Circular RNA circGSE1 promotes angiogenesis in ageing mice by targeting the miR-323-5p/NRP1 axis

Cell culture

Mouse aortic endothelial cells (MAECs) were obtained from the aortas of male C57BL/6 mice by trypsin digestion. The cells were cultured in DMEM (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS, Thermo Fisher Scientific) and 1% penicillin/streptomycin in a cell incubator at 37° C and 5% CO₂. Old mice were 18 months old, while young mice were 4 weeks old. Aortic ECs were obtained from 3 mice in the same batch, and cells were used at passages 3-5 in experimental procedures. All mouse studies were consistent with the Helsinki Declaration and were approved by the Experimental Animal Ethics Committee of the First Affiliated Hospital of Sun Yat-Sen University (Approval NO.2020-399).

Whole-transcriptome sequencing

The Hi-Seq platform was used for whole-transcriptome sequencing [21]. The sequencing library was constructed by removing rRNAs from the total RNA of 8 samples (MAECs from 4 young and 4 old mice). Clean reads were obtained by removing adapter sequences, N bases, and low-quality fragments. Then, the TopHat programme (v2.0.14) was used to map the fragments to the mouse genome. The DESeq algorithm was used to analyse differentially expressed genes, circular structures were predicted with Accurate CircRNA Finder Suite (ACFS) for circRNA identification. Gene Ontology and pathway analyses were performed to elucidate the biological significances and important pathways of differentially expressed genes. miRanda tools and RNAhybrid algorithms are used to explore competitive endogenous RNA (ceRNA) relationships.

Cell transfection

CircGSE1 siRNA (RiboBio) and plasmids (Geneseed) was transfected into cells using Lipofectamine 3000 (Thermo Fisher Scientific) in serum-free medium for 6 hours at 37° C cell and 5% CO₂; thereafter, the medium was replaced with medium supplemented with 10% serum. The transfection concentrations of siRNA and Overexpression plasmids were 50 nmol and 2.5 μg/ml, respectively. RNA was extracted 48 hours after cell transfection, and total protein was extracted 48 to 72 hours after cell transfection for subsequent experiments. The sequences of the siRNA used in this study are shown in Supplementary Table 2.

RNA extraction and qRT-PCR

Aortic ECs were collected from each group of mice, and total RNA was extracted using TRIzol reagent (Invitrogen). The concentration and purity of RNA were determined by spectrophotometry, and RNA from a single sample was reverse-transcribed into cDNA using a first-strand cDNA synthesis kit (TaKaRa). qRT-PCR was performed using TaKaRa SYBR Green reagent and Roche Light Cycler 480. GAPDH was used as the internal reference gene for circGSE1 and NRP1, and U6 was used as the internal reference gene for miRNA. qRT-PCRs were repeated three times, and differences were calculated by the 2^-ΔΔCt method. All primer sequences are listed in Supplementary Table 1.

RNase R experiment

RNA (10 μg) from MAECs was incubated at 37° C with RNase R (3 U/μg, Geneseed) for 30 minutes. Then, the treated RNA was reverse transcribed using divergent primers or polymeric primers, and nucleic acid electrophoresis was performed after qRT-PCR detection.

RNA immunoprecipitation

RNA immunoprecipitation (RIP) assay was performed using Magna RIP™ RNA-binding protein immunoprecipitation kit (Millipore). OECs was incubated with magnetic beads conjugated with anti-Argonaute 2 (AGO2) or anti-IgG antibody (Millipore) for 6 h at 4° C. The beads were washed and incubated with Proteinase K to remove proteins. Finally, isolated RNA was extracted using TRIzol Reagent (Takara), then, the purified RNA was subjected to agarose gel electrophoresis and qRT-PCR analysis.

Nucleic acid electrophoresis

cDNA and gDNA PCR products were detected using 3% agarose gel electrophoresis in TAE buffer solution at 100 V for 45 minutes, and DNA markers were used to identify 100-500bp fragments (Accurate Biology). The gel was illuminated with an ultraviolet detector.

