‒PCR in melanoma cells, and circSipa1l1 expression demonstrated the same trend as the RNA-seq results (Fig. 1B). Therefore, we focused on circSipa1l1 in our further investigations.

Fig. 1.

CircSipa1l1 expression was low in ATRA- or PB-4-treated melanoma B16 cells and high in melanoma tissues. (A) Heatmap of the differentially expressed circRNAs among the ATRA-treated (8 µM), PB-4-treated (4 mM), and control groups. (P < 0.05 and |fold change| >2.0). (B) qRT‒PCR analysis of the expression levels of three circRNAs in ATRA- (8 µM) and PB-4- (4 mM) treated B16 cells. (C, D) qRT‒PCR analysis of circSipa1l1 and β-actin amplified from cDNA or gDNA from melanoma B16 and A375 cells with divergent and convergent primers. (E, F) qRT‒PCR analysis of circSipa1l1 and Sipa1l1 mRNA levels in B16 and A375 cells with or without RNase R treatment. (G, H) qRT‒PCR analysis of circSipa1l1 and Sipa1l1 mRNA levels in B16 and A375 cells after treatment with actinomycin D. (I, J) RNA fluorescence in situ hybridization for the localization of circSipa1l1 in B16 and A375 cells. Scale bar: 20 μm. (K, L) qRT‒PCR analysis of circSipa1l1 expression in the cytoplasm or nucleus of B16 and A375 cells. (M) The level of circSipa1l1 was measured in B16, A375, and A875 melanoma cell lines and HaCaT cells by qRT‒PCR. (N, O) Representative images and positive percentages of circSipa1l1 in melanoma tissues and peritumor tissues. Scale bars: 100 μm. *P < 0.05, **P < 0.01, ***P < 0.001 versus corresponding controls

To explore the conservation of circSipa1l1, we aligned the murine-derived sequence circSipa1l1 with the human genome sequence and found a human-derived sequence with 71.13% homology (Supplementary Fig. 2A-B). To verify the existence of circSipa1l1 in melanoma cells, we designed convergent primers for the amplification of linear Sipa1l1 mRNA and divergent primers for the back-spliced form of circSipa1l1. Complementary DNA (cDNA) and gDNA were employed as the PCR templates, while β-actin was used as the control. Nucleic acid electrophoresis assays indicated that circSipa1l1 could be amplified from the cDNA but not gDNA of B16 and A375 cells by the divergent primers (Fig. 1C, D). Subsequently, B16 and A375 cells were treated with RNase R, an exoribonuclease that specifically degrades linear mRNA. The qRT‒PCR results confirmed that circSipa1l1 but not linear Sipa1l1 mRNA was resistant to digestion by RNase R, confirming its closed circular form (Fig. 1E, F). Furthermore, we used actinomycin D to inhibit the transcription of B16 and A375 melanoma cells, and the stability of circSipa1l1 and linear Sipa1l1 was further analyzed. Following actinomycin D treatment, circSipa1l1 expression was slightly changed, while Sipa1l1 mRNA expression was significantly downregulated, certifying that circSipa1l1 is more stable than the linear mRNA (Fig. 1G, H).

Furthermore, a FISH assay was used to examine the subcellular localization of circSipa1l1. CircSipa1l1 was mainly expressed in the cytoplasm of melanoma B16 and A375 cells rather than in the nucleus (Fig. 1I-J), consistent with the nuclear and cytoplasmic separation assay results obtained by qRT‒PCR (Fig. 1K-L). Moreover, the expression of circSipa1l1 was significantly higher in B16, A375, and A875 cells than in HaCaT cells (Fig. 1M), indicating that circSipa1l1 is highly expressed in melanoma cells. To further investigate the pathologic role of circSipa1l1 in melanoma development, 88 tissues from melanoma patients and seven controlled normal tissues were used to examine circSipa1l1 expression by RNA in situ hybridization. CircSipa1l1 was mainly enriched in the intratumoral regions rather than in the peritumoral regions (Fig. 1N, O), consistent with the higher level of circSipa1l1 observed in melanoma cells. Overall, these results demonstrated that circSipa1l1 is a highly conserved and stable transcript expressed at higher levels in tumor tissues from melanoma patients than in control tissues.

CircSipa1l1 silencing induces melanoma cell cycle arrest and differentiation

Three different siRNAs specifically targeting splicing junctions were synthesized to silence circSipa1l1 in B16 and A375 cells to explore whether circSipa1l1 contributes to melanoma cell differentiation (Fig. 2A, B). The CCK-8 assay showed that the proliferation of B16 and A375 melanoma cells was remarkably inhibited after silencing circSipa1l1 with siRNA-#1, siRNA-#2, and siRNA-#3 (Fig. 2C, D). According to the silencing and inhibitory efficiency, we finally selected circSipa1l1 siRNA-#1 (hereafter abbreviated as si-circSipa1l1) for subsequent exploration. Consistent with ATRA or PB-4 treatment, si-circSipa1l1 substantially reduced the cell number, and the cell morphology demonstrated a long spindle shape (Fig. 2E). Colony formation assays revealed that the colony formation ability of B16 and A375 cells was remarkably suppressed after silencing circSipa1l1 (Fig. 2F-H).

