Identification and characterization of circSMARCC1 in PCa

Arraystar Human CircRNA Array v2 (Kangcheng Biotech, Shanghai, China) was applied to analyze the circRNA microarray [18]. To explore the circRNA expression profile in PCa, we performed RNA sequencing analyses of ribosomal RNA-depleted total RNA from 4 pairs of plasma samples of patients with benign prostatic hyperplasia (BPH) or PCa. The cluster heat map revealed a more than 1.5 folds of change in differentially expressed circRNAs (Fig. S1A). The scatter plots demonstrated 98 up-regulated and 40 down-regulated circRNAs in the plasma samples of PCa patients compared to the control individuals (Fig. S1B). Among them, the top 20 dysregulated circRNAs were listed in Fig. 1A. We focused on the top 5 exonic circRNAs with up-regulated expression, namely: circMAPKBP1, circRAN, circAHI1, circRBM4 and circSMARCC1. Then, further detected the relative expression levels of 5 circRNAs by RT-PCR in 39 pairs plasma samples of PCa and BPH. The results showed that compared with the plasma derived from patients with BPH, the expression levels of circRAN and circAHI1 were not significantly different, while the expression levels of circMAPKBP1, circRBM4 and circSMARCC1 were up-regulated, with circSMARCC1 being the most significantly up-regulated (Fig. S1C). Therefore, we chose circSMARCC1 as the research target.

Fig. 1

circSMARCC1 validated and characterized in PCa cells. A The cluster heat map demonstrated the top 20 circRNAs differentially expressed in four pairs of plasma samples from PCa and BPH patients. B, C Schematic representation of the formation of circSMARCC1 by cyclization of exons 14, 15, and 16 of the SMARCC1 gene. The back splice junction sequence and RT-PCR product of circSMARCC1 were verified by Sanger sequencing and agarose gel electrophoresis, respectively. D The localization of circSMARCC1 observed in PCa tissues (scale bar, 20 μm) and cells (scale bar, 10 μm) detected by FISH. Nuclei were stained with DAPI. E Analysis of the cellular localization of circSMARCC1 by nuclear-cytoplasmic fractionation experiment. GAPDH was used as a control for cytoplasmic proteins and U6 was used as a nuclear control. F qRT-PCR was performed to determine the abundance of circSMARCC1 and SMARCC1 mRNA in PCa cells treated with RNase R at the indicated time points. G RNA expression of circSMARCC1 and SMARCC1 in PCa cells analyzed by qRT-PCR after treatment with actinomycin D for 4 h,8 h,12 h and 24 h.The data are presented as mean ± SD.**p < 0.01, ***p < 0.001, ****p < 0.0001

CircSMARCC1 (hsa_circ_0001296), which was formed from exons 14, 15, and 16 of the coding gene SMARCC1 by back-splicing on the basis of the annotation of circBase (http://www.circbase.org/).The melting curve of circSMARCC1 amplified product using divergent primers showed a single peak the same as GAPDH (Fig. S1D). Its back-splicing junction was validated by Sanger sequencing and the presence of circSMARCC1 was demonstrated by RT-PCR (Fig. 1B). We designed the divergent and convergent primers to amplify the circSMARCC1 circular transcripts and SMARCC1 linear transcripts. PCR results showed that circSMARCC1 was only detected in cDNA, thus ruling out the existence of circSMARCC1 in gDNA, whereas the convergent primers amplified SMARCC1 from both cDNA and gDNA (Fig. 1C). To clarify the subcellular localization of circSMARCC1, we performed nuclear-cytoplasmic fractionation and FISH experiments in PCa cells and tissues. It was found that circSMARCC1 was mainly present in the cytoplasm of PCa cells (Fig. 1D). Additionally, nuclear-cytoplasmic fractionation assays showed that circSMARCC1 was mainly localized in the cytoplasm of PCa cells by qRT-PCR (Fig. 1E). The Actinomycin D assay demonstrated that circSMARCC1 was more stable compared with linear SMARCC1 (Fig. 1F). Moreover, we found that circSMARCC1 was more resistant to RNase R digestion than linear SMARCC1 (Fig. 1G). These results suggested that circSMARCC1 exists as a circular form and is highly stable. Then, we predicted the function of the circSMARCC1 encoded protein by circRNADb (http://reprod.njmu.edu.cn/cgi-bin/circrnadb/circRNADb.php). The results indicated that no open reading frame was found, which means that the possibility of circSMARCC1 in encoding protein is low (Fig. S1E), implying that circSMARCC1 is not encoded protein.

