CircATP5B upregulation in glioma correlates with poor patient survival.
We first performed qRT-PCR on both normal brain tissue and glioma specimens, and the relative expression of circATP5B in glioma tissue was higher than in normal brain tissue, and was especially increased in higher WHO grade tissues (Fig. 1a). We found that circATP5B, also named hsa_circ_0027068 according to the annotation of circBase (http://www.circbase.org/), was spliced from exons 8 and 9 of the ATP5B gene (chr12: 57031958–57033091) and formed a sense-overlapping circular transcript of 451 nt (Fig. 1b). Sanger sequencing certified the head-to-tail splicing of circATP5B (Fig. 1c). We then determined whether the head-to-tail splicing of circATP5B results from trans-splicing or genomic rearrangement. To certify the stability of circATP5B, both GSC406 and GSC201 were treated with RNase R, which is a processive 3′ to 5′ exoribonuclease. It was found that circATP5B resisted digestion by RNase R, but the linear form of ATP5B was readily digested (Fig. 1f). Moreover, the results of nuclear–cytoplasm separation illustrated that circATP5B was predominantly localized in the cytoplasm (Fig. 1g) and indicated that it might be an appropriate diagnostic or prognostic marker. In addition, Kaplan–Meier survival analyses showed that the median survival times of lower-grade glioma patients, glioblastoma multiforme (GBM) patients, or total glioma patients with higher expression of circATP5B were all shorter than those in patients with lower circATP5B expression (Fig. 1e).
We cultured six patient-derived primary GSCs and hematoxylin and eosin were used to stain patient-derived glioma tissues (Fig. S1a). Immunofluorescence staining confirmed the enrichment of stem cell markers, CD133 and nestin (Fig. S1b). We also confirmed the differentiation capacity of GSCs with differentiation markers, GFAP and βIII tubulin (Fig. S1c). qRT-PCR showed that the expression of circATP5B was highest in WHO grade IV GSCs (GSC403 and GSC406), followed by WHO grade III GSCs (GSC302 and GSC305) and was lowest in WHO grade II GSCs (GSC201 and GSC203) (Figure. 1d). We found that the expression level of circATP5B in GSC406 was the highest and was the lowest in GSC201. Taken together, these results confirmed that circATP5B is overexpressed in glioma and correlates with poor patient survival.
CircATP5B regulates the proliferation of GSCs.
To detect the functions of circATP5B in GSCs, we selected GSC406 and GSC201 for circATP5B silencing or overexpression. qRT-PCR was performed to detect the transfection efficiency (Fig. S2a, b). Then, we evaluated the effects of circATP5B on the proliferation of GSCs via MTS and EDU assays. All of the results showed that the cell viability and EDU-positive rates were decreased in circATP5B-silenced GSC406, while the opposite results were acquired in circATP5B-overexpressed GSC201 (Fig. 1h, i). Furthermore, the relative size of the neurospheres formed by GSC406 was significantly smaller than those of the control group following circATP5B knockdown, while the opposite result was obtained in GSC201 after circATP5B overexpression (Fig. 1j). Limiting dilution assays also showed that the tumor formation incidence decreased in circATP5B-silenced GSC406 but increased in circATP5B-overexpressed GSC201 (Fig. 1k). Together, these results confirmed that circATP5B plays a vital role in promoting the proliferation of GSCs.
HOXB5 is overexpressed in glioma and correlates with poor patient survival.