Fig. 2.

CircSipa1l1 silencing induces cell cycle arrest and differentiation in melanoma cells. (A) qRT‒PCR analysis of the effect of transfecting si-circSipa1l1 on circSipa1l1 and Sipa1l1 expression in melanoma B16 cells after 72 h of transfection. (siNC, siRNA with scrambled sequences; si-circSipa1l1#1–3, three different siRNAs against the junction sites of mouse circSipa1l1). (B) qRT‒PCR analysis of the effect of si-circSipa1l1 transfection on circSipa1l1 and Sipa1l1 expression in melanoma A375 cells after 72 h of transfection. (siNC, siRNA with scrambled sequences; si-circSipa1l1#1–3, three different siRNAs against the junction sites of human circSipa1l1. (C, D) CCK-8 assays of B16 and A375 cell viability after transfection with si-circSipa1l1. (E) Cell morphological observation of B16 and A375 cells after circSipa1l1 silencing. (F‒H) A colony formation assay was used to assess cell survival in melanoma cells after transfection with si-circSipa1l1. (I, J) A flow cytometry assay was used to assess the cell cycle distribution in B16 and A375 cells after transfection with si-circSipa1l1. (K‒M) The levels of cell cycle-related proteins were measured by western blotting. (N, O) Melanin content was determined in B16 and A375 cells after transfection with si-circSipa1l1. (P, Q) Tyrosinase activity was measured in B16 and A375 cells after transfection with si-circSipa1l1. (R-T) The levels of melanin synthesis-related proteins were detected by western blotting. *P < 0.05, **P < 0.01, ***P < 0.001 versus corresponding controls

Moreover, flow cytometry assays showed that the proportion of G0/G1 phase cells was significantly increased after silencing circSipa1l1, suggesting that silencing circSipa1l1 induces melanoma cell cycle arrest at the G0/G1 phase (Fig. 2I, J, Supplementary Fig. 3A, B). Then, G0/G1 phase-associated proteins were further examined by western blotting, and the expression of cyclin D1, CDK4, and CDK6 was markedly suppressed after circSipa1l1 silencing in both the B16 and A375 cell lines (Fig. 2K-M). In addition, the melanin content and tyrosinase activity were significantly enhanced (Fig. 2N-Q), and the expression of melanin synthesis-related genes TYR, TRP1, and TRP2 was markedly increased in B16 and A375 cells after circSipa1l1 silencing (Fig. 2R-T). MITF is essential in governing the viability and functionality of melanocytes and in regulating proteins linked to the tyrosinase family, such as TYR, TRP1, and TRP2 [19]. Consistently, we also found an upregulation of the MITF protein after circSipa1l1 knockdown (Fig. 2R-T). These results indicated that silencing circSipa1l1 causes melanoma cell cycle arrest and induces melanoma cell differentiation.

We then overexpressed circSipa1l1 in B16 and A375 cells via transfection with a circSipa1l1 overexpression plasmid to further evaluate the biological function of circSipa1l1 in melanoma cells (Supplementary Fig. 4A, B). After circSipa1l1 overexpression, the CCK-8 assay, morphological observation, and colony formation assay showed increased cell viability and colony proliferation ability (Supplementary Fig. 4C-H). Interestingly, the overexpression of circSipa1l1 markedly decreased the melanin content and tyrosinase activity in both B16 and A375 melanoma cells (Supplementary Fig. 4I-L). Western blotting also demonstrated that circSipa1l1 overexpression could downregulate the expression of the cell differentiation-related proteins MITF, TRP1, TRP2, and tyrosinase (Supplementary Fig. 4M-O). These findings indicated that circSipa1l1 overexpression induces melanoma cell proliferation, which was opposite to circSipa1l1 silencing.

CircSipa1l1 enhances ARHGDIB mRNA stability through the recruitment of IGF2BP1

Next, we further explored the mechanism of circSipa1l1 in melanoma cell differentiation. Since circRNA has been extensively reported to function as an miRNA sponge in the cytoplasm and circSipa1l1 is stable and abundant in the cytoplasm, we speculated that circSipa1l1 might bind to miRNAs [20]. However, the RIP assay revealed that argonaut 2 (AGO2), the circRNA-miRNA interaction mediator, had no detectable enrichment with circSipa1l1 in A375 melanoma cells (Fig. 3A), indicating that circSipa1l1 might function via a different mechanism than miRNA sponging.

Fig. 3.

