CircASPM is up-regulated in glioblastomas tissues and correlated with the progression and poor prognosis.
CircASPM is derived from transcript 1 of the ASPM gene (chr1:197099044–197113230) and is formed into a loop by reverse splicing its exons 1–7 (Fig. 1b). We first found that the expression of circASPM was significantly higher in glioma tissues than in normal tissues based on Gene Expression Omnibus (GEO) datasets GSE109569 (Fig. 1a, d). To validate the bioinformatics analysis's conclusions, we selected clinical specimens from 55 glioma patients for qPCR assays and confirmed that circASPM expression was up-regulated with increasing WHO grade (Fig. 1c). Subsequently, we performed a Kaplan-Meier survival analysis on the prognostic significance of circASPM expression in 55 patients. The results showed that the average survival time of patients with high circASPM expression was significantly shorter than those with low expression (qPCR quantification, Cutoff: median, Fig. 1g). The clinical information of the above 55 samples is outlined in Supplementary Table 1.
We further extracted patient-derived GSCs from fresh clinical GBM specimens for culture, from which we selected six GSCs in the best growth state for subsequent studies. Supplementary Table 2 shows that the clinical information about the patients from whom the six GSCs were derived. In Supplementary Fig. 1a, we show hematoxylin and eosin (H&E) staining of the original patient tumors. Immunofluorescence staining was performed to confirm the expression of the stem cell markers CD133 + and Nestin + in the isolated neurospheres (Sup. Figure 1b). Moreover, Supplementary Fig. 1c fully demonstrates the multi-lineage differentiation capacity of GSCs. We then verified the stability of circASPM using RNase R assays. RNase R assays are commonly used to confirm RNAs' circular structure because it can degrade linear RNAs with short 3′-tails, whereas it cannot degrade circRNAs (35). After RNase R disposition, ASPM mRNA expression was significantly decreased in GSC11 and GSC18 cells, while circASPM was unaffected (Fig. 1e, f), indicating that circASPM is more resistant to RNase R digestion. Subsequently, we applied qPCR to determine the expression of circASPM in six cell lines, with GSC11 having the lowest expression and GSC18 the highest (Sup. Figure 1d). In summary, circASPM is a stable circRNA that is potentially involved in promoting glioblastomas' malignant phenotype.
Overexpression of circASPM promoted the proliferation of GSCs in vitro.
We performed a further series of gain-of-function assays to demonstrate the role of circASPM in GSCs, and we designed lentiviral-coated circASPM overexpression plasmids to infect GSC11 and GSC12, respectively, based on their low expression levels of circASPM (Sup. Figure 1d). qPCR confirmed that circASPM overexpression was optimal (Fig. 2a). First, we found by MTS assays that the absorbance values of GSCs after circASPM overexpression were significantly higher than those of controls (Fig. 2b, c), confirming that circASPM overexpression promoted the viability of GSCs. Next, we used Edu assays to examine the role of circASPM overexpression on tumor proliferative capacity. The proportion of Edu-positive cells in circASPM overexpressing cell lines was higher than that in controls, indicating that tumor cells' proliferative capacity was promoted after circASPM overexpression (Fig. 2d-f). Furthermore, in our neurosphere formation experiments, the relative size of spheres and the number of spheres formed per unit time after circASPM overexpression were greater than those in the EV group (Fig. 2g, h). Extreme limiting dilution assay also showed that overexpression of circASPM increased the rate of tumor formation (Fig. 2i, j). The above results showed that circASPM overexpression could promote the proliferation of glioma stem cells.
Knockdown of circASPM suppressed the proliferation of GSCs in vitro.
To continue investigating the function of circASPM in the malignant phenotype of GSCs, we knocked down circASPM by transfecting si-circASPM into GSCs and used qPCR to confirm the efficiency of circASPM knockdown (Fig. 3a). We first used MTS assays to find a significant decrease in absorbance values after circASPM knockdown, demonstrating that tumor viability was significantly inhibited (Fig. 3b, c). Subsequently, in Edu assays, we found that knockdown of circASPM significantly down-regulated the proportion of Edu-positive cells (Fig. 3d-f). Besides, the proliferative activity of neurospheres after circASPM knockdown was found to be significantly inhibited in neurosphere formation assays (Fig. 3g, j). The extreme limiting dilution assays also showed a significant decrease in tumor formation after knockdown of circASPM compared to the NC group (Fig. 2h, i). The above results indicate that knockdown of circASPM can suppress the malignant phenotype of GSCs and induce their apoptosis.
MiR-130b-3p could bind with circAPSM and mediated the function of GBM cells.
Existing studies suggest circRNAs have many microRNA response elements (MREs), and usually exert biological functions through miRNA sponging and regulate the target genes (2). To investigate the interaction between miRNAs and circAPSM, we searched the Starbase dataset and found a binding site for miR-130b-3p to circAPSM (Fig. 4a). We subsequently confirmed by qPCR assays that the expression of circASPM was up-regulated in GSC11 and GSC12 with the administration of miR-130b-3p inhibitor, whereas the opposite result was obtained in GSC17 and GSC18 with the administration of miR-130b-3p mimic (Fig. 4b, c). Further, we examined the relative expression of miR-130b-3p by qPCR assays after overexpression and knockdown of circASPM. MiR-130b-3p expression was also negatively regulated by circASPM levels (Fig. 4d, i). We then utilized luciferase reporter assays and found increased luciferase activity of wild-type circASPM in GSC11 and GSC12 after miR-130b-3p inhibitor treatment (Fig. 4e, f). In contrast, miR-130b-3p mimic treatment reduced the luciferase activity of wild-type circASPM luciferase in GSC17 and GSC18 (Fig. 4g, h). Since miRNAs bind to MREs via the RNA-induced silencing complex (RISC), and the Ago 2 (AGO2) protein is a key component of RISC (36), we performed anti-AGO2 RIP assays to determine whether miR-130b-3p and circAPSM are co-enriched in RISC, in which anti-AGO2 antibodies effectively pulled down both circASPM and miR-130b-3p compared to IgG. In addition, both circASPM and miR-130b-3p also showed significant enrichment after miR-130b-3p mimic treatment compared to miR-130b-3p negative control (Fig. 4m, p). In summary, miR-130b-3p is likely to bind with circAPSM, and the changes in their expression can negatively regulate each other's expression.