HOXB5, as a member of the homeobox gene family, is a vital transcription factor, and HOXB5 overexpression is significantly correlated with cancer progression and a poor prognosis (36, 37). However, the relationship between HOXB5 and glioma remains largely unknown. We found that the expression levels of HOXB5 in glioma tissues were higher than in normal brain tissues by qRT-PCR, western blotting, and immunohistochemical analysis, and were especially increased in higher glioma WHO grades (Fig. 2a, b, d). Moreover, Kaplan–Meier survival analyses showed that the median survival times of lower-grade glioma patients, GBM patients, or total glioma patients with higher HOXB5 expression were shorter than those for patients with lower HOXB5 expression (Fig. 2c, e, f). Both qRT-PCR and western blotting showed that HOXB5 was most highly expressed in WHO grade IV GSCs (GSC403 and GSC406), followed by WHO grade III GSCs (GSC302 and GSC305) and was lowest in WHO grade II GSCs (GSC201 and GSC203) (Fig. S1d, f). Furthermore, we found that the expression level of HOXB5 was higher in each type of GSC compared with other non-GSC types (Fig. S1e, g). Taken together, these results suggested that HOXB5 is overexpressed in glioma and associated with poor patient survival.
HOXB5 regulates the proliferation of GSCs
To confirm whether HOXB5 correlated with the proliferation of glioma, we firstly performed qRT-PCR and western blotting to detect the efficiency of HOXB5 knockdown or overexpression (Fig. S2c–f). Then, we performed MTS and EDU assays and the results showed that cell viability and the rates of EDU-positive GSCs were decreased in HOXB5-silenced GSC406 but increased in HOXB5-overexpressed GSC201 (Fig. 2g–i). Moreover, the relative size of neurospheres formed by GSC406 was significantly smaller than those of the control group following HOXB5 knockdown, while the opposite result was obtained in HOXB5-overexpressed GSC201 (Fig. 2k). In addition, limiting dilution assays showed that the incidence of tumor formation decreased in HOXB5-silenced GSC406, but increased in HOXB5-overexpressed GSC201 (Fig. 2j, l). In summary, our findings confirmed that HOXB5 is overexpressed in glioma and actively regulates the proliferation of GSCs.
MiR-185-5p negatively regulates HOXB5 expression.
We furtherly predicted that miR-185-5p was the only intersection bound to the 3′-UTR of HOXB5 according to microRNA, miRDB, TargetScan, and Starbase databases (Fig. 3a). We performed qRT-PCR and western blotting to confirm whether miR-185-5p regulated HOXB5 expression, and the results showed that HOXB5 expression levels were downregulated in miR-185-5p mimic-treated GSC406 but upregulated in miR-185-5p inhibitor-treated GSC201 (Fig. 3b-e). Pearson's correlation analyses also confirmed a negative correlation between the expression levels of miR-185-5p and HOXB5 in each WHO grade glioma and in all glioma samples (Fig. 3h). Furthermore, we constructed luciferase reporter plasmids with wild-type and mutant forms of the HOXB5 3′-UTR (Fig. 3f), and luciferase reporter assays showed that the luciferase activity of HOXB5-wt vector was significantly decreased in miR-185-5p mimic-treated GSC406, while obviously increased in miR-185-5p inhibitor-treated GSC201. However, the luciferase activity of the HOXB5-mt vector showed no significant changes (Fig. 3g). Taken together, these results suggested that miR-185-5p negatively regulates HOXB5 expression through binding to the 3′-UTR of HOXB5.
MiR-185-5p suppresses the proliferation of GSCs via HOXB5 inhibition.
To confirm the effects of miR-185-5p and HOXB5 in the proliferation of GSCs, we performed rescue experiments. Both MTS and EDU assays showed that the cell viability and rates of EDU-positive GSCs were decreased in miR-185-5p mimic-treated GSC406, while these effects were reversed after HOXB5 overexpression. The opposite results were obtained in miR-185-5p inhibitor-treated GSC201, and the reverse was observed after HOXB5 knockdown (Fig. 3i-k). Furthermore, the relative size of neurospheres formed by GSC406 was significantly smaller than that of the control group after miR-185-5p mimic treatment, but became larger after HOXB5 overexpression. The opposite results were obtained in miR-185-5p inhibitor-treated GSC201, and this effect was reversed following HOXB5 knockdown (Fig. 3l). Limiting dilution assays showed that the tumor formation incidence decreased in miR-185-5p mimic-treated GSC406, but increased following HOXB5 overexpression. The opposite results were obtained in miR-185-5p inhibitor-treated GSC201, and the effect was reversed following HOXB5 knockdown (Fig. 3m, n). Together, miR-185-5p negatively regulated HOXB5 expression and suppressed the proliferation of GSCs.