CircSipa1l1 plays a critical role in enhancing the stability of ARHGDIB mRNA by recruiting IGF2BP1. (A) A RIP assay was performed using anti-AGO2 or anti-IgG antibodies in A375 cell lysates, and the enrichment of circSipa1l1 was measured by qRT‒PCR. (B) Silver staining of potential circSipa1l1-interacting proteins after the RNA pull-down experiment using the specific biotin-labeled circSipa1l1 probe in A375 cell lysates. (C) Scatter image of the potential circSipa1l1-interacting proteins identified by RNA pull-down analysis. (D) The IGF2BP1 protein content was analyzed by western blotting after the RNA pull-down experiment. (E) A RIP assay was performed using anti-IGF2BP1 or anti-IgG antibodies in A375 cell lysates, and the enrichment of circSipa1l1 was measured by qRT‒PCR. (F, G) The levels of IGF2BP1 were detected by western blotting after circSipa1l1 silencing. (H, I) The co-localization of circSipa1l1 and IGF2BP1 was observed with FISH and IF assays. Scale bars: 100 μm. (J) Heatmap of the differentially expressed mRNAs among the si-circSipa1l1, si-IGF2BP1, and si-NC groups. (K) Venn diagram showing the overlap among the si-circSipa1l1, si-IGF2BP1, and IGF2BP1 RIP groups. (L, M) qRT‒PCR analysis of CHACI and ARHGDIB mRNA levels after circSipa1l1 or IGF2BP1 silencing in melanoma A375 cells. (N–P) The protein levels of ARHGDIB were measured by western blotting after silencing circSipa1l1 or IGF2BP1. (Q) A RIP assay was performed using anti-IGF2BP1 or anti-IgG antibodies in A375 cell lysates after transfection with OE-circSipa1l1 or empty vector, and the enrichment of ARHGDIB mRNA was measured by qRT‒PCR. (R) A375 cells were treated with actinomycin D at the indicated time points after silencing circSipa1l1 with/without co-transfection with OE-IGF2BP1, and the mRNA level of ARHGDIB was determined by qRT‒PCR. (S, T) The protein level of ARHGDIB was detected by western blotting after silencing circSipa1l1 with or without co-transfection with OE-IGF2BP1. *P < 0.05, **P < 0.01, ***P < 0.001 versus corresponding controls

As circRNAs have the protein-binding capacity [13], to further identify the proteins interacting with circSipa1l1, an RNA pull-down assay was performed in melanoma A375 cells. Then silver staining was employed to screen the potential circSipa1l1-interacting proteins, and we found a specific band at 50 kDa to 75 kDa (Fig. 3B). Meanwhile, mass spectrometry analysis implied that IGF2BP1 was the only protein with a molecular weight between 50 and 75 kDa in the top 10 quantified proteins (Fig. 3C). Combining the results of silver staining and mass spectrometry experiments, we speculate that IGF2BP1 could be the candidate binding partner of circSipa1l1. The following western blotting assay was conducted to validate this result, confirming that circSipa1l1 interacted with IGF2BP1 (Fig. 3D). The RIP assay also confirmed these findings (Fig. 3E).

We next detected the effect of circSipa1l1 silencing on IGF2BP1 expression. However, western blotting showed no altered abundance of IGF2BP1 protein expression (Fig. 3F-G). Intriguingly, FISH and IF assays revealed the co-localization of circSipa1l1 and IGF2BP1 in the cytoplasm of melanoma cells. Pearson’s correlation coefficient yielded a high correlation value (R = 0.83), confirming significant spatial co-localization (Fig. 3H). Moreover, line scan analysis along the designated region of interest further demonstrated highly overlapping fluorescence intensity profiles of circSipa1l1 and IGF2BP1, providing additional evidence for their co-localization (Fig. 3I). These results confirmed that circSipa1l1 and IGF2BP1 form an RNA‒protein complex in melanoma cells. Previous studies have confirmed that IGF2BP1, as an RNA-binding protein, can regulate mRNA stability [21]. To further verify the target genes of the circSipa1l1-IGF2BP1 complex, we analyzed mRNA expression profiles in circSipa1l1-silenced A375 cells and IGF2BP1-silenced A375 cells by RNA sequencing. The results showed that the levels of 126 mRNAs were significantly reduced in both the circSipa1l1-silenced and IGF2BP1-silenced groups (Fig. 3J). We intersected these 126 mRNAs with reported IGF2BP1 RIP-sequencing data, and two candidate IGF2BP1-binding mRNAs were obtained, namely ChaC glutathione specific gamma-glutamylcyclotransferase 1 (CHAC1) and ARHGDIB (Fig. 3K). A qRT‒PCR assay was used for further examination, and only ARHGDIB was downregulated after silencing circSipa1l1 or IGF2BP1 (Fig. 3L, M). Moreover, ARHGDIB protein expression also decreased after silencing circSipa1l1 or IGF2BP1 (Fig. 3N-P). Therefore, we selected ARHGDIB for further investigation.

A RIP assay was performed to explore whether IGF2BP1 could directly bind to ARHGDIB mRNA. We found significant ARHGDIB mRNA in the precipitate of the anti-IGF2BP1 group in comparison to that in the control group (Fig. 3Q). Moreover, circSipa1l1 overexpression markedly enhanced ARHGDIB mRNA interaction with IGF2BP1 (Fig. 3Q). The stability of ARHGDIB mRNA in melanoma cells with or without circSipa1l1 and IGF2BP1 interference was observed after actinomycin D treatment. The qRT‒PCR results indicated that the lower level of ARHGDIB mRNA induced by circSipa1l1 silencing was significantly attenuated after overexpressing IGF2BP1 (Fig. 3R). In addition, western blotting showed a similar result for ARHGDIB protein levels in A375 cells after silencing circSipa1l1 with or without overexpressing IGF2BP1 (Fig. 3S, T). Together, these results indicated that cicSipa1l1 recruits IGF2BP1 and stabilizes ARHGDIB mRNA in melanoma cells.

