The sample collection was based on the specimen bank of Nanjing Medical University. The study population included patients with breast cancer who underwent surgery at our hospital from May 2019 to January 2023, including 60 patients with bone metastasis of breast cancer and 110 patients with primary breast cancer. The samples of bone metastasis, primary lesions, and adjacent breast cancer tissues were obtained during the surgery. All the samples were quickly frozen in liquid nitrogen. All the patients and their families were informed of the study content and signed an informed written consent form. The study was approved by the Ethics Committee of the Jiangsu Cancer Hospital (Approval No. 2019017).

The circRNAs expressed differentially in metastatic breast cancer were screened from the GEO database GSE111504. GSE111504 circRNA expression profile was based on GPL21825 platform from parental MDA-MB-231 (231-PAR) cells, isogenic brain metastatic cells, and metastatic cells. The volcano map was conducted based on the condition of $\mid$logFC$\mid$ $\ge$1 and p 0.05. The online circInteractome [16] and circBank [17] databases were used to predict the targeting miRNAs of circFOXP1 and overlapped them using jvenn [18]. The binding mRNAs of miR-338-3p were derived from Targetscan [19], miRDB [20], miRWalk (http://mirwalk.umm.uni-heidelberg.de/), and starBase [21]. String and Cytoscape were used to analyze the PPI network and hub genes. The GO and KEGG enrichment analysis (https://www.bioinformatics.com.cn/) were performed using clusterProfilter 4.8.2 and R package version 4.3 [22, 23].

Human breast cancer cells MCF-7 (HTB-22), SK-BR-3 (HTB-30), T47D (HTB-133), and MDA-MB-231 (CRM-HTB-26), as well as non-tumorigenic epithelial cell line MCF-10A (CRL-10317), were acquired from American Type Culture Collection (ATCC; Manassas, VA, USA) with short tandem repeats (STR) verification. All cancer cell lines were maintained in liquid nitrogen and incubated in Dulbecco’s Modified Eagle Medium (DMEM; Gibco, Grand Island, NY, USA) with 10% fetal bovine serum (FBS; Gibco, Grand Island, NY, USA) at 37 °C in a 5% CO${}_2$ humidified incubator. MCF-10A cells were maintained in DMEM/F12 medium with 5% equine serum (heat-inactivated). Bone marrow mesenchymal stem cells (BMSCs) were purchased from Cyagen Biotechnology (Santa Clara, CA, USA) and maintained in a low-glucose DMEM medium with 10% FBS. The mycoplasma detection was carried out on a regular basis every one to three months using MycAwayTM Plus-Color One-Step Mycoplasma Detection Kit (Yeasen Biotechnology, Shanghai, China).

circFOXP1 siRNAs (si-circRNA-1, -2, -3), siRNA negative control (si-NC), miR-338-3p mimic (miR-338-3p), and mimic NC (miR-NC) were synthesized by GenePharma (Shanghai, China). Transfection of plasmids into breast cancer cells was performed with the help of lipofectamine 2000 (Invitrogen, Grand Island, NY, USA). After 48 h of transfection, a real-time quantitative reverse transcription polymerase chain reaction (RT-qPCR) was performed to detect the transfection efficiency. The sequences for circFOXP1 siRNAs were as follows: si-circRNA-1, 5${}^{\prime }$-AGGGAAAGGTTCCCGTGTCAG-3${}^{\prime }$; si-circRNA-2, 5${}^{\prime }$-CCAAAAGGGAAAGGTTCCCGT-3${}^{\prime }$; si-circRNA-3, 5${}^{\prime }$-GAAAGGTTCCCGTGTCAGTGG-3${}^{\prime }$; for si-NC, 5${}^{\prime }$-UUCUCCGAACGUGUCACGUTT-3${}^{\prime }$; miR-338-3p mimic, 5${}^{\prime }$-UCCAGCAUCAGUGAUUUUGUUG-3${}^{\prime }$; and mimic NC, 5${}^{\prime }$-CAGUAC UUUUAGUGUGUACAA-3${}^{\prime }$.

The co-culture model was performed by applying the medium supernatant of different transfected cells to the BMSCs culture. The co-cultured BMSCs were incubated for 1, 3, 5, 7, and 9 days to detect alkaline phosphatase (ALP) activity.

The clinical samples were cut into sizes measuring approximately 0.5 $\times$ 0.5 $\times$ 0.5 cm and were placed in 1.5 mL Eppendorf tubes. The supernatant of the cultured cells was discarded and rinsed thoroughly with phosphate buffered saline (PBS). Then Trizol reagent (Takara, Dalian, Liaoning, China) was added to the tissue samples and cells to extract total RNAs. Total RNA (treated with RNase R for circRNA/without RNase R for mRNA) was reverse transcribed into cDNA using a High-Capacity cDNA Reverse Transcription Kit (ThermoFisher, Applied Biosystems, Waltham, MA, USA). qPCR was performed using SYBR Green PCR Master Mix (Vazyme, Nanjing, Jiangsu, China) on a 7900HT Fast Real-Time PCR system. Finally, GAPDH was used as a reference to calculate the relative content of the target circFOXP1 and genes. The PCR sequences for circFOXP1 were forward 5${}^{\prime }$-CCTCCTCTGCACCTTCCAAG-3${}^{\prime }$ and reverse 5${}^{\prime }$-TGTTGCTGGAGGATCTGCTG-3${}^{\prime }$; for GAPDH were: forward 5${}^{\prime }$-GAGTCAACGGATTTGGTCGT-3${}^{\prime }$ and reverse 5${}^{\prime }$-TTGATTTTGGAGGCATCTCG-3${}^{\prime }$.