CircATP5B acts as a miRNA sponge of miR-185-5p.
CircRNAs have been confirmed to play crucial roles in several molecular mechanisms, such as miRNAs sponging, protein translation, and RNA-binding protein sponging. Increasing evidence has shown that miRNA sponging is the most common role of circRNAs in the development of tumors, including glioma (38–41). First, we predicted the potential target miRNAs of circATP5B according to Starbase and found that miR-185-5p possessed an accurate binding site for circATP5B (Fig. 4a). Second, qRT-PCR showed that the expression of circATP5B was decreased in miR-185-5p mimic-treated GSC406, but increased in miR-185-5p inhibitor-treated GSC201 (Fig. 4b). However, we found that the expression of miR-185-5p increased in circATP5B-silenced GSC406, but decreased in circATP5B-overexpressed GSC201 (Fig. 4c). To confirm the possibility that miR-185-5p binds directly to circATP5B, we constructed luciferase reporter plasmids with wild-type and mutant circATP5B (Fig. 4a). The luciferase activity of circATP5B-wt vector significantly decreased in miR-185-5p mimic-treated GSC406, while obviously increased in miR-185-5p inhibitor-treated GSC201. However, the luciferase activity of the circATP5B-mt vector did not significantly change (Fig. 4d). Moreover, previous studies have shown that miRNAs bind to microRNA response elements (MREs) through RNA-induced silencing complex (RISC), an important component of which is AGO2 protein (42, 43). Therefore, we performed an anti-AGO2 RIP assay to determine whether miR-185-5p and circATP5B were co-enriched in the RISC, and the results showed that both circATP5B and miR-185-5p were efficiently pulled down by anti-AGO2 antibody, compared with the IgG group. Furthermore, significant enrichment of circATP5B and miR-185-5p were observed after miR-185-5p mimic treatment, compared with the miR-185-5p negative control group (Fig. 4e). We also found a negative correlation between the expression levels of circATP5B and miR-185-5p in each WHO grade glioma and in all glioma samples via Pearson's correlation analyses (Fig. 4g). In summary, these results demonstrated the direct interaction between circATP5B and miR-185-5p, and indicated that circATP5B may sponge miR-185-5p.
CircATP5B promotes the proliferation of GSCs through miRNA sponging of miR-185-5p
To confirm the effects of circATP5B and miR-185-5p in the proliferation of GSCs, we performed rescue experiments. MTS and EDU assays showed that cell viability and the rates of EDU-positive GSCs were decreased in circATP5B-silenced GSC406, while this effect was reversed after miR-185-5p inhibitor treatment. However, the opposite results were obtained in circATP5B-overexpressed GSC201, and these upregulations were also reversed after miR-185-5p mimic treatment (Fig. 4f, h). In addition, the relative size of the neurospheres formed by GSC406 was significantly smaller than those of the control group following circATP5B knockdown, but became larger after miR-185-5p inhibitor treatment. While the relative size of the neurospheres formed by GSC201 was obviously larger than that of the control group after circATP5B overexpression, and this reversed following miR-185-5p mimic treatment (Fig. 4j). Limiting dilution assays showed that the tumor formation incidence decreased in circATP5B-silenced GSC406, but increased following miR-185-5p inhibitor treatment. While the opposite results were acquired in circATP5B-overexpressed GSC201, and the increased tumor formation incidence was reversed following miR-185-5p mimic treatment (Fig. 4i, k). Taken together, circATP5B promoted the proliferation of GSCs through sponging miR-185-5p, and there was a negative interaction between circATP5B and miR-185-5P.