We then performed a series of experiments to verify that miR-130b-3p could mediate GBM cells' function after binding with circAPSM. MTS assays revealed that circAPSM knockdown in GSC18 could inhibit proliferation. However, miR-130b-3p inhibitor treatment resulted in stronger tumor proliferation. In contrast, the pro-tumorigenic effect produced by overexpression of circAPSM completely disappeared after miR-130b-3p mimic treatment in GSC11, and inhibition even occurred compared to the control (Fig. 4j, k). In Edu assays, the proportion of Edu-positive cells that decreased due to knockdown of circASPM increased substantially after miR-130b-3p inhibitor treatment. In contrast, the increase in the proportion of Edu-positive cells due to circASPM overexpression decreased significantly after mimic treatment of miR-130b-3p (Fig. 4m, n, p). Neurosphere formation assays similarly revealed that knockdown treatment of circASPM resulted in GSC18-generated neurospheres with relatiely smaller size and fewer numbers. That neurosphere formation was excessively promoted after treatment with miR-130b-3p inhibitor. In contrast, circASPM overexpression induced GSC11 to produce more and larger neurospheres. After treatment with miR-130b-3p mimic, the size and number of neurospheres were reduced to be even smaller than the control (Fig. 4q, t). Extreme limiting dilution assays also showed a similar trend (Fig. 4r, s). These data suggest that miR-130b-3p suppresses the malignant phenotype of GBM by inhibiting circAPSM expression in GSCs.
CircAPSM regulated E2F1 expression via miR-130b-3p sponging.
Next, we intend to search for functional targets of miR-130b-3p to determine the mechanism by which circAPSM sponging with miR-130b-3p exerts biological function. We determined by StarBase 3.0 that the 3′-UTR of E2F1 has a binding site with miR-130b-3p (Fig. 5a). Luciferase reporter experiments showed that inhibitor treatment of miR-130b-3p triggered an approximate doubling of the luciferase activity of E2F1-wt compared to the E2F1-mt group. In contrast, the luciferase activity of E2F1-wt was significantly decreased after miR-130b-3p mimic treatment (Fig. 5b-e). Both qPCR assays and western blotting showed a significant increase in E2F1 expression after miR-130b-3p inhibitor treatment in GSC11 and GSC12, and a decrease in E2F1 expression after miR-130b-3p mimic treatment in GSC17 and GSC18 (Fig. 5f-k). Finally, we again performed qPCR and western blotting to explore the effect of circASPM on E2F1 expression via miR-130b-3p sponging, and the results demonstrated that knockdown and overexpression treatment of circASPM in GSC18 and GSC11 could positively regulate the expression of E2F1. Meanwhile, the expression level of E2F1 was significantly reversed after rescue experiments with miR-130b-3p inhibitor and mimic treatment, respectively (Fig. 5l,m,n). The above results prove that circAPSM regulates E2F1 expression through miR-130b-3p sponging.
CircAPSM mediated GSCs proliferation via E2F1 expression.
Since circASPM acts as a sponge of miR-130b-3p to regulate E2F1 expression, we further designed a series of experiments to determine whether circASPM up-regulates E2F1 expression in GSCs via the miR-130b-3p-mediated ceRNA mechanism to promote the malignant phenotype of tumors. MTS assays (Fig. 6a, b), EDU assays (Fig. 6c-e), neurosphere formation assays (Fig. 6f, i) and extreme limiting dilution assays (Fig. 6g, h) demonstrated that knockdown treatment of circASPM in GSC18 significantly decreased the viability, proliferation capacity and formation rate of neurospheres. On the contrary, circASPM overexpression treatment of GSC11 up-regulated tumor cells' viability, proliferation capacity, and neurospheres formation rate. Then, we further performed additional E2F1 knockdown or overexpression treatment on the base of circAPSM overexpression or knockdown, respectively. The results showed an over-reversal compared to the levels without the circASPM knockdown or overexpression treatment after manipulating the levels of E2F1. Taken together, these results strongly suggest that circASPM plays a biological role in regulating E2F1 expression through the miR-130b-3p-mediated ceRNA mechanism and promotes the malignant phenotype of GBM by up-regulating E2F1 expression in GSCs.
CircAPSM promoted GSCs tumorigenesis in vivo.
We explored the role of circAPSM in vivo by establishing a tumor xenograft model. We found that knockdown of circAPSM resulted in a significant reduction in GSC18 tumor size (Fig. 7a, c) and a significantly longer median survival time (mean survival: 20.6 ± 6.19 and 29 ± 7.18 days; Fig. 7c). Comparatively, overexpression of circAPSM significantly promoted intracranial tumor growth in GSC11 (Fig. 7d, e) and reduced survival time (mean survival: 25.4 ± 6.50 days, 16.2 ± 6.02 days; Fig. 7f) compared to controls. By further H&E staining and immunohistochemical staining, we found that the staining intensities and expression levels of circAPSM, E2F1 and ki-67 were significantly higher in the circAPSM overexpression group than in the control group, while we obtained opposite results in the knockdown group. The above results demonstrated that circAPSM promoted GSCs tumorigenesis in vivo.