After co-culture, the cells were collected and homogenized with RIPA lysis buffer to obtain total protein. The protein purity and concentration were measured by standard methods of BCA assays (Beyotime, Shanghai, China). 10 µg per sample were separated by SDS-PAGE for 90 min, then blotted onto nitrocellulose membranes for 1 h, and incubated with primary antibodies (Abcam, Cambridge, UK) overnight at 4 °C. Nitrocellulose membranes were washed four times and treated with appropriate secondary HRP-conjugated goat anti-rabbit or goat anti-mouse antibodies (Abcam, Cambridge, UK) at 37 °C for 1 h. Finally, membranes were scanned and transferred to the Image J program (version 1.53, National Institutes of Health, Bethesda, MD, USA) for gel gray measurement.

The cell proliferative capacities were assessed using cell counting kit-8 (CCK-8, Dojindo, Tokyo, Japan). The transfection-treated breast cancer cells were incubated in 96-well plates (approximately 5000 cells/well). The CCK-8 reagent was added to each well at 0, 24, 48, and 72 h time points. The cells were incubated for further 1 h. The micro-plate Reader (Bio-Rad, Hercules, CA, USA) was used to determine the absorbance at a wavelength of 450 nm (OD value).

ALP activity of co-cultured BMSCs was measured using an Alkaline Phosphatase Activity Assay Kit (Beyotime, Shanghai, China).

DMEM medium without serum was put on the upper layer of the Transwell culture plate (pore size = 8 µm) coated with or without Matrigel (Corning, Corning, NY, USA), and DMEM medium containing 10% FBS was added to the lower layer. The cell suspension (5 $\times$ 10${}^4$ cells/well) was added to Transwell plates and incubated for 24 h at 37 °C in a 5% CO${}_2$ incubator. The filter membrane was removed, and the cells that migrated or invaded the lower layer were stained and counted under a microscope.

The circFOXP1-wild type (WT-circRNA) or circFOXP1-mutant (MUT-circRNA) binding sites with miR-338-3p were inserted into the pGL3 promoter vector (Realgene, Shanghai, China). Breast cancer cells were cultured and seeded in 24-well plates. miR-338-3p mimic, mimic NCs, and luciferase reporter plasmids (WT-circRNA or MUT-circRNA) were co-transfected into the cells when cells grew to 80% confluence. After 48 h of incubation, firefly luciferase activities were measured using Dual-luciferase Reporter Assay System (Promega, Madison, WI, USA).

The statistical data are expressed as mean $\pm$ standard deviation (SD) and analyzed using GraphPad 9.0 software (Dotmatics, Boston, MA, USA). All experiments repeated at least three times. One-way analysis of variance (ANOVA) or two-way ANOVA was performed for comparison among groups and the Student t-test was used for pairwise comparison. p 0.05 indicates statistical significance.

According to GSE111504 circRNA profile data, the volcano map was based on the condition of $\mid$logFC$\mid$ $\ge$1 and p 0.05. circFOXP1 (hsa_circ_0008234) was one of the upregulated circRNAs in metastatic breast cancer cells (Fig. 1A). We measured the circFOXP1 expression in breast cancer tissues using RT-qPCR. circFOXP1 expression was high in breast cancer tissues (BCa/T), particularly in tissues with bone metastasis of breast cancer (BM-BCa/T) compared with adjacent normal tissues (BCa/N and BM-BCa/N) (p 0.001, Fig. 1B). Furthermore, circFOXP1 levels were higher in breast cancer cells, compared with normal MCF-10A cells (p 0.01, Fig. 1C).