CircATP5B can upregulate the expression of HOXB5 through miRNA sponging of miR-185-5p
Since both circATP5B and HOXB5 had specific binding sites for miR-185-5p, to confirm whether circATP5B regulated HOXB5 expression via a miR-185-5p-mediated ceRNA mechanism in GSCs, we firstly detected the expression of HOXB5 via qRT-PCR and western blotting. The results showed that HOXB5 expression was downregulated in circATP5B-silenced GSC406, but upregulated in circATP5B-overexpressed GSC201 (Fig. 5a–c). In addition, we performed rescue experiments by additional treatment with miR-185-5p mimic or miR-185-5p inhibitor. Both the qRT-PCR and western blotting results showed that HOXB5 expression was increased in circATP5B-silenced GSC406 after miR-185-5p inhibitor treatment, while the expression of HOXB5 was decreased in circATP5B-overexpressed GSC201 after miR-185-5p mimic treatment (Fig. 5d-f). Pearson's correlation analyses among clinical glioma specimens showed strong positive correlations between circATP5B and HOXB5 expression in each WHO grade glioma and among the total glioma samples (Fig. 5g). In summary, circATP5B upregulated HOXB5 expression through sponging miR-185-5p.
CircATP5B promotes the proliferation of GSCs by upregulating the expression of HOXB5.
To confirm the effects of circATP5B and HOXB5 in the proliferation of GSCs, we performed rescue experiments. Both MTS and EDU assays showed that cell viability and the rates of EDU-positive GSCs were decreased in circATP5B-silenced GSC406, while these effects were reversed after HOXB5 overexpression. However, cell viability and the rates of EDU-positive GSCs were increased in circATP5B-overexpressed GSC201, and these effects were reversed after HOXB5 knockdown (Fig. 5h-j). Furthermore, the relative size of the neurospheres formed by GSC406 was significantly smaller than that of the control group after circATP5B knockdown, but became larger following HOXB5 overexpression. The opposite results were obtained in circATP5B-overexpressed GSC201, and were reversed after HOXB5 knockdown (Fig. 5k). Limiting dilution assays showed that the incidence of tumor formation decreased in circATP5B-silenced GSC406, but increased following HOXB5 overexpression. The opposite results were obtained in circATP5B-overexpressed GSC201, but reversed after HOXB5 knockdown (Fig. 5l, m). Taken together, these results suggested that circATP5B actively regulates HOXB5 expression through a miR-185-5p-mediated ceRNA mechanism, and promotes the proliferation of GSCs by upregulating HOXB5 expression.
HOXB5 transcriptionally regulates IL6 expression and activates JAK2/STAT3 signaling.
To confirm the possible downstream mechanism of HOXB5 on glioma, we performed GSEA based on the expression of HOXB5. Both TCGA and CGGA datasets showed that higher HOXB5 expression was associated with enrichment of IL6-mediated JAK2/STAT3 signaling (Fig. 6a). Moreover, Pearson’s correlation analyses among clinical glioma specimens revealed significant positive correlations between HOXB5 and IL6 expression in each WHO grade glioma and among the total glioma samples (Fig. 6b). Then, qRT-PCR, western blotting, and ELISA assays showed that IL6 expression was downregulated in HOXB5-silenced GSC406, whereas IL6 expression was upregulated in HOXB5-overexpressed GSC201 (Fig. 6c, d, l, m). Furthermore, since HOXB5 is a transcription factor, we investigated whether HOXB5 transcriptionally regulated the expression of IL6 according to the Jaspar database (Fig. 6e). We performed luciferase reporter assays and found that the luciferase activity of the IL6-wt vector significantly decreased in HOXB5-silenced GSC406, while obviously increased in HOXB5-overexpressed GSC201. However, the luciferase activity of the IL6-mt vector showed no significant changes (Fig. 6f). ChIP assays also revealed that the enrichment of IL6 was decreased in GSC406 following HOXB5 knockdown, whereas it was increased in GSC201 after HOXB5 overexpression (Fig. 6g). In addition, we detected the downstream molecules of the JAK2/STAT3 signaling pathway by western blotting and found that the expression levels of p-JAK2 and p-STAT3 were significantly downregulated in HOXB5-silenced GSC406, whereas the opposite results were obtained in HOXB5-overexpressed GSC201 (Fig. 6l, m). In summary, HOXB5 could transcriptionally regulate IL6 expression and activate JAK2/STAT3 signaling.
