SRSF10 inhibits biogenesis of circ-ATXN1 to regulate glioma angiogenesis via miR-526b-3p/MMP2 pathway

All human normal brain tissues and glioma specimens were obtained from the Department of Neurosurgery of Shengjing Hospital, China Medical University. All participants signed and provided informed consent. In the present study, the research methods were approved by the Institutional Review Board at Shengjing Hospital. The glioma tissues were classified by two experienced clinical pathologists into two groups, that is, LGGs (low grade gliomas: WHO grade I-II) and HGGs (high grade gliomas: WHO grade III-IV), according to the WHO classification.

The immortalized human cerebral microvascular endothelial cell line hCMEC/D3 was provided by Dr. Couraud (Institut Cochin, Paris, France). Cells applied in this study were within 28–32 passages. Endothelial cells (ECs) were cultured in endothelial basal medium (EBM-2; Lonza, Walkersville, MD, USA). Human glioma U87MG and human embryonic kidney 293 T (HEK293T) cell lines were obtained from the Shanghai Institutes and they were cultured in Dulbecco’s modified Eagle’s medium (DMEM) of high glucose with 10% fetal bovine serum. Primary NHA were acquired from the Sciencell Research Laboratories (Carlsbad, CA) and cultured under the conditions instructed by the manufacturer. All cells were maintained in a humidified incubator at 37 °C with 5% CO₂. Glioma conditioned medium and astrocyte conditioned medium were collected as previously described [28]. Astrocyte conditioned medium was used as a negative control (NC).

Cell Counting kit-8 (CCK-8) assay (Beyotime Institute of Biotechnology, Jiangsu, China) was performed to determine the viability of GECs. Cells were seeded in 96-well plates in triplicate and incubated in glioma conditioned medium for 24 h, respectively. Each well was incubated with 10 μl CCK-8 for 2 h and the absorbance was measured at 450 nm using a spectrophotometer (Molecular Devices, United States).

GECs migration in vitro was assessed using a Transwell chamber (Costar, Corning, NY, USA) with a polycarbonic membrane (6.5 mm in diameter and 8 μm pore size). GECs were suspended into single cells in serum-free medium at the density of 5 × 10⁵ cells/ml. 100 μl suspension was added to the upper compartment and 600 μl of glioma conditioned medium supplemented with 10% FBS was added into the lower chamber. Cells were incubated for 48 h at 37 °C. Non-migrating cells on the top surface of membrane were removed with cotton swabs. Cells that migrated to the lower surface of the membrane were fixed with 3:1 methanol: glacial acetic acid, stained with 10% Giemsa solution for 30 min at 37 °C, and washed twice with phosphate buffer saline (PBS). Then the pictures of stained cells were taken with an inverted microscope. Then, stained cells in five randomly fields were randomly chosen for statistics.

To identified circ-ATXN1 and miR-526b-3p rearrangement in GECs, circ-ATXN1 probe (green-labeled; Biosense, Guangzhou, China) and miR-536b-3p probe (red-labeled; Exiqon, Copenhagen, Denmark) were used. Briefly, Slides were blocked with prehybridization buffer (3% BSA in 4× saline-sodium citrate, SSC). Then, the slides were treated with PCR-grade proteinase-k (Roche Diagnostics, Mannheim, Germany). The hybridization mix was prepared with circ-ATXN1 probe or miR-536b-3p probe in hybridization solution. And the slides were washed with washing buffer. Then the sections were stained with anti-digoxin rhodamine conjugate (1:100, Exon Biotech Inc., Guangzhou, China) at 37 °C for 1 h. Subsequently, the sections were stained with 4′,6-diamidino-2-phenylindole (DAPI, Beyotime, China). All fluorescence images (100×) were captured using a fluorescence microscope (Leica, Germany).

RNA binding protein, circ-RNA, and miRNA analysis, Sample preparation and microarray hybridization were performed by Kangchen Bio-tech (Shanghai P.R. China).