Fig. 1.

circFOXP1 expression in breast cancer. (A) Volcano plot of circRNAs from GES111504 and screened upregulated circFOXP1 (hsa_circ_0008234). (B) circFOXP1 expression was detected in breast cancer bone metastasis tissues, primary breast cancer tissues, and surrounding tissues. (C) circFOXP1 expression was increased in breast cancer cells in contrast with that in normal MCF-10A cells. **p 0.01, ***p 0.001. FOXP1, Forkhead box protein P1; circRNAs, circular RNAs; BCa/T, breast cancer tissues; BM-BCa/T, bone metastasis breast cancer tissues; BCa/N, breast cancer adjacent normal tissues; BM-BCa/N, bone metastatic breast cancer adjacent normal tissues.

circFOXP1 siRNAs (si-circRNA-1, si-circRNA-2, si-circRNA-3) were transfected into SK-BR-3 MDA-MB-231 cells. Fig. 2A,B depict that si-circRNA-1 exhibited the highest inhibitory effect on circFOXP1 expression (p 0.001). Thus, si-circRNA-1 (termed si-circRNA) was used for subsequent experiments. CCK-8 assay results suggested that interference of circFOXP1 decreased the cell proliferative abilities (p 0.05, Fig. 2C). A similar inhibitory trend caused by si-circFOXP1 was observed in migration assay (p 0.001, Fig. 2D) and invasion assay (p 0.001, Fig. 2E).

Fig. 2.

Functional role of circFOXP1 in breast cancer cells. (A,B) circFOXP1 siRNAs decrease circFOXP1 expression in SK-BR-3 and MDA-MB-231 cells. (C) CCK-8 assay measured the influence of circFOXP1 siRNA in breast cancer cells. (D,E) circFOXP1 interference decreased breast cancer cell migration (D) and invasion (E) abilities. *p 0.05, ***p 0.001. CCK-8, cell counting kit-8; si-NC, siRNA negative control; si-circRNA, circFOXP1 siRNA.

To investigate the effects of circFOXP1 on the interaction between breast cancer cells and BMSCs, the supernatant of MDA-MB-231 cells cultured in si-circRNA group and si-NC group were collected and co-cultured with BMSCs. The increased ALP activity of BMSCs co-cultured with MDA-MB-231/si-circRNA group was observed (p 0.05, Fig. 3A). Fig. 3B,C depicted that the five osteogenic genes/proteins of OSX, RUNX2, BSP, OPG, and BMP2 expression were increased in BMSC + MDA-MB-231/si-circRNA group contrast with control group (p 0.01).

Fig. 3.

circFOXP1 silencing in MDA-MB-231 improves the BMSCs osteogenesis. (A) Alkaline phosphatase activity analysis of BMSCs co-cultured with transfected MDA-MB-231 culture supernatant. (B) The osteogenic genes OSx, RUNX2, BSP, OPG, and BMP2 in BMSCs after co-culture. (C) Osteogenic proteins expression of BMSCs after co-culture. *p 0.05, **p 0.01, ***p 0.001. BMSCs, bone marrow-derived mesenchymal stem cells.

Two circRNA-miRNA interaction databases (circInteractome and circBank) were used to screen the intersection of miRNAs that may play regulatory roles. These databases overlapped five potential miRNAs (Fig. 4A). Among the five miRNAs, miR-338-3p exhibited the highest levels after silencing of circFOXP1 in MDA-MB-231 cells, which was selected to validate the targeting relationship with circFOXP1 (p 0.001, Fig. 4B). Fig. 4C displays the binding sites between circFOXP1 and miR-338-3p. Dual-luciferase reporter assay indicated that miR-338-3p overexpression suppressed the luciferase activities of circFOXP1 wild-type of MDA-MB-231 cells (p 0.001, Fig. 4D).

Fig. 4.

circFOXP1 could sponge miR-338-3p. (A) circInteractome and circBank databases overlapped the potential predictive miRNAs of circFOXP1. (B) Relative expression of miRNAs was measured in si-circFXOP1 transfected MDA-MB-231 cells. (C) Binding sites between circFOXP1 and miR-338-3p. (D) Dual-luciferase reporter assay confirmed the relationship between circFOXP1 and miR-338-3p. *p 0.05, ***p 0.001, ns, no significance. WT-circRNA, circFOXP1-wild type; MUT-circRNA, circFOXP1-mutant; miR-NC, mimic NC; miR-338-3p, miR-338-3p mimic.

To explore the potential molecular mechanism underlying circFOXP1, TargetScan, miRDB, miRWalk, and starBase algorithms were used to overlap the potential target mRNAs of miR-338-3p (Fig. 5A). The STRING online database predicted PPI using the overlapped 92 mRNAs of miR-338-3p (Fig. 5B) and the top ten target mRNAs (PTEN was the leading mRNA) were visualized by Cytoscape (Fig. 5C). The GO (Fig. 5D) and KEGG (Fig. 5E) pathway enrichment analysis indicated that the targets genes were principally enriched in dendrite development, morphogenesis, and other biological processes, and pathway enriched in MAPK signaling pathway, regulation of actin cytoskeleton, tight junction, Ras signaling pathway, and PI3K-AKT signaling pathway.

Fig. 5.

Downstream mRNAs of miR-338-3p and pathway enrichment analysis. (A) Downstream mRNAs of miR-338-3p were overlapped using four online databases. (B) STRING conducted a protein-protein interaction (PPI) network using overlapped 92 mRNAs. (C) The leading ten hub genes were identified using the Cytoscape network. (D) Gene Ontology (GO) analysis was performed using the related genes in the network. (E) Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway enrichment analysis of the genes in the network genes.