CircSMARCC1 is up-regulated in PCa and associated with clinical characteristics

To evaluate the expression and clinical value of circSMARCC1, qRT-PCR was carried out to detect the expression level of circSMARCC1 in the plasma of 39 pairs of age-matched patients with PCa and BPH. The results indicated that the expression of circSMARCC1 was significantly elevated in the plasma of PCa patients, which was consistent with the results of RNA-seq (Fig. 2A). Likewise, we validated the expression levels of circSMARCC1 in 22 pairs para-cancerous and PCa tissues. The results showed that circSMARCC1 was also significantly up-regulated in PCa tissues (Fig. 2B). Moreover, the diagnostic value of circSMARCC1 for PCa screening was further assessed utilizing the receiver operating characteristic (ROC) curve. Based on ROC curve analysis, the area under the curve (AUC) for circSMARCC1 was 0.713, with a specificity and sensitivity of 74.4 and 66.7%, respectively, and a cut-off value of 5.937 (Fig. 2C). It suggested that circSMARCC1 has the ability to identify PCa, but not as well as prostate specific antigen (PSA). Subsequently, we assessed the relationship between circSMARCC1 expression and clinicopathological features. The results showed that circSMARCC1 levels were positively correlated with Gleason score and T-stage in PCa patients, but not with age or PSA (Table 1).

Fig. 2

circSMARCC1 up-regulated in PCa and correlated with pathological parameters. A, B The relative expression of circSMARCC1 in plasma (n = 39) and tissue (n = 22) specimens from PCa patients and BPH patients was detected by qRT-PCR. C ROC curve analysis of the diagnostic value of circSMARCC1 and PSA for PCa. D The relative expression of circSMARCC1 in PCa cell lines by qRT-PCR. E The correlation between the expression of circSMARCC1 and Gleason score (GS) in PCa tissues through FISH experiment (scale bar, 20 μm). The data are presented as the mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Table 1 Correlation between circSMARCC1 expression and clinicopathological features in plasma from patients with BPH (n = 39) and PCa (n = 39)

Furthermore, the expression of circSMARCC1 was validated in PCa tissues and cell lines. We analyzed the endogenous expression of circSMARCC1 by qRT-PCR. The results demonstrated that the expression of circSMARCC1 is higher in five PCa cell lines (PC-3, DU145, 22Rv1, C4–2 and LNCaP) than in the normal prostate epithelial cell line (Fig. 2D). Meanwhile, we examined the expression of circSMARCC1 in PCa tissues by RNA-FISH. The results showed that circSMARCC1 was significantly up-regulated in PCa tissues, and down-regulated in normal adjacent tissues (Fig. 2E). More importantly, it was clear that as the Gleason score of PCa increased, the expression of circSMARCC1 was also significantly up-regulated. In addition, using publicly available sequencing data for circRNAs [21], we validated the prognostic relationship between circSMARCC1 and biochemical recurrence (BCR) in PCa patients. We identified a trend that PCa patients with high circSMARCC1 expression were more likely to experience BCR early after radical prostatectomy, implying that high circSMARCC1 expression indicates a poorer prognosis, although we did not observe a statistically significant difference (Fig. S1F). These results suggested that circSMARCC1 might act as a tumor promoter, and high expression of circSMARCC1 could be a predictor of poor prognosis in PCa patients.