HOXB5 regulates the proliferation of GSCs via IL6/JAK2/STAT3 signaling.
We furtherly performed rescue experiments to confirm the effects of HOXB5 and IL6 in the proliferation of GSCs. Both MTS and EDU assays showed that cell viability and the rates of EDU-positive GSCs were decreased in HOXB5-silenced GSC406, while these effects were reversed following additional human recombinant IL6 treatment. The opposite results were obtained in HOXB5-overexpressed GSC201, and these effects were also reversed following additional IL6-neutralizing antibody treatment (Fig. 6h, i). In addition, the relative size of neurospheres formed by GSC406 was significantly smaller than that of the control group after HOXB5 knockdown, but became larger following additional human recombinant IL6 treatment. The opposite results were obtained in HOXB5-overexpressed GSC201, and the effects were reversed following additional IL6-neutralizing antibody treatment (Fig. 6j). Limiting dilution assays showed that the incidence of tumor formation decreased in HOXB5-silenced GSC406, but increased after additional human recombinant IL6 treatment. The opposite results were obtained in HOXB5-overexpressed GSC201 and reversed following additional IL6-neutralizing antibody treatment (Fig. 6k). Taken together, HOXB5 transcriptionally regulated IL6 expression and promoted the proliferation of GSCs via JAK2/STAT3 signaling.
SRSF1 can bind to and upregulate circATP5B expression.
Splicing factor SRSF1 is a typical splicing factor protein that, in addition to its function in splicing, also plays a crucial role in nonsense-mediated mRNA decay, mRNA export, and translation (44). Splicing is considered the main mechanism by which circRNAs originate, and SRSF1 is upregulated and functions as an oncoprotein in several cancers (45). We found that SRSF1 was the most probable RBP to interact with circATP5B, according to the Starbase database with the highest "ClipExpNum". We firstly selected GSC406 and GSC201 to perform SRSF1 knockdown and overexpression assays, and both qRT-PCR and western blotting were used to detect the efficiency of SRSF1 knockdown or overexpression (Fig. S2g–j). Then, qRT-PCR showed that circATP5B expression was downregulated in SRSF1-silenced GSC406, but upregulated in SRSF1-overexpressed GSC201 (Fig. 7a). Furthermore, we performed a RIP assay to detect whether SRSF1 bound to circATP5B and found that the relative enrichment of circATP5B in the anti-SRSF1 group was significantly increased compared with the IgG-treated group. The relative enrichment of circATP5B in the anti-SRSF1 group was obviously decreased after SRSF1 knockdown, but increased after SRSF1 overexpression. However, the relative enrichment of circATP5B in the IgG-treated group showed no significant changes (Fig. 7b, c). Moreover, RNA pull-down assays showed that circATP5B-wt pulled down SRSF1 in GSC406 and GSC201, while this effect was not seen with circATP5B-mt (Fig. 7d, e). Together, these results suggested that, as an RBP and splicing factor, SRSF1 promotes the expression of circATP5B.
SRSF1 regulates the proliferation of GSCs by upregulating circATP5B expression.
To confirm the effects of SRSF1 and circATP5B in the proliferation of GSCs, MTS and EDU assays were performed. Both assays showed that cell viability and the rates of EDU-positive GSCs were decreased in SRSF1-silenced GSC406, and these effects were reversed following circATP5B overexpression. The opposite results were obtained in SRSF1-overexpressed GSC201, and the effects were also reversed following circATP5B knockdown (Fig. 7f, h). In addition, the relative size of neurospheres formed by GSC406 was significantly smaller than that of the control group after SRSF1 knockdown, but became larger following circATP5B overexpression. The opposite results were obtained in SRSF1-overexpressed GSC201, and the effects were reversed following circATP5B knockdown (Fig. 7j). Limiting dilution assays showed that the tumor formation incidence decreased in SRSF1-silenced GSC406, but increased after circATP5B overexpression. While the opposite results were obtained in SRSF1-overexpressed GSC201, and these effects were reversed following circATP5B knockdown (Fig. 7i, k). Taken together, SRSF1 promoted the proliferation of GSCs by binding to and upregulating circATP5B expression.