Silencing circSipa1l1 contributes to melanoma cell cycle arrest and differentiation by combining with the IGF2BP1-ARHGDIB axis

We then determined whether IGF2BP1 mediated circSipa1l1 silencing-induced cell differentiation. An IGF2BP1 overexpression plasmid (abbreviated as OE-IGF2BP1) was established. qRT-PCR (Supplementary Fig. 5A) and western blotting (Supplementary Fig. 5B, C) confirmed that both IGF2BP1 mRNA and protein levels were significantly increased in A375 cells transfected with OE-IGF2BP1 compared to the empty vector control. Subsequently, OE-IGF2BP1 was co-transfected with si-circSipa1l1. Then, the OE-IGF2BP1 was co-transfected with si-circSipa1l1. CCK-8 and colony formation assays indicated that OE-IGF2BP1 could significantly reverse the proliferation inhibition associated with circSipa1l1 silencing in A375 cells (Fig. 4A-C). Moreover, OE-IGF2BP1 also remarkably attenuated A375 cell cycle arrest at the G0/G1 phase induced by silencing circSipa1l1 (Fig. 4D, Supplementary Fig. 6A). Western blotting revealed that OE-IGF2BP1 remarkably restored the decreased levels of cyclin D1, CDK4, and CDK6 in circSipa1l1-silenced A375 cells (Fig. 4E, F). In addition, compared with the circSipa1l1-silenced group, the increased melanin content and tyrosinase activity in A375 cells were also repressed by OE-IGF2BP1 (Fig. 4G, H). Importantly, OE-IGF2BP1 significantly blocked the expression levels of key protein markers (MITF, TRP1, TRP2, and tyrosinase) in circSipa1l1-silenced A375 cells (Fig. 4I, J). Combined with these results, we confirmed that IGF2BP1 participates in circSipa1l1 silencing-induced melanoma cell differentiation.

Fig. 4.

IGF2BP1 participates in circSipa1l1 silencing-induced melanoma cell differentiation. Melanoma A375 cells treated with si-circSipa1l1 with or without co-overexpression with the empty vector or OE-IGF2BP1. (A) CCK-8 assays of A375 cell viability. (B, C) A375 cell survival was assessed using a colony formation assay. (D) A375 cell cycle distribution was assessed by flow cytometry. **P < 0.01, ***P < 0.001 versus the si-NC group; ^###P < 0.001 versus the si-circSipa1l1 group. (E, F) Cell cycle-related proteins were analyzed by western blotting. (G, H) Melanin content and tyrosinase activity were measured. (I, J) The melanin synthesis-related proteins were analyzed by western blotting. *P < 0.05, **P < 0.01, ***P < 0.001

Next, to verify the regulatory effect of ARHGDIB on silencing circSipa1l1-induced cell differentiation in melanoma cells, we constructed an ARHGDIB overexpression plasmid (abbreviated as OE-ARHGDIB). qRT-PCR (Supplementary Fig. 5D) and western blotting (Supplementary Fig. 5E, F) indicated that ARHGDIB expression was notably upregulated at both the mRNA and protein levels in A375 cells transfected with OE-ARHGDIB. CCK-8 and clonogenicity assays revealed that the si-circSipa1l1-induced reduction in cell proliferation was partially counteracted by OE-ARHGDIB (Fig. 5A-C). OE-ARHGDIB also markedly attenuated the A375 cell cycle arrest at the G0/G1 phase induced by silencing circSipa1l1 (Fig. 5D, Supplementary Fig. 6B) and remarkably restored the decreased levels of cyclin D1, CDK4, and CDK6 in circSipa1l1-silenced A375 cells (Fig. 5E, F). In addition, OE-ARHGDIB markedly suppressed circSipa1l1 silencing-induced increases in melanin content and tyrosinase activity in A375 cells (Fig. 5G, H). Western blotting showed that the upregulated levels of the cell differentiation-related proteins MITF, TRP1, TRP2, and tyrosinase in circSipa1l1-silenced A375 cells were partially reversed by OE-ARHGDIB (Fig. 5I, J). These findings indicated that ARHGDIB participates in circSipa1l1 silencing-induced melanoma cell differentiation.

Fig. 5.

ARHGDIB was involved in circSipa1l1 silencing-induced melanoma cell differentiation. Melanoma A375 cells treated with si-circSipa1l1 with or without co-overexpression with the empty vector or OE-ARHGDIB. (A) CCK-8 assays of A375 cell viability. (B, C) A375 cell survival was assessed by colony formation assay. (D) A375 cell cycle distribution was assessed by flow cytometry. *P < 0.05, ***P < 0.001 versus the si-NC group; ^###P < 0.001 versus the si-circSipa1l1 group. (E, F) Cell cycle-related proteins were analyzed by western blotting. (G, H) Melanin content and tyrosinase activity were measured. (I, J) The melanin synthesis-related proteins were analyzed by western blotting. *P < 0.05, **P < 0.01, ***P < 0.001

The circSipa1l1/IGF2BP1/ARHGDIB axis affects the melanoma cell cycle and differentiation through the ERK signaling pathway