HEK293T cells were seeded into a 96-well plate. The potential binding sequence of miR-526-3p in circ-ATXN1 gene and its mutant sequence was amplified by PCR, synthesized and cloned into the pmirGLO dual-luciferase vector (Promega, Madison, WI, USA). Wild-type pmirGLO-circ-ATXN1 (or circ-ATXN1 mutant) reporter plasmid and miR-526-3p agomir (or agomir NC) were co-transfected into HEK293T cells, when they reached 50–70% confluence. The pmirGLO empty vector was transfected as “Control” group. Luciferase activity was measured at 48 h after transfection through the Dual-Luciferase Reporter System (Promega). Wild-type MMP2–3′UTR reporter plasmid (MMP2-wt) and mutated-type MMP2–3′UTR reporter plasmid (MMP2-mut) were constructed with pmirGLO-promoter vector. Following transfection approach and measurement of luciferase activities were performed as described above.

RIP was performed using a Magna RNA-binding protein immunoprecipitation kit (Millipore, Billerica, MA, USA) following the manufacturer’s instructions. Bio-pre-circ-ATXN1-Wt1 (5′ flanking, 5′-500 nt), Bio-pre-circ-ATXN1-Wt2 (3′ flanking, 3′ + 500 nt), and Bio-pre-circ-ATXN1-Wt3 (5′-500 nt + 3′ + 500 nt) were constructed for possible SRSF10 binding sites. In addition, Bio-pre-circ-ATXN1-Mut1 (5′ flanking, 5′-500 nt), Bio-pre-circ-ATXN1-Mut2 (3′ flanking, 3′ + 500 nt), and Bio-pre-circ-ATXN1-Mut3 (5′-500 nt + 3′ + 500 nt) for possible SRSF10 binding sites were constructed respectively. IgG (Millipore) was used as a control. Three wild-type circATXN1 or three mutant circATXN1 of SRSF10 binding site plasmids were transfected into GECs respectively. The Cells were lysed in complete RNA lysis buffer containing magnetic beads conjugated with human anti-SRSF10 antibody. The samples were incubated with Proteinase K and then immunoprecipitated RNA was isolated. Purified RNA was obtained and then applied in qPCR with reverse transcription analysis.

Silencing plasmid of circ-ATXN1 was constructed in pGPU6/Hygro vector (Genechem Co, Shanghai, China). And a non-targeting sequence was used as a NC. Silencing plasmids of SRSF10 (NM_054016.3) with pGPU6/GFP/Neo (GenePharma, Shanghai, China) was constructed. Overexpression plasmid of SRSF10 and circ-ATXN1 were constructed in pIRES/EGFP vector (GenScript, Piscataway, NJ, USA). An empty vector was used as a blank control. GECs were seeded in 24-well plates and transfected using Opti-MEM I and LTX and Plus reagents (Life Technologies) when they were at about 80% confluence. Geneticin (G418; Sigma-Aldrich, StLouis, MO, USA) and Hygromycin (Solarbio, China) were selected. Furthermore, miR-526b-3p agomir (miR-526-3p (+)), miR-526-3p antagomir (miR-526-3p (−)), and their NC sequence (miR-526-3p (+) NC and miR-526-3p (−)NC) were synthesized (GenePharma, Shanghai, China) and transiently transfected into GECs using Opti-MEM and Lipofectamine 3000 reagents (Life Technologies Corporation, Carlsbad, CA, USA), respectively. Cells were collected 48 h after transfection. Sequences of sh-circ-AXTN1, sh-SRSF10 and sh-NC were shown in Table 1. The transfection efficiency of SRSF10, circ-AXTN1, and miR-526b-3p were shown in Supplementary Figure S2A-C.

The gelatinolytic activity was identified by gelatin zymographic analysis, as previously described [29]. In brief, the total of 10 ml protein from the GECs was subjected to electrophoresis. Then, the gels were washed twice using 2.5% Triton X-100 in 50 mM Tris-HCL for 30 min and were incubated in calcium assay buffer overnight at 37 °C. The gels were stained with Coomassie Brilliant Blue in 50% methanol and 10% acetic acid and destained in 10%acetic acid. Finally, the gels were scanned by densitometry using CS Analyzer 3.0.