HOXB5 transcriptionally regulates SRSF1 expression in GSCs
Since HOXB5 is a transcription factor and SRSF1 is an RBP, we determined whether HOXB5 transcriptionally regulates SRSF1 expression in GSCs. We designed two binding sites for HOXB5 in the promoter of SRSF1 according to the Jaspar database (Fig. 7l), then performed luciferase reporter assays. The luciferase activity of SRSF1-wt vector was significantly decreased in HOXB5-silenced GSC406, while obviously increased in HOXB5-overexpressed GSC201. However, the luciferase activity of SRSF1-mt vector showed no obvious changes (Fig. 7m, n). In addition, Pearson's correlation analyses among clinical glioma specimens showed significant positive correlations between HOXB5 and SRSF1 expression in each WHO grade glioma and among the total glioma samples (Fig. 7g). ChIP assays also showed that the enrichment of SRSF1 was decreased in HOXB5-silenced GSC406, while increased in HOXB5-overexpressed GSC201 (Fig. 7o). Finally, qRT-PCR and western blotting showed that SRSF1 expression was downregulated in HOXB5-silenced GSC406, while upregulated in HOXB5-overexpressed GSC201 (Fig. 7p, q, r). Together, these results suggested that HOXB5 transcriptionally regulates SRSF1 expression in GSCs.
The SRSF1/circATP5B/miR-185-5p/HOXB5 feedback loop regulates glioma tumorigenesis in vivo.
We performed orthotopic xenografts to confirm the effects of the SRSF1/circATP5B/miR-185-5p/HOXB5 axis in glioma tumorigenesis in vivo. Compared with the control group, tumor volumes were decreased in the circATP5B knockdown group, the miR-185-5p mimic group, and the SRSF1 overexpression combined with circATP5B knockdown group. In contrast, tumor volumes were increased in the HOXB5 overexpression group, the SRSF1 overexpression group, the circATP5B knockdown combined with the HOXB5 overexpression group, and the miR-185-5p mimic combined with the HOXB5 overexpression group (Fig. 8a, b). Kaplan–Meier survival analysis showed similar results with the circATP5B knockdown group, the miR-185-5p mimic group, and the SRSF1 overexpression combined with circATP5B knockdown group showing longer median survival times compared with the normal control group. The opposite results were obtained in the HOXB5 overexpression group, the SRSF1 overexpression group, the circATP5B knockdown combined with HOXB5 overexpression group, and the miR-185-5p mimic combined with the HOXB5 overexpression group (Fig. 8c). Immunohistochemistry was performed to detect the effects of the SRSF1/circATP5B/miR-185-5p/HOXB5 axis on tumor tissues. The results confirmed that the circATP5B knockdown group, the miR-185-5p mimic group, and the SRSF1 overexpression combined with circATP5B knockdown group had lower expression of HOXB5, SRSF1, IL6, and Ki67, whereas higher expression of HOXB5, SRSF1, IL6, and Ki67 was found in the HOXB5 overexpression group, the SRSF1 overexpression group, the circATP5B knockdown combined with HOXB5 overexpression group, and the miR-185-5p mimic combined with the HOXB5 overexpression group (Fig. 8d). To illustrate our findings, the schematic diagram in Fig. 8e shows that the SRSF1/circATP5B/miR-185-5p/HOXB5 feedback loop promotes the tumorigenesis and proliferation of glioma stem cells through the IL6-mediated JAK2/STAT3 signaling pathway. In summary, our results suggested that the SRSF1/circATP5B/miR-185-5p/HOXB5 axis regulates glioma tumorigenesis and proliferation in vivo.