CircSMARCC1 accelerates PCa cell proliferation, migration, and invasion in vitro

As C4–2 cells exhibited the highest levels of circSMARCC1, and DU145 and PC-3 cells expressed lower levels of circSMARCC1 compared to other cells, we selected DU145, PC-3, and C4–2 cells for the subsequent studies. To clarify the potential role of circSMARCC1 in promoting PCa progression, three siRNAs (si-#01,02,03) targeting circSMARCC1 were constructed to silence its expression. The si-#03 exhibited the highest silencing efficiency measured by qRT-PCR and was chosen for the following experiments (Fig. 3A). We up-regulated circSMARCC1 expression in DU145 and PC-3 cells and down-regulated circSMARCC1 in C4–2 cells using a lentiviral vector and confirmed no change of SMARCC1 expression in parental cells (Fig. 3B). The colony formation, EdU, and CCK-8 assays were used to detect cell proliferation and viability. The results revealed that knockdown of circSMARCC1 significantly inhibited the proliferative capacity of PCa cells, while overexpression of circSMARCC1 resulted in increased cell viability (Fig. 3C-E). Furthermore, flow cytometry analysis was performed to confirm whether changes in proliferation were attributed to alterations in the cell cycle profile. The results showed that overexpression of circSMARCC1 in PC-3 and DU145 cells accelerated cell cycle progression at the S phase, whereas knockdown of circSMARCC1 in C4–2 cells induced G1 phase arrest (Fig. 3G). In addition, Western blot analysis showed that down-regulation of circSMARCC1 decreased CDK2 expression and increased p21 Waf1/Clip1 levels, while up-regulation of circSMARCC1 showed the opposite results (Fig. 3F). These results suggested that circSMARCC1 enhances cell growth, at least partially by inducing the G1/S transition in PCa cells. Similarly, we found that down-regulation of circSMARCC1 suppressed the migratory and invasive capacity of PCa cells through transwell and wound healing assays, whereas up-regulation of circSMARCC1 exhibited the opposite effect (Fig. 3H-I). Meanwhile, Western blot analysis showed that overexpression of circSMARCC1 distinctly reduced E-cadherin expression and increased Vimentin levels, while silencing of circSMARCC1 revealed the opposite effect (Fig. 3F). These findings indicated that circSMARCC1 has an oncogenic role in PCa cells.

Fig. 3

CircSMARCC1 promotes proliferation, migration and invasion of PCa cells. A three siRNAs (si-#01,02,03) targeting circSMARCC1 were constructed to silence its expression and confirmed by qRT-PCR. B The relative expression of circSMARCC1 and SMARCC1 in PCa cells transfected with circSMARCC1 overexpressing or knockdown lentivirus by qRT-PCR. C-E Assessment of cell proliferation capacity by colony formation, EdU assay (scale bar, 100 μm) and CCK-8 assay. F Western blot analysis evaluated expression of cell cycle-associated proteins and EMT biomarkers following overexpression or knockdown of circSMARCC1. G Flow cytometric analysis of changes in the cell cycle profile of PCa cells stably transfected with circSMARCC1. H, I Transwell assays and wound healing assays assessed the migration and invasion abilities of PCa cells stably transfected with circSMARCC1 (Scalebar, 50 μm). The data are presented as the mean ± SD, **p < 0.01, ***p < 0.001, ****p < 0.0001

CircSMARCC1 directly binds to miR-1322 and suppresses miR-1322 activity

Endogenous circRNAs have been found to act as microRNA (miRNA) sponges in human cancers. As circSMARCC1 is mainly localized in the cytoplasm, we hypothesized that circSMARCC1 could regulate the expression of downstream molecules by binding to specific miRNAs. To investigate whether circSMARCC1 could sponge miRNA in PCa cells, we selected the 9 top potential miRNAs with a score ≥ 90 predicted by the CircInteractome database (Supplementary Table S4). To confirm the interaction between circSMARCC1 and the candidate miRNAs, dual-luciferase reporter assays were performed to detect the binding between circSMARCC1 and miRNAs. The wild-type and mutant dual-luciferase reporter plasmids of circSMARCC1 were constructed (Fig. 4A). The luciferase-circSMARCC1 reporters were transfected into HEK-293 T cells along with miRNA mimics or negative controls, and the data indicated that miR-1322 significantly reduced luciferase activity (Fig. 4B). Subsequently, RNA pull-down experiments were performed using biotinylated circSMARCC1 probes. The results demonstrated that miR-1322 was substantially pulled down by the biotin-coupled circSMARCC1 probe rather than the oligo probe in DU145 cells with circSMARCC1 overexpression, suggesting that circSMARCC1 might directly binds to miR-1322 (Fig. 4C). Besides, qRT-PCR assays were performed to investigate the effect of circSMARCC1 on miR-1322 expression. The results revealed that overexpression or knockdown of circSMARCC1 resulted in down-regulation or up-regulation of miR-1322 in PCa cells (Fig. 4D), while the expression of circSMARCC1 after transfection with miR-1322 mimics or inhibitors shows no significant changes (Fig. S2A). Furthermore, luciferase reporter assays were conducted in DU145 cells that were transfected with the wild-type and mutant dual-luciferase reporter plasmids of circSMARCC1. The luciferase activity was obviously decreased in cells co-transfected with the miR-1322 mimics and the circSMARCC1-Wt luciferase reporter when compared with the mutants (Fig. 4E). In addition, RNA-FISH was utilized to confirm the subcellular co-localization of circSMARCC1 and miR-1322 in PCa cells. We found that circSMARCC1 co-localized with miR-1322 in the cytoplasm (Fig. 4F). These results demonstrated that circSMARCC1 could directly bind to miR-1322 in PCa cells.