Fluorescence in situ hybridization (FISH)

The localization of circGSE1 and miR-323-5p in the vascular ECs was detected by FISH. The circGSE1 probe labelled with CY3 and the miR-323-5p probe labelled with FITC were designed and synthesized by Geneseed Biology. The circGSE1 and miR-323-5p probes were hybridized overnight according to the manufacturer’s instructions. Images were captured by a Zeiss LSM710 laser scanning confocal microscope (Zeiss Instruments). The immunofluorescence signals of circGSE1 and miR323-5p were quantified with software Fiji setting threshold automatically by Otsu filter. Colocalization values (Pearson, Manders’ coefficients) were calculated from eight representative cell images using the colocalization plugin JACoP (Just another colocalization plugin) [22]. All the experiments were repeated three times. The sequences of circGSE1 and miR-323-5p are listed in Supplementary Table 2.

Western blotting

WB was used to determine the relative expression levels of NRP1. ECs were incubated under normal culture conditions for 48-72 hours and then frozen in RIPA buffer (Thermo Fisher Scientific). Protein concentrations were determined using the BCA protein assay kit (Thermo Fisher Scientific). Equivalent (20 μg) amounts of protein were separated by 10% SDS-PAGE and transferred to a PVDF membrane (Millipore,). After blocking with TBST supplemented with 5% bovine serum albumin (BSA) (BioFroxx), the PVDF membrane was incubated with anti-rabbit NRP1 primary antibodies at 4° C and then with an anti-rabbit secondary antibody. NovexECL (Invitrogen) was used for detection, and ImageJ software (version 1.8.0) was used for signal quantification. The Primary antibodies are listed in Supplementary Table 3.

Luciferase assays

CircGSE1 sequences with a mutated miR-323-5p-binding site were obtained using a QuikChange II site-directed mutation kit (Stratagene). Wt-miR-323-5p and mut-miR-323-5p were designed by Giesee. The negative control (NC) or circGSE1 sequence was cotransfected with wt-miR-323-5p or mut-miR-323-5p into HEK-293T cells using Lipofectamine 3000 (Thermo Fisher Scientific). Twenty-four hours after transfection, the luciferase efficiency was measured by a dual-luciferase reporter and detection kit (TransGen Biotech). All the experiments were repeated three times.

In vitro wound healing assay

Transfected cells seeded in 6-well plates were cultured until 90% confluence. A 1-ml sterile pipette tip was used to generate a scratch at a point relative to the reference, after which the cells were washed with sterile PBS and observed under a microscope (Olympus Co., Ltd.) to obtain reference images. Then, the cells were cultured in serum-free medium for 24 hours, and a second set of images of the same area was obtained. ImageJ software was used for the statistical analysis of scratch healing.

Transwell assay

A chamber (Corning Life Sciences) with a pore diameter of 8 μm was inserted into a 24-well plate. Then, 500 μl of medium containing 10% FBS was added to the lower chamber, and the treated ECs as well as 300 μl of serum-free medium were added to the upper chamber for 24 hours. After the culture period ended, 4% paraformaldehyde for 20 minutes and then attained with crystal violet was used for staining for 15 minutes. The chamber wall was then gently wiped with a cotton ball, and the migrated cells were counted in five random areas using an optical microscope (Olympus Co., Ltd).

Tube formation experiment

A 24-well plate was precooled at 4° C overnight, filled with 200 μl of Matrigel matrix (Corning Life Sciences), incubated at 4° C for 6 minutes, and placed in a cell culture incubator for 30 minutes. Then, 150000 cells/well were added to the solidified matrix, placed in a cell culture incubator for 6 hours, and imaged using an optical microscope. ImageJ software was used for tube length measurements.

