CircFOXM1 promotes proliferation and metastasis of hepatocellular carcinoma via regulating miR-1179/SPAG5 axis

Hepatocellular carcinoma (HCC) predominantly occurs in patients with chronic liver disease, accounting for 70–90% of all liver cancer cases. The role of circFOXM1/miR-1179/SPAG5 axis in HCC has not been reported. This study aimed to explore the regulatory mechanism of circFOXM1 in HCC proliferation and metastasis. RNA polymerase inhibitor actinomycin D and RNase R exonuclease were used to identify circFOXM1 in HCC cells. The qRT-PCR was used to detect circFOXM1 expression. Specific siRNA for circFOXM1 was designed, and the sequence of circFOXM1 was inserted in pLCDH-ciR to overexpress circFOXM. Cell proliferation was detected by CCK8 in vitro, by tumor volume and tumor weight of HCC xenograft in vivo. Cell migration was detected by transwell test. Binding status of circFOXM1 with miR-1179 was detected by luciferase reporter gene assay. Rescue experiments were applied to identify the oncogenic mechanism of circFOXM1 in HCC cells. Actinomycin D assay confirmed the cyclization of circFOXM1. RNase R treatment showed that circFOXM1 was not affected by RNase R exonuclease. CCK8 assay, tumor volume and tumor weight showed that circFOXM1 effectively promoted HCC cell proliferation. Transwell assay showed that circFOXM effectively promoted migration and invasion abilities of HCC cells. Luciferase reporter gene activity assay showed that miR-1179 had complementary binding sites with circFOXM1 and SPAG5. CircFOXM1 silencing inhibited malignant phenotypes in HCC cells were partly rescued by either miR-1179 silencing or SPAG5 overexpression. CircFOXM1 promoted HCC cell proliferation and metastasis by regulating miR-1179/SPAG5 axis.

Actinomycin D assay. Cells were exposed to 2 mg/ml actinomycin D (Sigma) for transcription inhibition.
And the RNA stabilities of circFOXM1 and FOXM1 were analyzed by qRT-PCR.
RNase R treatment. Incubated 10 μg total RNA with 3 U/μg RNase R (Epicentre Technologies) at 37 °C for 15 min, and the RNA stabilities of circFOXM1 and FOXM1 were analyzed by qRT-PCR.
qRT-PCR. Total RNAs were extracted using TRIzol reagent (Invitrogen). qRT-PCR for mRNA detection was performed with SYBR Premix Ex Taq (Takara) system. qRT-PCR for miRNA detection was performed with SYBR Prime Script miRNA RT-PCR kit (Takara). U6 was used as an internal control for miRNA; GAPDH was used as an internal control for circRNA and mRNA. ABI 7500 real-time PCR system (Life Technology Corporation) was used for qRT-PCR reaction. The relative gene expression was calculated by 2 −∆∆CT . The relative expressions of circRNA and mRNA were normalized to GAPDH. The relative miRNA expression was normalized to human U6. CCK-8 proliferation assay. A 10 μL of CCK-8 reagent (Dojindo, Japan) was added into a 96-well plate with 3000 cells per well, and incubated at 37 °C for 2 h. The OD was measured at 450 nm with a microplate reader (Thermo).
Transwell analysis. Cells (0.5 × 10 5 ) were suspended in 200 μl medium without FBS and seeded into the upper compartments of "Transwell" chambers (BD Biosciences, Heidelberg, Germany). The medium containing 10% FBS was injected into the bottom chamber, and the cells were stained with crystal violet for 15 min. After 24 h incubation, cells migrated to the lower surface were fixed and stained. For cell invasion assay, Matrigel (Sigma) was plated into the upper chamber surfaces.  Statistical analysis. SPSS 24.0 (IBM, Chicago, USA) was used to perform statistical analysis on the experimental results obtained in this research. The data differences between the groups were analyzed by Student's t test and one-way analysis. p < 0.05 was used as a statistically significant standard.
Ethical approval. This study has been reported in accordance with ARRIVE guidelines.

Identification of circFOXM1 in HCC cells.
To identify the expression status of circFOXM1 in HCC cells, circFOXM1 expression levels in different HCC cells (Huh7, HepG2, Hep3B, HCCLM3 and MHCC97-H) and the normal human hepatocytes (MIHA) were compared, which showed that circFOXM1 expressions in Huh7, HepG2, Hep3B, HCCLM3 and MHCC97-H cells were all significantly upregulated versus MIHA, with the highest expression level in Huh7 cells (Fig. 1A). Herein, the stability of circFOXM1 in HCC cells was further investigated using Huh7 cells. After both the circFOXM1 and FOXM1 in Huh7 cells were exposed to Actinomycin D ( Fig. 1B) or Ribonuclease R (RNase R, Fig. 1C), the expressions of circFOXM1 and FOXM1 were detected by qRT-PCR respectively, which revealed that circFOXM1 was more stable and more resistant to RNase R digestion than FOXM1.