Total RNA was extracted from the cultured cells with Trizol reagent as described by the manufacturer (Life Technologies Corporation, Carlsbad, CA, USA). One Step PrimeScript™ RT-PCR Kits (Takara, RR064A, Japan) were used for measurement of circ-ATXN1. In addition, RNase-R was used to confirm the existence of circ-ATXN1 and eliminate the influence of liner RNAs. The expression of miR-526b-3p and U6 were detected via TaqMan MicroRNA Reverse Transcription kit and Taqman Universal Master Mix II (Applied Biosystems). One Step SYBR® PrimeScript™ RT-PCR Kit (Takara Biomedical Technology, Dalian, China) was used for the detections of SRSF10 and GAPDH. Their expressions were normalized to endogenous control GAPDH and fold change was determined as 2^−ΔΔCt in gene expression. Primers and probes used in this study are shown in Table 2.

Matrigel assay was performed to evaluate in vitro angiogenesis activity via quantifying tube formation as previously described. In total, 96-well culture plates were coated with 100 μL Matrigel (BD Biosciences, Bedford, MA, USA) per well and then allowed to polymerize for 30 min at 37 °C. Then, cells were resuspended in 100 μl glioma conditioned medium at a density of 4 × 10⁵cells/mL and added to Matrigel-coated wells. After maintained in 37 °C for 6 h, digital camera system (Olympus, Tokyo, Japan) was used to acquire three or more images at random from each culture, and Image J software was used to measure the number of tubes.

Total proteins from the cells were lysed with ice-cold RIPA buffer with protease inhibitors (Beyotime Institute of Biotechnology). Equal amounts of each protein samples (40 μg) was run on SDS/PAGE gels, transferred to PVDF membranes. Membranes were incubated in 5% fat-free milk in TBST and then incubated with primary antibodies against SRSF10 (1:500, Proteintech, Chicago, IL, USA), MMP2 (1:500, Proteintech, Chicago, IL, USA) and GAPDH (1:1000, Proteintech, Chicago, IL, United States).

To measure the angiogenesis, Matrigel plug assay was conformed as previously described [30]. We purchased nude mice from the National Laboratory Animal Center (Beijing, China). Four-week-old BALB/c athymic nude mice were fed with autoclaved water and food during the experiment. All the experiments with the nude mice were performed strictly in according to the protocol approved by the Administrative Panel on Laboratory Animal Care of Shengjing Hospital. Briefly, GECs resuspended in 400 μL of solution containing 80% Matrigel at a density of 3 × 10⁵ cells per ml, and then it were subcutaneously injected. After 4 days, plugs were obtained. Weighed, photographed, and dispersed in 400 μL of PBS at 4 °C overnight incubation to collect the hemoglobin. Drabkin’s solution (Sigma) was conducted to measure hemoglobin content, according to manufacturer’s instructions.