Fig. 4

CircSMARCC1 acts as a sponge for miR-1322 and miR-1322 reverses the oncogenic effects of circSMARCC1 on proliferation, invasion and migration in PCa cells. A Schematic diagram of circSMARCC1 luciferase reporter vectors carrying wild-type (Wt) or mutant (Mut) miR-1322 binding sites. B Luciferase reporter assay to analyze the effects of 9 candidate miRNAs on the luciferase activity of circSMARCC1. C The RNA pull-down assay performed in DU145 cells using circSMARCC1 and negative control probes. D The relative expression of miR-1322 in PCa cells after transfection of circSMARCC1 was detected by qRT-PCR. E The relative luciferase activities measured in 293 T cells co-transfected with circSMARCC1-Wt or circSMARCC1-Mut and miR-1322 mimics or miR-nc by luciferase reporter assay. F The co-localization of circSMARCC1 and miR-1322 observed using RNA-FISH in DU145 cells (scale bar, 5 μm). The nuclei were stained with DAPI. G-I The viability of PCa cells in miR-1322 rescue experiments was analyzed by colony formation assay, EdU assay (scale bar, 100 μm) and CCK8 assay, respectively. J, K The migration and invasion capacity of PCa cells in the miR-1322 rescue experiment was analyzed by transwell assays (scale bar, 50 μm) and wound healing (scale bar, 100 μm) assays. The data are presented as the mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

MiR-1322 reverses the tumor-promoting effect of circSMARCC1 in PCa cells

To determine whether circSMARCC1-mediated miR-1322 affected the cell viability as well as migration and invasion of PCa cells, several rescue experiments by co-transfection of miR-1322 mimics or miR-1322 inhibitors with lv-circSMARCC1 or sh-circSMARCC1 were performed. The results showed that ectopic expression of miR-1322 significantly attenuated the proliferation, migration and invasion promotion induced by circSMARCC1 up-regulation, while miR-1322 inhibitor counteracted the inhibitory effect of circSMARCC1 down-regulation in proliferation, migration and invasion by colony formation assays (Fig. 4G), EdU analysis (Fig. 4H), CCK8 experiments (Fig. 4I), transwell assays (Fig. 4J) and wound healing assays (Fig. 4K). Overall, these experiments suggested that circSMARCC1 serves as a sponge for miR-1322.

CCL20 is a direct target of miR-1322 and activates PI3K-Akt pathway through circSMARCC1/miR-1322/CCL20 axis

To further explore the underlying mechanism, we performed RNA-seq in DU145-lv-circSMARCC1 and DU145-vector cells and found that 151 genes were up-regulated and 209 genes were down-regulated in DU145-lv-circSMARCC1 cells compared to DU145-vector cells (fold change > 2 and p < 0.05). The cluster heat map and volcano plots of the differential genes were shown in Fig. 5A, B. Subsequently, we conducted bioinformatics analyses using Targetscan (http://www.targetscan.org), miRDB (http://www.mirdb.org/) and miRDIP (https://ophid.utoronto.ca/mirDIP/index.jsp) to predict possible downstream targets for miR-1322 binding. Combined with the results of RNA sequencing, the data showed that 9 molecules containing conserved target sites of miR-1322 may serve as downstream targets of mir-1322 (Fig. 5C). Next, we used qRT-PCR to verify the 9 candidate molecules. The results showed that CCL20, but not other candidate targets, was up-regulated in DU145-lv-circSMARCC1 cells and down-regulated in C4–2-sh-circSMARCC1 cells (Fig. 5D). Further, we successfully constructed miR-1322-overexpressing and miR-1322-knockdown cells transfected with miR-1322 mimics or miR-1322 inhibitors (Fig. S2B). As shown in Fig. 5E-F, CCL20 expression was significantly down-regulated or up-regulated at both mRNA and protein levels in PCa cells transfected with miR-1322 mimics or miR-1322 inhibitors. Furthermore, wild-type and mutant dual luciferase reporter plasmids of CCL20 were constructed (Fig. 5G). Using a dual luciferase reporter assay, we found that transfection of miR-1322 mimics can significantly reduce the activity of the wild-type luciferase reporter gene, but not the mutant of CCL20 (Fig. 5H). These data suggested that miR-1322 directly targets CCL20.