CircFOXM1 promotes proliferation and metastasis of HCC cells in vitro, and HCC xenograft growth in vivo.
To explore the biological function of circFOXM1 in HCC cells, circFOXM1 was either silenced by transfection of circFOXM1 specific siRNA in Huh7 and HepG2 HCC cells with higher circFOXM1 expressions, or overexpressed by transfection of circFOXM1 sequence containing plasmid pLCDH-circFOXM1 in HCCLM3 and MHCC97-H HCC cells with lower circFOXM1 expressions. As shown in Fig. 2A, circFOXM1 was effectively silenced in Huh7 and HepG2 cells and overexpressed in HCCLM3 and MHCC97-H cells determined by qRT-PCR. Then the influences of circFOXM1 genetic modification on HCC cell phenotypes were further investigated both in vitro and in vivo. The results of CCK8 showed that down-regulation of circFOXM1 expression plays an important role in reducing the proliferation of both Huh7 and HepG2 HCC cells, while increased circFOXM1 expression promoted proliferation of HCCLM3 and MHCC97-H HCC cells (Fig. 2B). Furthermore, transwell analysis showed that circFOXM1 silence effectively reduced the migration (  www.nature.com/scientificreports/ CircFOXM1 sponges miR-1179 in HCC cells to increase SPAG5 expression. To investigate the molecular mechanism of circFOXM1 in HCC carcinogenesis and development, miR-1179 and SPAG5 expressions were compared between HCC cells and MIHA cells, which showed that miR-1179 expression was significantly downregulated in HCC cells than in MIHA (Fig. 3A), and SPAG5 expression was significantly upregulated in HCC cells than in MIHA cells (Fig. 3B); furthermore, the potential miRNAs had complementary binding sites with circFOXM1 were first predicted using circular RNA Interactome (https:// circi ntera ctome. nia. nih. gov/) (Fig. 3C), followed by dual-luciferase reporter activity analysis in Huh7 and HepG2 HCC cells. Luciferase activities of wild-type circFOXM1 3′UTR reporter gene in Huh7 and HepG2 HCC cells were significantly inhibited when miR-1179 was overexpressed with the mimics, while the inhibited luciferase activities of circFOXM1 3′UTR reporter gene were rescued when the predicted binding sites of circFOXM1 3′UTR with miR-1179 were mutated (Fig. 3C,D). Furthermore, the direct regulation of miR-1179 by circFOXM was investigated by detecting miR-1179 expression in HCC cells after circFOXM1 was silenced or overexpressed respectively. Our results showed that miR-1179 expression was significantly upregulated in Huh7 and HepG2 HCC cells after circFOXM1 was silenced, while miR-1179 expression was significantly downregulated in HCCLM3 and MHCC97-HHCC cells after circFOXM1 was overexpressed (Fig. 3E).
To further reveal the ceRNA network triggered by circFOXM1, potential mRNAs had complementary binding sites with miR-1179 were first predicted with online tool targetscan (http:// www. targe tscan. org/), which showed that the 3′UTR of sperm-associated antigen 5 (SPAG5) had potential binding sites with miR-1179 (Fig. 3G). Herein, whether circFOXM1 sponges miR-1179 thus to upregulate SPAG5 expression in HCC cells was investigated. The Expression DIY module in the GEPIA database was used to analyze SPAG5 expression profile in the HCC (liver hepatocellular carcinoma, LIHC) and the normal tissues, which showed an increased trend of circFOXM1 transcripts in HCC tissues than in normal control tissues (Fig. 3F). Furthermore, the Survival Plots module in the GEPIA database was used to analyze the association between SPAG5 expression levels with the overall survival of HCC patients, which showed that HCC patients with high SPAG5 expression exhibited a significant poor overall survival (Fig. 3H). Moreover, our results showed that miR-1179 silencing with inhibitor in Huh7 and HepG2 HCC cells considerably upregulated SPAG5 expressions at both mRNA (Fig. 3I) and protein (Fig. 3J) levels, while miR-1179 overexpression with mimics in Huh7 and HepG2 HCC cells considerably downregulated SPAG5 expressions at both mRNA (Fig. 3I) and protein (Fig. 3J) levels. Pearson correlation analysis showed that circFOXM1 expression was significantly positively associated with SPAG5 expression (Fig. 3K), while miR-1179 expression was significantly negatively associated with SPAG5 expression (Fig. 3L). These findings suggested that circFOXM1 promoted HCC oncogenesis and development by sponging miR-1179 to upregulate SPAG5 expression.