RBPs are involved in the regulation of tumor cells’ biology process. To determine the functional role of RBPs in GECs, we first performed microarray to assess the expression patterns of RBPs in AECs and GECs. As shown in Supplementary Figure S1A, SRSF10 and RNPC1 are two of the most abundant RBPs in GECs. Subsequently, qRT-PCR was used to detect that expression of both SRSF10 and RNPC1 were elevated in GECs with the former being more evident (Supplementary Figure S1B). Further, we established GECs with downregulated SRSF10 to demonstrate the potential roles of SRSF10 in cell viability. Thus, SRSF10 was selected to perform the subsequent analyses. Further, the expression of SRSF10 protein was detected through western blot in brain tissues (normal brain tissues, NBTs), low-grad glioma tissues (LGGTs, Word Health Organization, grad I-II), high-grad glioma tissues HGGTs, Word Health Organization, grad III-IV), as well as in astrocyte-associated endothelial cells (AECs) and GECs. As shown in Fig. 1a, the expression of SRSF10 protein was upregulated in LGGTs group and HGGTs group compared with NBTs. Meanwhile, SRSF10 was significantly increased in GECs compared with AECs (Fig. 1b, P < 0.01). Then we demonstrated the impact of SRSF10 knockdown on the process of migration and tube formation in GECs. CCK-8 assay was used to detect cell proliferation. Interestingly, downregulation of SRSF10 decreased cell viability while SRSF10 overexpression on the basis of SRSF10 knockdown reversed the effect of SRSF10 knockdown on reducing cell viability (Fig. 1c). RNPC1 exerted no effects on cell viability (Supplementary Figure S1C). Transwell assay was conducted to assess migration. There was no significant difference in migration between Control group and sh-NC group. Migration of sh-SRSF10 group was significantly attenuated compared with sh-NC group. And SRSF10 overexpression on the basis of SRSF10 knockdown reversed the effect of SRSF10 knockdown on reducing migration (Fig. 1d, P < 0.01). Matrigel plug assay was conducted to detect tube formation with knockdown of SRSF10. Knockdown of SRSF10 inhibited the tube formation, while SRSF10 overexpression on the basis of SRSF10 knockdown reversed the effect of SRSF10 knockdown on inhibiting tube formation (Fig. 1e, P < 0.01). Subsequently, we assessed the effect of SRSF10 knockdown on MMP2 and VEGFA with western blot assay. The expression of MMP2 and VEGFA in sh-SRSF10 group were both significantly decreased compared with sh-NC group. SRSF10 overexpression on the basis of SRSF10 knockdown reversed the effect of SRSF10 knockdown on reducing MMP2 and VEGFA protein expression (Fig. 1f-g, P < 0.01). Meanwhile, to identify the gelatinolytic activity of MMP2, gelatin zymography was performed. The zymography studies demonstrated that the activity of MMP2 were markedly decreased in sh-SRSF10 group compared with sh-NC group, and SRSF10 overexpression on the basis of SRSF10 knockdown reversed the effect of SRSF10 knockdown on reducing the activity of MMP2. (Fig. 1h, P < 0.05). The above results proved SRSF10 knockdown lowered MMP2 expression and activity to inhibit GEC-induced angiogenesis in vitro.

Using circ-RNA microarray, we found circ-ATXN1 was significantly downregulated after knockdown of SRSF10 in GECs (Supplementary Figure S1D-E). Thus, we hypothesized that circ-ATXN1 was involved in SRSF10-mediated regulation on glioma angiogenesis. Figure 2a showed that the composition of circ-ATXN1 by presenting the genomic DNA sequence of ATXN1 in contrast to circ-ATXN1. The junction sites of circ-ATXN1 CATCCAGAGC and GCCAATTGCA, which were further validated by Sanger sequencing experiment. Subsequently, divergent primers detected the presence of circ-ATXN1 in cDNA while no circ-ATXN1 was detected in the genomic DNA (gDNA) (Fig. 2b). Then, qRT-PCR assay was used to verify that circ-ATXN1 was significantly increased in GECs compared with AECs, while there was no significant difference in linear-ATXN1 expression between GECs and AECs (Fig. 2e-f, P < 0.01). RNase R (an exoribonuclease that degrades linear RNAs instead of circRNAs) was applied to distinguish circ-ATXN1 from linear-ATXN1. Circ-ATXN1 exhibited resistance to RNase R, whereas linear-ATXN1 was degraded after treatment with RNase R (Fig. 2c-d). Thus, circ-ATXN1 instead of linear-ATXN1 may regulate the function of GECs. To investigate the functional role of circ-ATXN1, we further detected the location and expression of circ-ATXN1 by FISH in GECs and AECs and verified that circ-ATXN1 was located in the cytoplasm and its expression was increased in GECs (Fig. 2g). Subsequently, we established glioma cell lines with circ-ATXN1 knockdown or overexpression, then cultured ECs in glioma conditioned medium generated with the above-mentioned cell line. Cell viability, migration and tube formation were analyzed. As shown in Fig. 2h-j, the cell viability, migration and tube formation in sh-circ-ATXN1 group were significantly repressed compared with sh-NC group. Also, circ-ATXN1 overexpression on the basis of circ-ATXN1 knockdown reversed the effect of circ-ATXN1 knockdown on inhibition of cell viability, migration, and tubes formation. Besides, the expression of MMP2 and VEGFA as well as MMP2 activity were markedly attenuated after circ-ATXN1 knockdown. Overexpression of circ-ATXN1 reversed the reduced expression of MMP2 and VEGFA, and the activity of MMP2 (Fig. 2k-m, P < 0.01).