Fig. 5

CCL20 is a direct target of miR-1322 and activates PI3K-Akt pathway through circSMARCC1/miR-1322/CCL20 axis. A, B The heatmap and volcano plot of differentially expressed mRNAs in DU145 cells transfected with vector or circSMARCCC1. Each sample was mixed with three replicates. C Venn diagram illustrating the overlapping of miR-1322 targeting mRNAs and up-regulated mRNAs in DU145 cells. D qRT-PCR was performed to detect the mRNA expression of 9 candidate molecules (MS4A1, CBFB, ARPP21, PHF6, FTO, CCL20, MED7, CD40 and SLC25A15). E, F The expression of CCL20 in PCa cells transfected with miR-1322 mimics or miR-1322 inhibitors was detected by qRT-PCR and Western blotting. G Schematic illustration of CCL20 wide type (Wt) and mutant (Mut) luciferase reporter vectors were shown. H The relative luciferase activities measured in 293 T cells co-transfected with CCL20-Wt or CCL20-Mut and miR-1322 mimics or miR-nc by luciferase reporter assay. I IF analysis detected the protein expression of CCL20 in PCa cells transfected or co-transfected with lv-circSMARCC1, sh-circSMARCC1, miR-1322 mimics or inhibitors (scale bar, 50 μm). J The relative protein levels of CCL20, EMT biomarkers and cell cycle-related molecules detected by Western blotting in miR-1322 rescue experiments. K IHC detection of CCL20 expression in PCa tissues and para-cancerous tissues (scale bar, 100 μm and 20 μm). L the Secretion of CCL20 in the supernatants of circSMARCC1-overexpressing DU145 and PC-3 cells cultured in RPMI-1640 medium or 10% FBS medium was measured by ELISA. M, N The CCL20 recombinant and CCL20 neutralizing antibodies were applied to transwell assays and CCK8 assays to explore the effects of CCL20 on PCa cell proliferation, migration and invasion. O, P KEGG and GSEA analyses of DEGs in DU145 cells. Q The relative protein levels of CCL20 and AKT pathway-related molecules measured by Western blotting in miR-1322 rescue experiments. The data are presented as the mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

To explore whether circSMARCC1/miR-1322 affects CCL20 expression, immunofluorescence (IF) and Western blot analysis were performed on lv-circSMARCC1 and sh-circSMARCC1 tumor cells. The results showed that lv-circSMARCC1 remarkably increased CCL20 expression, while sh-circSMARCC1 distinctly decreased CCL20 levels. Furthermore, miR-1322 mimics or inhibitors could reverse the increase or decrease of CCL20 induced by circSMARCC1 overexpression or knockdown (Fig. 5I-J). Importantly, we further found that CCL20 was up-regulated in PCa tissues compared with para-cancer tissues via IHC assay (Fig. 5K). Since CCL20 is a secreted protein, we also detected a significant increase in CCL20 secretion in the supernatants of DU145 and PC-3 cells overexpressing circSMARCC1 by ELISA, both in RPMI-1640 medium and in 10% FBS medium (Fig. 5L). These data suggested that CCL20 is a direct downstream of miR-1322.

As reported, CCL20 induces EMT in ovarian cancer cells and contributes to tumor progression [22]. We further assessed whether CCL20 could enhance the proliferative and metastatic capacity of PCa cells. CCL20 was found to be significantly up-regulated in PCa cells (Fig. S2C). We designed three small interfering RNAs (siRNAs) targeting CCL20 (si-#01, 02, 03), and found that si-#01 had the highest knockdown efficiency and used it for follow-up studies (Fig. S2D). The CCK8 experiments indicated that knockdown of CCL20 significantly down-regulated the proliferation ability of cells (Fig. S2E). Meanwhile, transwell assay showed that downregulation of CCL20 significantly reduced the number of migrating PCa cells (Fig. S2F). We then applied CCL20 recombinant protein and CCL20 neutralizing antibody to transwell assays and CCK8 assay. The results showed that CCL20 neutralizing antibody impaired the metastasis and proliferation caused by overexpression of circSMARCC1, and CCL20 recombinant protein reversed the reduction in metastasis and proliferation caused by knockdown of circSMARCC1(Fig. 5M, N). Furthermore, we interpreted those events in terms of changes in cell cycle proteins and EMT-related proteins by Western blotting. It was found that up-regulation of circSMARCC1 increased Vimentin and CDK2 expression and decreased E-cadherin and p21 Waf1/Clip1 expression, while down-regulation of circSMARCC1 had the opposite effects. Importantly, these effects could be abolished by miR-1322 mimics or inhibitors, respectively (Fig. 5J).