Cell proliferation evaluation via the EdU assay

ECs were inoculated into a 96-well plate and cultured normally until reaching. 60-70% confluence, after which a c10310-1/-2/-3 kit (RiboBio) was used to assess the proliferation of cultured cells according to the instructions, and a fluorescence microscope (Leica) was used assess their proliferative by measuring the proportion of EdU-positive cells.

Animal model

C57BL/6 mice (GemPharmatech Co., Ltd) ageing 18 months and 4 weeks were used in the experiment. All procedures were approved by the Animal Ethics Committee of the First Affiliated Hospital of Sun Yat-Sen University, and animal models of unilateral and hindlimb ischemia were established based on previous studies. The left groin area of the mouse was anaesthetized with isoflurane, and a 1-cm oblique incision was made. After the subcutaneous tissue was separated, the femoral artery sheath was opened, the femoral vein and femoral nerve were separated, and the femoral artery was located. The femoral artery was ligated with 7-0 silk thread, and removed, and the incision was sutured. The right hind leg served as a control [23]. Two weeks before the surgery, the mice were divided into four groups as follows: elderly mice injected with AAV9-circGSE1, elderly mice injected with AAV9-NC, elderly mice injected with PBS and young mice injected with PBS. In the two-week before preoperative, AAV9 (30 μl, 10¹⁰ vg in each leg) and PBS (30 μl, each leg) were injected into the bilateral posterior calf gastrocnemius muscle [24]. The right hind limb was the control. Blood flow was observed pre- and postoperatively at 0, 3, 7, 14, 21 and 28 days using an animal blood flow imaging perfusion apparatus. On the 28th day after the operation, the mice were sacrificed by the intraperitoneal injection of high-dose pentobarbital, and gastrocnemius muscle immunohistochemistry, qPCR, and other experiments were performed.

Blood flow measurements

Blood flow measurements were performed using a small-animal blood flow imaging perfusion machine. At 0, 3, 7, 14, 21 and 28 days after surgery, blood flow was measured by PeriCamPS I (Perimed) in a constant temperature, humidity and sound insulation environment (room temperature of 24° C, no excessive light and noise). The imaging data were analysed using PIMSoft (Perimed) software [24], and the ratio of ischemic limb blood flow to normal limb blood flow was used as the index of ischemic limb blood flow recovery.

Immunohistochemistry

Mouse gastrocnemius muscle sections (3 μm) were treated with a hydrogen peroxide blocker and incubated with primary anti-CD31 and anti-NRP1 antibody. After overnight incubation and washing, the muscle sections were incubated with a secondary antibody, exposed to 3,3-diaminobenzidine reagent (DAB; Solarbio), and stained with haematoxylin. The muscle slices were observed and imaged using an Olympus BX51 microscope (Olympus).

Statistical analysis

All data are presented as the mean ± standard deviation (SD). GraphPad Prism 8.0.1 was used for statistical analysis. Independent sample t-tests were used for comparisons between two groups, and one-way analysis of variance (ANOVA) was used for comparisons among three or more groups.

Data availability statement

The data that support the results of this study are available from the corresponding author upon reasonable request.

CircRNA sequencing identifies differentially expressed circRNAs in endothelial cell aging

To study ageing EC-related angiogenesis, we performed whole-transcriptome sequencing. We identified 26 significantly differentially expressed circRNAs by comparing YEC and OEC samples. The circogram shows the distribution of the detected and significantly expressed circRNAs on mouse chromosomes (Figure 1A). We statistically analysed the length distributions statistics of the detected circRNAs, which indicated that most of the circRNA were less than 1000 nt in length (Figure 1B). Regarding their genomic origins, most of the circRNAs originated from exons (Figure 1C). The volcano plot shows the detected and significantly expressed circRNAs, with red and blue dots representing significantly up- and downregulated circRNAs in YECs compared with OECs, respectively (Figure 1D).