In order to clarify the molecular mechanism of SRSF10 in regulating the angiogenesis of GECs, we conducted RIP analysis to determine the interaction between SRSF10 and circ-ATXN1. Enrichment of Bio-pre-circ-ATXN1-Wt3 in SRSF10 co-immunoprecipitation group was significantly elevated compared with IgG co-immunoprecipitation group. Enrichment of Bio-pre-circ-ATXN1-Wt3 in SRSF10 co-immunoprecipitation group was significantly decreased compared with Bio-pre-circ-ATXN1-Mut3 (Fig. 3a, P < 0.01). We subsequently assessed circ-ATXN1 expression after knockdown of SRSF10 using qRT-PCR. As shown in Fig. 3b, the expression of circ-ATXN1 in sh-SRSF10 group was significantly decreased. Cell lines with co-transfection of SRSF10 and cir-ATXN1 were established. Meanwhile, we detected that the cell viability of GECs in sh-SRSF10 + sh-circ-ATXN1 group was significantly attenuated compared with sh-SRSF10-NC + sh-circ-ATXN1-NC, sh-SRSF10 + sh-circ-ATXN1-NC and sh-SRSF10-NC + sh-circ-ATXN1-NC groups, respectively (Fig. 3c, P < 0.01). Migration and tube formation analysis yielded similar results (Fig. 3d-e, P < 0.01). The above results demonstrated simultaneous knockdown of circ-ATXN1 and SRSF10 intensified inhibition of cell viability, migration and tube formation induced by SRSF10 knockdown. Likewise, the expression of MMP2 and VEGFA, as well as the activity of MMP2 were significantly decreased in sh-SRSF10 + sh-circ-ATXN1 group (Fig. 3f-h, P < 0.01). The above results suggested simultaneous knockdown of SRSF10 and circ-ATXN1 inhibited the expression of MMP2 and VEGFA more significantly. We verified that circ-ATXN1, involved in the regulation of GECs angiogenesis, was regulated by SRSF10.

Using microarray, we found miR-92a and miR-526b-3p were two of upregulated genes in GECs treated with sh-circ-ATXN1 (Supplementary Figure S1F). Meanwhile, qRT-PCR was used to assess that miR-526b-3p expression was higher than miR-92a in GECs treated with sh-circ-ATXN1 (Supplementary Figure 1G). Subsequently, we detected that miR-526b-3p was significantly decreased in GECs compared with AECs (Fig. 4a, P < 0.01). In addition, FISH was used to demonstrate that miR-526b-3p was located in cytoplasm and downregulated in GECs (Fig. 4b). In order to further clarify the effect of miR-526b-3p on regulating biological behaviors of GECs, we established glioma cell lines with overexpression and knockdown of miR-526b-3p. ECs were cultured in glioma conditioned medium generated with the above-mentioned cell lines. Cell viability, migration and tube formation were analyzed. The proliferation of pre-miR-526b-3p group was significantly inhibited compared with pre-NC group and the proliferation of anti-miR-526b-3p group was significantly elevated compared with anti-NC group (Fig. 4c, P < 0.01). Likewise, the migration and tube formation of pre-miR-526b-3p group were significantly inhibited compared with pre-NC group (P < 0.01). Anti-miR-526b-3p group yielded the opposite results compared with anti-NC group (Fig. 4d-e, P < 0.01). The above results indicated that miR-526b-3p suppressed angiogenesis of GECs.