The significantly differentially expressed mRNAs in circSipa1l1-silenced A375 cells (Fig. 3J) were analyzed by importance performance analysis (IPA) to explore the downstream pathways of the circSipa1l1/IGF2BP1/ARHGDIB axis. ARHGDI signaling was the top canonical pathway, consistent with the results of our previous study (Fig. 6A). Further cluster analysis of differential mRNAs showed a network direct to the ERK signaling pathway (Fig. 6B). Subsequently, western blotting was performed to measure ERK signaling molecule expression in circSipa1l1-silenced melanoma cells with/without OE-IGF2BP1 or OE-ARHGDIB. As expected, p-ERK/ERK expression was significantly downregulated after circSipa1l1 silencing compared to that in the control group, and OE-IGF2BP1 or OE-ARHGDIB remarkablyrescued the inhibitory effects of silencing circSipa1l1 (Fig. 6C-F).

Fig. 6.

The ERK signaling pathway downstream of circSipa1l1/IGF2BP1/ARHGDIB axis regulates melanoma cell cycle arrest and differentiation. (A, B) The top canonical pathways (A) and networks (B) were obtained by performing an IPA of the RNA sequencing data shown in Fig. 3J. (C, D) The levels of ARHGDIB and ERK signaling proteins were analyzed by western blotting in A375 cells after silencing circSipa1l1 with or without overexpressing IGF2BP1 (OE-IGF2BP1). (E, F) The levels of ERK signaling proteins were analyzed by western blotting of A375 cells after silencing circSipa1l1 with or without ARHGDIB overexpression (OE-ARHGDIB). (G, H) A375 cell survival was assessed by colony formation assay after the indicated treatments. (I) A375 cell cycle distribution was assessed by flow cytometry. **P < 0.01, ***P < 0.001 versus the si-NC group; ^#P < 0.001 versus the si-circSipa1l1 group. (J, K) Cell cycle-related proteins were analyzed by western blotting. (L, M) Melanin content and tyrosinase activity were measured in A375 cells after the indicated treatments. (N, O) The levels of melanin synthesis-related proteins in A375 cells after the indicated treatment were analyzed by western blotting. *P < 0.05, **P < 0.01, ***P < 0.001

To functionally validate the involvement of ERK signaling, we treated cells with the ERK activator tert-butylhydroquinone (TBHQ). Colony formation assays showed that the circSipa1l1 silencing-induced inhibition of A375 cell proliferation was markedly reversed by TBHQ (Fig. 6G, H). TBHQ also markedly attenuated A375 cell cycle arrest at the G0/G1 phase induced by silencing circSipa1l1 (Fig. 6I, Supplementary Fig. 6C) and markedly restored the decreased levels of cyclin D1, CDK4, and CDK6 in circSipa1l1-silenced A375 cells (Fig. 6J, K). In addition, TBHQ remarkably inhibited the circSipa1l1 silencing-induced increase in melanin content and tyrosinase activity in A375 cells (Fig. 6L, M). Western blotting revealed that the increased levels of MITF, TRP1, TRP2, and tyrosinase in circSipa1l1-silenced A375 cells were partially attenuated by TBHQ (Fig. 6N, O).

To further establish the position of ERK as a downstream effector of ARHGDIB, we employed the ERK inhibitor U0126 in ARHGDIB-overexpressing cells (OE-ARHGDIB). U0126 significantly suppressed the enhanced proliferation and clonogenicity conferred by ARHGDIB overexpression (Supplementary Fig. 7A-C). Cell cycle analysis demonstrated that U0126 reversed the reduction in the G0/G1 population induced by OE-ARHGDIB (Supplementary Fig. 7D), correlating with decreased levels of Cyclin D1, CDK4, and CDK6 (Supplementary Fig. 7E, F). Additionally, U0126 partially restored melanin production and tyrosinase activity, which were reduced by ARHGDIB overexpression (Supplementary Fig. 7G, H), and rescued the expression of differentiation markers MITF, TRP1, TRP2, and tyrosinase (Supplementary Fig. 7I, J). Taken together, these data suggested that the ERK signaling pathway downstream of the circSipa1l1/IGF2BP1/ARHGDIB axis regulates melanoma cell cycle arrest and differentiation.

Silencing circSipa1l1 inhibits melanoma cell growth in vivo

A B16 melanoma xenograft model was established to investigate whether circSipa1l1 affects tumor growth. When the tumor volume reached approximately 100 mm³, shRNA circSipa1l1 or shRNA negative control (shRNA-NC) lentivirus was injected in situ into the tumor. As shown in Supplementary Fig. 8A-D, the tumor weights, and volumes were significantly reduced without any body weight alteration in shRNA circSipa1l1-treated mice, consistent with our in vitro results. Moreover, IHC and western blotting showed that ARHGDIB and p-ERK/ERK levels were significantly downregulated in tumor tissues from shRNA circSipa1l1-treated mice (Supplementary Fig. 8E-H). The levels of cell cycle-related proteins Cyclin D1, CDK4, and CDK6 were also significantly decreased in tumor tissues from shRNA circSipa1l1-treated mice (Supplementary Fig. 8I, J). Furthermore, we confirmed that the melanin content and tyrosinase activity were significantly enhanced in tumor tissues from shRNA circSipa1l1-treated mice (Supplementary Fig. 8K, L). Meanwhile, MITF, TRP1, TRP2, and tyrosinase levels were remarkably upregulated in tumor tissues from shRNA circSipa1l1-treated mice (Supplementary Fig. 8M, N). These results indicated that silencing circSipa1l1 inhibits xenograft melanoma growth in vivo.