Figure 1. Profiling of circRNAs in old endothelial cells (OECs) and young endothelial cells (YECs). (A) The circle shows the distribution and expression of differentially expressed circRNAs and mRNAs on mouse chromosomes. The outer, middle and inner layers represent the mouse chromosomes, mRNAs and circRNAs, respectively. Red represents upregulation, and green represents downregulation in YECs. (B) The distribution of circRNAs of different lengths in YECs compared with OECs. (C) Different mouse genomic sources of differential circRNAs. (D) Volcano map of circRNAs in YECs compared with OECs. Red represents upregulation, green represents downregulation, and blue represents no significant difference. (E) The heat map shows the differentially expressed circRNAs in OECs and YECs. The red bars represent upregulation, and the blue bars represent downregulation. (F) qRT-PCR shows ten of the most significant differences circRNAs expression in YECs compared with OECs; the expression was normalized to that of GAPDH (n=3). (G) qRT-PCR analysis of circGSE1 expression in the lower limb gastrocnemius muscles of mice with femoral artery ligation ischaemia; expression was normalized to that of GAPDH (n=3). (The data are expressed as the mean ± SD, *P < 0.05, **P < 0.01 versus the negative control).

CircGSE1 is significantly downregulated in MAECs of ageing mice and ischemic muscles

The heat map shows differentially expressed circRNAs in OECs and YECs, and circGSE1 was significantly upregulated in YECs compared with OECs (Figure 1E). We used qRT-PCR to verify the 10 most significantly different circRNAs (5 upregulated and 5 downregulated), which were basically consistent with the sequencing results. And we found circGSE1 was the most significantly upregulated in YECs (Figure 1F). Meanwhile, we found that circGSE1 was expressed at significantly lower levels in the ischemic gastrocnemius muscle samples than that in control samples (Figure 1G).

CircGSE1 characteristics and differences in organization

Our study confirmed the head-to-tail splicing of circGSE1 in the circGSE1 PCR product by Sanger sequencing and determined its genome size and sequence, as reported by the circBase database (Figure 2A). As a result, we also designed convergent primers for GSE1 RNA and specific divergent primers for the amplification of GSE1. cDNA and gDNA were extracted from OECs and analysed by PCR and agarose gel electrophoresis, revealing that circGSE1 was only amplified from cDNA, and not from gDNA (Figure 2B). We also employed RNase R to confirm the stability of circGSE1 using specially designed divergent and convergent primers for PCR and agarose gel electrophoresis. CircGSE1 resisted digestion by RNase R, while linear GSE1 and GAPDH were sensitive to RNase R. In addition, qRT-PCR was used to quantify the expression levels of the back spliced and typical forms of GSE1 with or without RNase R in the cDNA and gDNA of OECs (Figure 2C, 2D). FISH experiments showed that circGSE1 was mostly located in the cytoplasm (Figure 2E). We detected the relative expression of circGSE1 in different mouse organs by qRT-PCR, revealing that circGSE1 expression was higher in muscle, artery and splenic tissues than that in other organs (Figure 2F).

Figure 2. Validation of circGSE1 and differences in organization. (A) Sanger sequencing of the band amplified with divergent primers confirms head-to-tail splicing of circGSE1. (B) Agarose gel electrophoresis validated the expression of circGSE1 in MAECs. CircGSE1 was amplified by divergent primers in cDNA but not gDNA, and GAPDH was used as a negative control. (C, D) In the presence or absence of RNase R, the expression of circGSE1 and linear GSE1 RNA in MAECs was detected by qRT-PCR, and nucleic acid electrophoresis or qRT-PCR was then performed. (E) FISH experiments proved that most circGSE1 is localized in the cytoplasm. Nuclei were stained blue, and cytoplasmic circGSE1 was stained red. (magnification, 400×, scale bar, 20 μm). (F) The circGSE1 expression levels in different mouse organs were analysed by qRT-PCR, revealing higher expression in muscle, artery and splenic organs. (The data are expressed as the mean ± SD, *P < 0.05, **P < 0.01 versus the negative control).