It has been widely reported that miRNAs specifically bound to 3’UTR of mRNAs to inhibit their translation or facilitate their degradation. Western blot assay was conducted to detect MMP2 and VEGFA expression with alteration of miR-526b-3p. In the present study, we measured that MMP2 and VEGFA expression of pre-miR-526b-3p group was significantly inhibited compared with pre-NC group and the expression of anti-miR-526b-3p group was significantly elevated compared with anti-NC group (Fig. 4f-g, P < 0.01). Likewise, the activity of MMP2 in pre-miR-526b-5p group was significantly reduced compared with pre-NC, while MMP2 activity was increased in anti-miR-526b-3p group (Fig. 4h). We predicted 3’UTR of MMP2 and VEGFA harbored binding sites with miR-526b-3p respectively using Targetscan (http://​www.​targetscan.​org/​vert_​72/​) (Fig. 4i and k). Next, dual-luciferase reporter analysis demonstrated firefly/renilla luciferase activity of MMP2–3′-UTR-Wt and miR-526b-3p (+) co-transfection group was significantly decreased compared with MMP2–3′-UTR-Wt + pre-NC group. Similar results were observed in VEGFA (Fig. 4j and l). In summary, miR-526-3p affected angiogenesis of GECs via regulating the expression of MMP2 and VEGFA.

The existence of targeted binding between circ-ATXN1 and miR-526-3p were predicted using Starbase (Fig. 5a), thus we presumed that circ-ATXN1 regulated angiogenesis of GECs via regulating miR-526-3p. Dual-luciferase reporter analysis was conducted to verify the binding between circ-ATXN1 and miR-526-3p. Luciferase activity of circ-ATXN1-Wt and miR-526b-3p(+) co-transfection group was significantly decreased compared with circ-ATXN1-Wt + miR-526b-3p(+)NC group, suggesting circ-ATXN1 specifically bound to the predicted site of miR-526-3p (Fig. 5b, P < 0.01). RIP analysis was conducted to determine whether circ-ATXN1 and miR-526-3p were in the expected Ago2 immunoprecipitation complex. Expression of circ-ATXN1 and miR-526-3p in Ago2 co-immunoprecipitation group were significantly increased compared with IgG co-immunoprecipitation group (P < 0.01). Knockdown of miR-526-3p significantly lowered the enrichment of circ-ATXN1 and miR-526-3p in Ago2 co-immunoprecipitation groups (Fig. 5c-d, P < 0.01). These results suggested that circ-ATXN1 and miR-526-3p were in the expected Ago2 immunoprecipitation complex. We established glioma cell lines stably expressing circ-ATXN1 knockdown with up-regulation and down-regulation of miR-526-3p. ECs were cultured in glioma conditioned medium generated with the above-mentioned cell lines. Knockdown of miR-526-3p reversed the decreased proliferation, migration and tube formation of GECs induced by knockdown of circ-ATXN1 (Fig. 5e-g, P < 0.01). The down-regulated expression of MMP2 and VEGFA induced by knockdown of circ-ATXN1 was reversed after knockdown of miR-526b-3p (Fig. 5h-i, P < 0.01). Meanwhile, the activity of MMP2, which was decreased after knockdown of circ-ATXN1, was reversed by silencing miR-526b-3p (Fig. 5j). The above results demonstrated that miR-526-3p was involved in the regulation of GECs angiogenesis by circ-ATXN1. MiR-526-3p was also critical in the regulation of MMP2 and VEGFA by circ-ATXN1. In this study, we further investigated the effects of co-transfection of knockdown of SRSF10 and knockdown of miR-526b-3p. The results showed that, knockdown of miR-526b-3p reversed the effects of SRSF10 knockdown on the proliferation, migration and tube formation of GECs, as well as the expression of MMP2, VEGFA, and the activity of MMP2 (Supplementary Figure S1H-L).

In our study, Matrigel plug assay was conducted to assess whether SRSF10 and circ-ATXN1 were able to regulate the angiogenesis of GECs in vivo. As shown in Fig. 6a-b, the hemoglobin content in sh-SRSF10 group, sh-circ-ATXN1 group, and pre-miR-526b-3p group, and sh-SRSF10 + sh-circ-ATXN1 + pre-miR-526b-3p group were decreased compared with Control group (P < 0.05). Meanwhile, the hemoglobin content in sh-SRSF10 + sh-circ-ATXN1 + pre-miR-526b-3p group was significantly decreased compared with sh-SRSF10 group, sh-circ-ATXN1 group, and pre-miR-526b-3p group respectively. Thus, the above results indicated that the combination of SRSF10 knockdown, circ-ATXN1 knockdown and miR-526b-3p overexpression presented the strongest inhibitory effect on glioma angiogenesis in vivo.