Furthermore, a xenograft mouse model was utilized to evaluate the effect of ARHGDIB overexpression on circSipa1l1 silencing-induced melanoma growth inhibition in vivo. A375 cells stably transfected with sh-circSipa1l1 or sh-NC plasmid were subcutaneously inoculated into nude mice, and intratumoral injection was used to infect the cells with the ARHGDIB overexpression lentivirus or relevant vector after tumor formation. Consistently, our data demonstrated a decrease in tumor weight and volume without any effect on body weight in shRNA circSipa1l1-treated mice, while these effects were partly reversed by OE-ARHGDIB (Fig. 7A-C). IHC staining and western blotting showed that the decreased levels of ARHGDIB and p-ERK/ERK in tumor tissues from shRNA circSipa1l1-treated mice were significantly reversed by OE-ARHGDIB (Fig. 7D-F, Supplementary Fig. 9A). Moreover, the downregulated levels of cyclin D1, CDK4, and CDK6 in tumor tissues from shRNA circSipa1l1-treated mice were remarkably recovered by OE-ARHGDIB (Fig. 7G, H). The increased melanin content and tyrosinase activity and the upregulated levels of MITF, TRP1, TRP2, and tyrosinase were all remarkably attenuated by OE-ARHGDIB when compared with the shRNA circSipa1l1-treated mice (Fig. 7I-L). In conclusion, our results demonstrated that ARHGDIB overexpression could recover circSipa1l1 silencing-induced melanoma growth inhibition and differentiation in vivo.

Fig. 7.

CircSipa1l1 silencing inhibits tumor growth in vivo. (A, B) Representative image of tumors and tumor weights after the indicated treatments. (C) Tumor volume and body weight were measured every 3 days. (D) The percentage of ARHGDIB-positive cells in tumor tissues was assessed by IHC. (E, F) The levels of ARHGDIB and ERK signaling proteins in tumor tissues were analyzed by western blotting. (G, H) The levels of cell cycle-related proteins in tumor tissues were analyzed by western blotting. (I, J) Melanin content and tyrosinase activity were measured in tumor tissues. (K, L) The levels of melanin synthesis-related proteins in tumor tissues were analyzed by western blotting. *P < 0.05, **P < 0.01, ***P < 0.001

Discussion

Recent studies shown that some novel therapies, such as immune checkpoint inhibitors targeting regulatory mechanisms or strategies modulating the tumor microenvironment, exhibits great potential for melanoma treatment [22, 23]. However, further randomized clinical data are still needed to validate their efficacy across primary, secondary, and tertiary treatment scenarios. Moreover, tumor-infiltrating lymphocyte therapy exhibit effectiveness in anti-programmed cell death-1(PD-1) refractory melanoma [24]. Nevertheless, the high cost and associated toxicity pose significant barriers, constraining its applicability in most of patients. Despite significant advancements achieved over the last decade, approximately half of individuals diagnosed with advanced melanoma die from the disease, highlighting the urgent need for novel therapeutic strategies [3].

Induction differentiation therapy, which induces malignant cells to redifferentiate into less aggressive or normal-like phenotypes while inhibiting proliferation, represents a promising alternative [25]. Melanomas originate from neural crest-derived melanocytes, suggesting they retain potential for redifferentiation [10]. Trans-differentiation of melanoma cells can be induced by specific chemical inducers, such as ATRA or PB-4 (as used in our study), which trigger cell cycle arrest and upregulate differentiation markers like tyrosinase [26]. Compared to other noncoding RNAs (ncRNAs), circRNAs possess greater stability, which grants them a distinct advantage as biomarkers or targets in disease treatment [11, 27]. Numerous circRNAs have been shown to be dysregulated in melanoma, but most studies focus on their roles in proliferation, metastasis, or drug resistance—far less is known about their regulation of melanoma differentiation [15, 28, 29]. Our findings revealed circSipa1l1 as a novel player in melanoma characterized by stability, a long half-life, and conservation. Notably, circSipa1l1 expression was upregulated in melanoma in vitro and in vivo in comparison with that in HaCaT cells and pericarcinomatous tissues. Further observations demonstrated that the silencing of circSipa1l1 induced cell cycle arrest and differentiation in both B16 and A375 melanoma cells, whereas overexpression of circSipa1l1 reversed these effects. In conclusion, circSipa1l1 may serve as an oncogenic factor in melanoma.

Emerging evidence strongly suggests that circRNAs may function as competitive endogenous RNAs (ceRNAs) in tumors, thereby participating in miRNA-mediated mRNA destabilization as endogenous miRNA “sponges,” ultimately regulating the expression of target mRNAs [20, 30]. AGO2 has been identified as a mediator of circular RNA-microRNA interactions, responsible for the miRNA-mediated degradation of target RNAs [31]^, [32, 33].