CircGSE1 promotes the proliferation, migration and tube formation of MAECs in vitro

To explore the function of circGSE1 in MAECs, we designed three siRNAs according to the splice site of circGSE1 to inhibit its expression and designed a circGSE1 overexpression vector (Figure 3A). We used qRT-PCR assay to detect the interference effect and found that all three siRNAs could knock down circGSE1, while GSE1 mRNA did not change. We chose to utilize si-circGSE1-2 for subsequent experiments because it had the strongest inhibitory effect (Figure 3B, 3C).

Figure 3. CircGSE1 promotes the proliferation, migration and tube formation of MAECs in vitro. (A) Schematic diagram of the circGSE1 siRNA and overexpression vectors. (B) qRT-PCR detection of YECs demonstrating successful knockdown of circGSE1; however, the linear GSE1 RNA expression was not altered significantly (n=3). (C) qRT-PCR detection demonstrating successful circGSE1 overexpression in OECs, while the linear GSE1 RNA expression was not changed significantly (n=3). (D) Wound healing assay of OECs treated with vector and OE-circGSE1, YECs treated with NC and si-circGSE1 (n=3, scale bar, 200 μm). (E) Tube formation assay of OECs treated with vector and OE-circGSE1, YECs treated with NC and si-circGSE1 (n=3, scale bar, 200 μm). (F) Migration assay of OECs treated with vector and OE-circGSE1, YECs treated with NC and si-circGSE1 (n=3, scale bar, 200 μm). (G) EdU assay of OECs treated with vector and OE-circGSE1, YECs treated with NC and si-circGSE1 (n=3, scale bar, 200 μm). (n=5, scale bar, 100 μm). (The data are expressed as the mean ± SD, *P < 0.05, **P < 0.01 versus the negative control).

We also conducted a variety of functional experiments to explore the role of circGSE1 in angiogenesis. The wound healing assay showed that compared with the empty vector overexpression and blank control groups, the circGSE1 overexpression group exhibited an enhanced OEC migration ability, while the YEC migration in the circGSE1-knockdown group was inhibited compared with that in the siRNA-NC and the blank control groups, (Figure 3D). The tube formation assay demonstrated that compared with that in the empty vector overexpression and blank control groups, the OEC tube formation ability in the circGSE1 overexpression group was enhanced, while that of YECs in the circGSE1-knockdown group was inhibited compared with that in siRNA-NC and blank control groups (Figure 3E). The migration assay showed that compared with that in the empty vector overexpression and blank control groups, the OEC migration ability in the circGSE1 overexpression group was enhanced, while that of YECs in the circGSE1-knockdown group was inhibited compared with that in the siRNA-NC and blank control groups (Figure 3F). The EdU assay indicated that compared with that in the empty vector overexpression and blank control groups, the OEC proliferation in the circGSE1 overexpression group was enhanced, while that of YECs was inhibited, in the circGSE1- knockdown group compared with that in the siRNA-NC and blank control groups (Figure 3G). These functional experiments show that circGSE1 can promote the proliferation, migration and tube formation of MAECs, thereby promoting angiogenesis.

miR-323-5p is the downstream target of circGSE1

To further explore the specific mechanism by which circGSE1 promotes angiogenesis, we used three databases (TargetScan, miRanda and RNAhybrid) to predict the downstream microRNA targets of circGSE1. The prediction results showed that 8 microRNAs are potentially downstream targets of circGSE1, miR-323-5p, miR-320-3p, miR-6973a-5p, miR-6924-5p, miR-6769b-5p, miR-6943-5p, miR-7036b-3p, and miR-6917-5p (Figure 4A). We used qRT-PCR to quantify the relative expression levels of the 8 predicted microRNAs in OECs overexpressing circGSE1 and in YECs after siRNA-mediated knockdown of circGSE1, and only miR-323-5p was differentially expressed (Figure 4B, 4C). We also found that miR323-5p was upregulated in OECs compared with YECs (Figure 4D). Meanwhile, considering that circRNAs could serve as miRNA sponges in the cytoplasm, FISH experiments showed that circGSE1 and miR-323-5p were mostly colocalized in the cytoplasm (Figure 4E, 4F). In addition, bioinformatic analysis was used to identify the portion of the circGSE1 sequence that may bind to miR-323-5p (Figure 4G).