In addition, we examined the downstream signal pathways to which the miR-1322/CCL20 axis might transmit signals. We performed pathway analysis (KEGG and GSEA) based on RNA-seq profiles and found that overexpression of circSMARCC1 was positively correlated with the PI3K-Akt signaling pathway (Fig. 5O, P). To confirm this result, further studies showed that overexpression of circSMARCC1 increased the expression of CCL20, P-Akt473 and P-Akt308, while knockdown of circSMARCC1 exerted the opposite effect. Similarly, these effects could be eliminated by miR-1322 mimics or inhibitors (Fig. 5Q). Furthermore, western blot analysis showed that knocking down CCL20 down-regulated the expression of P-Akt473 and P-Akt308 (Fig. S2G). Finally, to further confirm whether the PI3K-Akt was required for circSMARCC1-mediated promotion of PCa progression, we performed rescue experiments by using LY294002, a PI3K-Akt inhibitor. The results showed that LY294002 could eliminate the increased cell proliferation (Fig. S2H, I) and migration (Fig. S2J) due to overexpression of circSMARCC1. In addition, Western blotting analysis was conducted to evaluate the effect of treatment with LY294002 on the levels of related proteins. Our results demonstrated that LY294002 partially restored p21 Waf1/Clip1 and E-cadherin expression while decreased Vimentin and CDK2 levels in PCa cells with ectopic expression of circSMARCC1 (Fig. S2K). These results suggested that circSMARCC1 functions as sponge of miR-1322 to promote PCa progression via activating Akt pathway.

CircSMARCC1 promotes PCa cells growth and metastasis in vivo

To investigate the biological function of circSMARCC1 in vivo, we constructed a xenograft tumor model by injecting circSMARCC1-overexpressing or vector DU145 cells into the axilla of BALB/c mice (Fig. 6A). The results showed heavier tumors and faster tumor growth in the overexpression group compared to the vector group (Fig. 6B, C). To study the effect of metastasis, a metastatic tumor model was generated by tail vein injection of nude mice using circSMARCC1-overexpressing or vector PC-3 cells. The PC-3 cells were pre-labelled with the luciferase gene to facilitate observation of tumor metastases in vivo by bioluminescence. We observed that mice in the overexpression group developed more lung and abdominal metastases compared to the vector group (Fig. 6D and S2L). H&E staining shows the pathological features of subcutaneous tumor tissue and isolated pulmonary metastases foci (Fig. 6E, F). Besides, RNA-FISH experiments in tumors showed a more extensive fluorescent area in the xenograft tumor tissue of lv-circSMARCC1, confirming that circSMARCC1 expression was indeed up-regulated (Fig. 6G). The up-regulation of circSMARCC1 expression was accompanied by an increase in the proportion of proliferating cells (Ki67+) via IHC experiments (Fig. 6H). These results suggested that circSMARCC1 promotes the growth of PCa in vivo. Next, IHC results showed a significant up-regulation of CCL20 expression in the lv-circSMARCC1 xenograft tumor group, which is consistent with the results we observed in human PCa tissues (Fig. 6I). In addition, we examined the expression of CD31 in tumor tissue, and found that CD31 expression was up-regulated in the lv-circSMARCC1 group, indicating active tumor angiogenesis (Fig. S2M). These findings demonstrated that circSMARCC1 promoted PCa growth and metastasis in vivo.