CircFOXM1/miR-1179/SPAG5 axis promotes HCC cell malignant phenotypes.
To explore the involvement of circFOXM1/miR-1179/SPAG5 axis in HCC carcinogenesis and progression, the malignant phenotypes of HCC cells were measured following the genetic modifications. Viabilities of Huh7 and HepG2 cells after circFOXM1 silencing with/without miR-1179 inhibitor or SPAG5 overexpression were determined by CCK-8 assay. The results showed that circFOXM1 silencing inhibited OD values were partly rescued by either miR-1179 silencing or SPAG5 overexpression in a time (0-96 h) dependent manner (Fig. 4A); meanwhile, viabilities of HCCLM3 and MHCC97-H cells after circFOXM1 overexpression with/without miR-1179 mimics or SPAG5 silencing were determined by CCK-8 assay. The results showed that circFOXM1 overexpression increased OD values were partly rescued by either miR-1179 mimics or SPAG5 silencing in a time (0-96 h) dependent manner (Fig. 4B). Migration capabilities of Huh7 and HepG2 cells after circFOXM1 silencing with/ without miR-1179 inhibitor or SPAG5 overexpression were determined by transwell assay without Matrigel. The results showed that circFOXM1 silencing inhibited migration capabilities were partly rescued by either miR-1179 silencing or SPAG5 overexpression (Fig. 4C). Invasion capabilities of Huh7 and HepG2 cells after circFOXM1 silencing with/without miR-1179 inhibitor or SPAG5 overexpression were determined by transwell assay with Matrigel. The results showed that circFOXM1 silencing inhibited invasion capabilities were partly rescued by either miR-1179 silencing or SPAG5 overexpression (Fig. 4D). Migration capabilities of HCCLM3 and MHCC97-H cells after circFOXM1 overexpression with/without miR-1179 mimics or SPAG5 silencing were determined by transwell assay without Matrigel. The results showed that circFOXM1 overexpression increased migration capabilities were partly rescued by either miR-1179 mimics or SPAG5 silencing (Fig. 4E). Invasion capabilities of HCCLM3 and MHCC97-H cells after circFOXM1 overexpression with/without miR-1179 mimics or SPAG5 silencing were determined by transwell assay with Matrigel. The results showed that circFOXM1 overexpression increased invasion capabilities were partly rescued by either miR-1179 mimics or SPAG5 silencing (Fig. 4F). These findings indicated that miR-1179 knockdown or SPAG5 overexpression partially reversed circFOXM1 silencing inhibited malignant phenotypes in HCC cells, meanwhile, miR-1179 overexpression or SPAG5 silencing partially reversed circFOXM1 overexpression promoted malignant phenotypes in HCC cells.

Discussion
CircFOXM1 has been reported to contribute to sorafenib resistance of HCC cells by regulating miR-1324/ MECP2 17 , indicating the potential significance of circFOXM1 in HCC management. In this study, Actinomycin D treatment verified the cyclization of circFOXM1; RNase R treatment showed that circFOXM1 was not affected by RNase R exonuclease; HCC xenograft experiment showed that circFOXM1 effectively promoted the growth of HCC cells in vivo; CCK8 assay showed that circFOXM1 effectively promoted the proliferation of HCC cells; transwell assay showed that circFOXM1 effectively promoted the migration and invasion abilities of HCC cells.
Studies have reported that SPAG5 serves a promising prognostic factor in HCC and acts as an oncogene 18 www.nature.com/scientificreports/ inhibitor or SPAG5 overexpression reversed circFOXM1 siRNA-mediated inhibitory effect on HCC cell proliferation, migration and invasion abilities; and miR-1179 mimics or SPAG5 silencing reversed pLCDH-circFOXM1mediated promotion on HCC cell proliferation, migration and invasion abilities. These findings implicate that circFOXM1 is an oncogenenic factor and plays an important role in promoting oncogenesis and metastasis of HCC.
Studies have confirmed that miR-1179 is an important tumor regulator, which directly regulates Notch signaling pathway and inhibits breast cancer cell metastasis 23 . Meanwhile, studies have also confirmed that miR-1179 can target HMGB1 to inhibit the proliferation of gastric cancer cells 24 . It has also been pointed out that miR-1179 can target the E2F5 gene in regulating pancreatic cancer cell proliferation, thereby inhibiting tumor cell migration and invasion 25 ; plays an important role in regulating the proliferation of glioblastoma cells by directly targeting E2F transcription factor 5 26 ; and plays important role in cancer through Akt signaling 27,28 . Moreover, circFOXM1 has been confirmed to accelerate the progression of papillary thyroid carcinoma by sponging miR-1179 10 . Literature review also revealed that SPAG5 was the downstream of miR-1179 in lung adenocarcinoma 29 .
SPAG5, also named mitotic spindle associated protein, has been reported to influence the separation of sister chromatids by regulating spindles, thus affecting the cell cycle. SPAG5 overexpression is associated with carcinogenesis and development of lung, breast, cervical and bladder urothelial cancers 22 . Therefore, the underlying mechanism of circFOXM1/miR-1179/SPAG5 axis in HCC remains to be further elucidated.
CircRNAs are derived from canonical splice sites [30][31][32] , with the sequences from exons, introns, antisense, intragenic, intergenic sequences or both of the exons and introns. However, the exon-derived circRNAs are the most widely studied. Regarding the biogenesis of circRNAs, multiple hypotheses have been proposed. Reverse complementary intron sequence pairing and RNA-binding protein regulation are the most commonly accepted mechanisms of biogenesis 33,34 . However, the biogenesis of circFOXM1 still need to be further explored.
In summary, our current work demonstrated that circFOXM1 promoted HCC proliferation and metastasis, which could be partially reversed by upregulating miR-1179 or downregulating SPAG5. Our results provide a theoretical basis for HCC therapy.