Given circSipa1l1’s cytoplasmic localization, we initially hypothesized it might function as a miRNA sponge. However, RIP assays showed no enrichment of AGO2 with circSipa1l1, ruling out this mechanism. In addition to ceRNA mechanisms, circRNAs can interact with RNA-binding proteins (RBPs), thereby regulating their availability within the cell and affecting the posttranscriptional fates of RBP-interacting mRNAs, such as stability or translation [13, 34, 35]. For example, circPABPN1 extensively binds to human antigen R (HuR), thereby preventing HuR from binding to polyadenylate-binding protein nuclear 1 (PABPN1) mRNA and subsequently reducing PABPN1 translation [36]. Moreover, circACTN4 recruits Y-Box binding protein 1 (YBX1), enhancing the interaction between Yes-associated protein 1 (YAP1) and β-catenin and activating the Hippo and Wnt signaling pathways [37]. This activation ultimately promotes the proliferation and metastasis of tumor cells in intrahepatic cholangiocarcinoma. Herein, our findings demonstrate that circSipa1l1 specifically binds to IGF2BP1, as confirmed by direct interaction assays and co-localization analysis in melanoma cells. Notably, circSipa1l1 silencing did not alter IGF2BP1 protein abundance, indicating it modulates IGF2BP1 function rather than expression.

IGF2BP1 has been documented as a multitarget mRNA binding protein that regulates the stability of target mRNAs at the posttranscriptional level [38]. IGF2BP1 exhibits high expression levels in various malignant tumors and is often considered an oncogene [21, 39]. Notably, IGF2BP1 has been identified as a potential prognostic indicator in melanoma since its overexpression in metastatic melanoma confers resistance to chemotherapeutic agents. IGF2BP1 depletion in melanoma cells suppresses cell proliferation, enhances apoptosis, and diminishes the tumorigenicity of vemurafenib-resistant melanoma [40, 41]. Specifically, IGF2BP1 can stabilize fermitin family member 2 (FERMT2), MITF2, and protein kinase C alpha-3 (PKCα3) mRNA in melanoma [42–44]. Our study provides the first evidence that IGF2BP1 plays a critical role in circSipa1l1-mediated melanoma progression. Upregulation of IGF2BP1 counteracts the effects of silencing circSipa1l1 on melanoma cell cycle arrest and differentiation, highlighting the crucial function of IGF2BP1 as a mediator of circSipa1l1 silencing effects in melanoma. Through RNA-seq and RIP-seq intersection, we identified ARHGDIB as the key downstream mRNA target of the circSipa1l1-IGF2BP1 complex. IGF2BP1 directly binds to ARHGDIB mRNA to enhance its stability, and circSipa1l1 strengthens this interaction.

ARHGDIB, also known as RhoGDI2, is a member of the RhoGDI family. Substantial evidence suggests that ARHGDIB is frequently overexpressed in various cancers and contributes to aggressive characteristics, such as motility, invasion, and metastasis, by disrupting the Rho GTPase signaling pathway [45–47]. Thus, ARHGDIB is a promising target for cancer treatment. In line with this, we revealed that the RHOGDI signaling pathway was the top-ranked canonical pathway in the IPA of circSipa1l1- and IGF2BP1-silenced melanoma cells. Our investigation also demonstrated that ARHGDIB counteracts the tumor-inhibitory effects of silencing circSipa1l1 in melanoma both in vivo and in vitro.

Moreover, we identified ERK as the downstream pathway of the circSipa1l1/IGF2BP1/ARHGDIB axis. ERK is a member of the MAPK signal transduction enzyme family, which plays a crucial role in regulating the cell cycle, proliferation, and differentiation [48–50]. Importantly, our current study reveals that circSipa1l1 silencing reduces p-ERK/ERK ratios, while IGF2BP1/ARHGDIB overexpression rescues this effect. ERK agonists (TBHQ) could abolish the effects of circSipa1l1 silencing on melanoma cell cycle arrest and differentiation. Conversely, inhibiting ERK with U0126 abolishes ARHGDIB-mediated proliferation and differentiation suppression. This finding further supports the notion that the circSipa1l1/IGF2BP1/ARHGDIB axis could modulate melanoma progression through the ERK pathway. While our present study did not directly assess the impact of ATRA on ERK phosphorylation levels in melanoma cells, ATRA-induced differentiation was observed to be concomitant with the downregulation of circSipa1l1 and a corresponding decrease in p-ERK1/2 levels (Figs. 1 and 6). Interestingly, Nefedova et al. reported that ATRA directly activates ERK1/2 phosphorylation in myeloid-derived suppressor cells (MDSCs), and that this activation is essential for both ATRA-induced upregulation of GSS and MDSC differentiation [51]. These collective findings suggest that the role of ATRA in modulating ERK signaling is highly context-dependent. In melanoma, the anti-proliferative and pro-differentiative effects of ATRA may require suppression of the oncogenic ERK pathway, whereas in MDSCs, ERK activation may be necessary for immunomodulatory differentiation. Moreover, it has been reported that the expression levels of phosphorylated ERK and P38 significantly decrease in ARHGDIB-overexpressing SNU-484 gastric cancer cells, which is crucial for ARHGDIB-induced cisplatin resistance [52]. However, it is important to note that in response to cisplatin, while the activation of p38 only promotes cell death, the induction of ERK can play dual roles, either promoting survival or cell death [53]. Therefore, the increased expression of ERK in ARHGDIB-upregulated gastric cancer cells may involve negative feedback regulation.