Fig. 6

circSMARCC1 promotes PCa tumorigenesis and metastasis in vivo. A Images of subcutaneous xenograft tumors derived from DU145 cells transfected with vector and circSMARCC1. B Tumor weight comparison in lv-circ group and vector group. C Tumor volumes measured once a week and the growth curves were drawn. D Fluorescence image of metastatic transplanted tumor from PC-3 cells transfected with vector or circSMARCC1. E H&E staining of subcutaneous xenograft tumors in mice (scale bar, 100 μm and 20 μm). F H&E staining of metastatic lung tumors in mice (scale bar, 100 μm and 20 μm). G RNA-FISH experiment showed the fluorescence area in xenograft tumor tissues of vector and lv-circSMARCC1 groups (scale bar, 100 μm and 20 μm). H, I The expression of Ki-67 and CCL20 in xenograft tumors detected by IHC (scale bar, 100 μm and 20 μm). *p < 0.05, ***p < 0.001, ****p < 0.0001

Overexpression of circSMARCC1 increases the number of M2 macrophages in human PCa tissues and mouse allograft tumors

Chemokines regulate tumor progression in the TME through chemotaxis of macrophages or immune cells [23]. To further assess the role of CCL20, we analyzed the differential function of CCL20 as well as CCL20 co-expressed genes in PCa from TCGA data using Gene Ontology (GO) analysis. The results showed that the most differentially expressed genes were associated with immune system processes (Fig. S3A). Next, the correlation between CCL20 and 28 tumor-infiltrating lymphocytes types in the TME of PCa was analyzed using the TISIDB database (http://cis.hku.hk/TISIDB/). The results showed a significant positive correlation between CCL20 and macrophages (Fig. S3B). Similarly, we found that CCR6, the specific receptor for CCL20, also had a significant positive correlation with macrophages and was more closely related to M2-type macrophages among them (Fig. S3C). These evidences support our exploration of the relationship between the CCL20-CCR6 axis and macrophages in the TME of PCa. In addition, several studies have shown that CCL20 regulates macrophage recruitment to drive tumor growth in colon cancer [24] and to promote breast tumorigenesis [25]. Kfoury et al. described the PCa bone metastases and revealed an immune mechanism of M2 macrophage infiltration mediated by the CCL20-CCR6 axis [26]. Accordingly, we tested circSMARCC1 expression and macrophage infiltration by IHC in human PCa tissues and adjacent tissues. Interestingly, the number of TAMs (CD68+, CD163+ and CD206+) in PCa tissues with circSMARCC1 up-regulation was significantly higher than that in adjacent tissues with circSMARCC1 down-regulation (Fig. 7A). Furthermore, we obtained consistent results in BALB/c mice xenograft tumors. The proportion of CD68+/CD163+/CD206+ positive cells in xenograft tumors was significantly increased in the mice overexpressing circSMARCC1 group compared to the vector group (Fig. 7B). These results suggested that the expression of circSMARCC1 is positively correlated with M2 macrophage infiltration, and circSMARCC1 may induce M2 macrophage colonization.

Fig. 7

circSMARCC1 promotes M2 macrophage polarization and recruitment via the CCL20-CCR6 axis. A IHC detected the infiltration of CD68+/ CD163+/ CD206+ macrophages in human PCa tissue and Para-cancerous tissues (scale bar, 100 μm and 20 μm). B IHC detected the infiltration of CD68+/ CD163+/ CD206+ macrophages in mouse xenograft tumors (scale bar, 100 μm and 20 μm). C, D The expression of M1 phenotype (TNF-a, CD80 and CD86) and M2 phenotype (IL-10, ARG-1 and CD163) markers were detected by qRT-PCR in THP-1 cells, THP-1-Mø and THP-1-M2 cells. E The proportion of CD68 and CD163 positive cells detected by flow cytometry. F The CM prepared from lv-circSMARCC1 or vector and sh-circSMARCC1 or sh-nc tumor cells was used to evaluate the migration ability of macrophages through transwell experiments (scale bar, 50 μm). G The migration ability of TAM was detected by transwell assay using CCL20 recombinant protein and CCR6 neutralizing antibody (scale bar, 50 μm). H The CM prepared from THP-1-Mø or THP-1-M2 cells for PCa cells migration experiments (scale bar, 50 μm). I, J qRT-PCR tested the expression of M2 phenotypes marker (IL-10, ARG-1, CD68 and CD163) when co-cultured with DU145-lv-circSMARCC1 cells or vector cells or using CCL20 recombinant protein, with IL-4 and IL-3 polarization as the reference. K The expression of CD68, CD163 and CCR6 in THP-1 cells, THP-1-Mø cells stimulated by CCL20 recombinant protein, and THP-1-Mø cells (with and without co-culture) were detected by Western blotting. L The expression of CD68, CD163 and CCR6 in THP-1-Mø cells and THP-1-M2 cells (with and without anti-CCR6) were detected by Western blotting. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

CircSMARCC1 promotes M2 macrophage recruitment through CCL20-CCR6 axis

The above results indicated that the expression levels of circSMARCC1 and CCL20 were positively correlated both in PCa cells and tissues. We hypothesized that circSMARCC1 could regulate M2 macrophage recruitment through the CCL20-dependent process. We treated THP-1 cells with 100 ng/ml PMA for 24 h and found that THP-1 cells became adherent and extended their pseudopod. qRT-PCR was performed to detect macrophage markers and we found that the expression of CD68, CD80 were significantly up-regulated and CD14 was down-regulated in THP-1 cells after PMA treatment compared to the control (Fig. 7C). The data suggested that THP-1 cells were induced to become THP-1-Mø. Next, THP-1-Mø were incubated with 20 ng/ml IL-4 and 20 ng/ml IL-13 for 48 h, and then the typical M1 phenotypes (TNF-a, CD80, and CD86) and M2 phenotypes (IL-10, ARG-1, and CD163) markers were investigated using qRT-PCR. The results showed a significant increase in mRNA of ARG-1 and CD163 expression stimulated by IL-4 and IL-13, while CD86 expression was down-regulated (Fig. 7D). In addition, flow cytometry demonstrated a substantial increase in the proportion of CD68+ and CD163+ macrophages (Fig. 7E). These results indicated a polarization of THP-1 towards THP-1-M2.

To investigate the effect of circSMARCC1 on the migration of TAMs, the CM prepared from lv-circSMARCC1 or vector and sh-circSMARCC1 or sh-nc tumor cells were added to the lower chamber for macrophage migration experiments. Consistent with the expected results, the number of migrating TAMs increased significantly when induced with the CM from lv-circSMARCC1 tumor cell, whereas TAM migration was significantly reduced in sh-circSMARCC1 group (Fig. 7F). CC-chemokine ligand 20 (CCL20) and its selective receptor CCR6 [27], known to be responsible for the chemoattraction of TAMs in homeostatic conditions [28], have been recently found to be involved in metastatic progression of colon cancer [24]. we hypothesized that the chemotaxis of TAMs may be due to the binding of CCL20 to CCR6 in PCa. By analyzing the TCGA database, we found a strong positive correlation between CCL20 and CCR6 levels in PCa tissues (Fig. S3D). To determine whether the fact that circSMARCC1 induces migration of TAMs is critically mediated by CCL20-CCR6 axis, we performed transwell migration assays using CCL20 recombinant protein and CCR6 neutralizing antibody. The results showed that the CCL20 recombinant protein significantly promoted the migration of TAMs, however, this promotion was reversed by the introduction of the CCR6 neutralizing antibody (Fig. 7G). These findings suggested that circSMARCC1 plays a key role in mediating the migration of M2 macrophages via CCL20-CCR6 axis.

TAM had been shown to promote EMT in tumor cells [29]. We prepared the CM from THP-1-Mø and THP-1-M2 separately and used them to perform migration assays in PCa cells. We observed that the CM of THP-1-M2 significantly increased the migration of DU145 and PC3 cells compared to the CM of THP-1-Mø (Fig. 7H). In addition, we also examined the effect of CM from THP-1-M2 on the proliferation of stably transfected PC cells. CCK8 results showed that the CM of TAMs did not alter the proliferation of PCa cells (Fig. S3E). These results suggested that circSMARCC1 enhances the recruitment of TAMs via the CCL20-CCR6 axis, thereby facilitating the progression of PCa.

CircSMARCC1 promotes M2 macrophage polarization via the CCL20-CCR6 axis

To explore the effect of circSMARCC1 on the polarization of TAMs, we co-cultured the stably transfected PCa cells with THP-1-Mø. The PCa cells of lv-circSMARCC1 or vector, sh-circSMARCC1 or sh-nc were seeded in the 0.4 μm pore size upper inserts and then transferred to the 6-well plates pre-inoculated with THP-1-Mø. After 48 h, macrophages were collected for the following experiments. qRT-PCR test confirmed that the expression of M2 phenotype (IL-10, ARG-1 and CD163) markers was significantly up-regulated in the co-cultured with DU145-lv-circSMARCC1 cells compared with the vec