Some limitations of our study should be acknowledged. siRNA technology is now well known to generate an enormous amount of off-targets, given the fact that every siRNA molecule acts as an artificial miRNA [54, 55]. To mitigate this risk, we employed three independent siRNAs specifically targeting the back-splicing junction of circSipa1l1—thus avoiding interference with linear Sipa1l1 mRNA. All three siRNAs consistently induced the same phenotypic effects (Fig. 2A-D). We also included scrambled siRNA controls (si-NC) to account for non-specific effects. Importantly, the core findings from siRNA-based knockdown were corroborated using shRNA-mediated silencing in in vivo xenograft models and were successfully reversed through IGF2BP1 or ARHGDIB overexpression in rescue experiments. These complementary approaches strengthen the specificity of our conclusions and reduce the likelihood that off-target effects significantly influenced our results. Moreover, the utilization of mass spectrometry in our study revealed that circSipa1l1 might bind to various other proteins, implying that these proteins could also mediate the role of circSipa1l1 in melanoma. In addition, it is likely that ARHGDIB is not the only target gene stabilized by the circSipa1l1-IGF2BP1 complex, and further investigations are needed to elucidate these mechanisms and deepen our understanding of circSipa1l1’s functions. Furthermore, larger cohorts are needed to verify circSipa1l1’s prognostic value and association with differentiation therapy responses.

Conclusions

In summary, our study uncovers a novel regulatory axis—circSipa1l1/IGF2BP1/ARHGDIB/ERK—that controls melanoma cell differentiation and proliferation. Silencing circSipa1l1 disrupts IGF2BP1-mediated ARHGDIB mRNA stabilization, inhibits ERK signaling, and induces differentiation [Fig. 8]. These findings provide new insights into circRNA-mediated melanoma progression and identify circSipa1l1 as a potential target for melanoma differentiation therapy.

Fig. 8.

Schematic of the CircSipa1l1–IGF2BP1–ARHGDIB–ERK–differentiation pathway. Silencing of circSipa1l1 inhibits the binding between IGF2BP1 and ARHGDIB mRNA and destabilizes ARHGDIB mRNA, thereby inducing cell differentiation via the ERK signaling pathway in melanoma

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ATRA

All-trans-retinoic acid

ARHGDIB

Rho GDP-dissociation inhibitor 2

cDNA

Complementary DNA

ceRNAs

Competitive endogenous RNAs

CDK4

Cyclin-dependent kinase 4

CDK6

Cyclin-dependent kinase 6

CHAC1

ChaC glutathione specific gamma-glutamylcyclotransferase 1

FERMT2

Fermitin family member 2

FISH

Fluorescence in situ hybridization

gDNA

Genomic DNA

HuR

Human antigen R

H&E

Hematoxylin and eosin

IGF2BP1

Insulin-like growth factor 2 mRNA binding protein 1

IHC

Immunohistochemical

IF

Immunofluorescence

IPA

Importance performance analysis

MITF

Microphthalmia-associated transcription factor

OE

Overexpression

PABPN1

Polyadenylate-binding protein nuclear 1

PD-1

Anti-programmed cell death-1

PKCα3

Protein Kinase C alpha-3

qRT-PCR

RNA sequencing and quantitative RT‒PCR

RBPs

RNA-binding proteins

RIP

RNA immunoprecipitation

RNA-seq

RNA sequencing

siRNAs

Small-interfering RNAs

shRNAs

Small hairpin RNAs

TYR

Tyrosinase

TRP1

Tyrosinase-related protein 1

TRP2

Tyrosinase-related protein 2

YAP1

Yes-associated protein 1

YBX1

Y-box binding protein 1

Author contributions

BL and DL supervised the project. BL, QZ, JL and DL designed the research. BL, LL DS and DL wrote the article. BL, LL, DS and XW performed the experiments. TM, XX, JZ, XH, GW, TA and QJ analyzed the data. SB, ZP and JL helped to revise the manuscript. All authors read and approved the final manuscript.

Funding

This research was supported by the the Shandong Provincial Natural Science Foundation (ZR2021QH323 to BL), National Natural Science Foundation of China (82073313 to DL), the joint project of State Administration of Traditional Chinese Medicine and Health Commission of Shandong Provincial (GZY-KJS-SD-2023-094 to DL), the Traditional Chinese Medicine Special Project of Binzhou Medical University Affiliated Traditional Chinese Medicine Hospital (2023ZYZX02 to DL), 2023 Qilu Biancang Traditional Chinese Medicine Talent Cultivation Project (to DL), and the Introduction and Cultivation Project for Young Creative Talents of Higher Education of Shandong Province (to GW).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethical approval and consent to participate

The collection of human tissue microarrays was approved by the Zhuoli Biotechnology Company Biomedical Ethics Committee (ZLL-15-01). All the guidelines for the care and use of laboratory animals were followed, and the study was approved by the Animal Ethics and Experimental Safety Committee of Binzhou Medical University (BZMU-2021-221; BZMU-2022-252).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.