Figure 4. miR323-5p is the downstream target of circGSE1. (A) The downstream miRNA targets of circGSE1 as predicted by the TargetScan, miRanda and RNAhybridB databases. (B) The relative OEC expression of the downstream miRNA target of circGSE1 predicted by the database as determined by qRT-PCR. (C) The relative YEC expression of the downstream miRNA target of circGSE1 predicted by the database as determined by qRT-PCR. (D)The relative of miR323-5p expression in YEC and OEC. (E, F) The FISH experiment showed that most circGSE1 (red) and miR-323-5p (green) are colocalized in the cytoplasm (magnification, 400×, scale bar, 20 μm) (n=3, analysed 8 cells). (G) Schematic diagrams of the circGSE1-WT and circGSE1-mutant dual-luciferase reporter vectors. (H) Dual-luciferase reporter assays showed that the circGSE1 wt 3’UTR but not the mut 3’UTR was downregulated by miR-323-5p mimics or upregulated by the miR-323-5p inhibitor. (I, J) The Anti-Ago2 RIP assay was executed in OEC, followed by nucleic acid electrophoresis and qRT-PCR to detect circGSE1. (The data are expressed as the mean ± SD, *P < 0.05, **P < 0.01 versus the negative control).

To verify the binding of circGSE1 and miR-323-5p, we synthesized circGSE1-wt-psiCHECK2 and circGSE1-mut-psiCHECK2 dual-luciferase assay vectors. Herein, circGSE1 wt 3’UTR but not the mut 3’UTR was downregulated by miR-323-5p mimics or upregulated by the miR-323-5p inhibitor, which suggested that a direct interaction occurs between circGSE1 and miR-323-5p (Figure 4H). It was widely acknowledged that miRNAs regulated target mRNA expression by binding to Argonaute 2 (Ago2), an essential component of RNA-induced silencing complex (RISC). Subsequently, RNA immunoprecipitation (RIP) assays were performed in OECs to pull down the RNA transcripts that bound to Ago2. Results of agarose gel electrophoresis and qRT-PCR assays indicated that circGSE1 was pulled down by anti-Ago2 as compared with those in the input control (Figure 4I, 4J).

MiR323-5p inhibits the proliferation, migration and tube formation of MAECs

To further study the interaction among circGSE1 and miR-323-5p, we conducted a series of functional experiments. Wound healing experiments showed that the miR-323-5p inhibitor enhanced the migration ability of OECs, but siRNA targeting circGSE1 suppressed this effect. Moreover, the miR-323-5p mimic decreased the migration ability of YECs, but this inhibitory effect was reversed by co-transfection of OE-circGSE1 (Figure 5A). Tube formation experiments showed that the miR-323-5p inhibitor enhanced the tube formation ability of OECs, while the siRNA targeting circGSE1 suppressed this effect. Moreover, miR-323-5p mimics decreased the tube formation ability of YECs, but this inhibitory effect was reversed by co-transfection of OE-circGSE1 (Figure 5B). Migration experiments showed that inhibiting miR-323-5p enhanced the migration ability of OECs, but co-transfection of the miR-323-5p inhibitor with the siRNA targeting circGSE1, decreased their migration ability. Overexpressing miR-323-5p via mimics decreased the migration ability of YECs, while co-transfection with OE-circGSE1 increased their migration ability (Figure 5C). EdU experiments showed that inhibiting miR-323-5p enhanced the proliferation of OECs, while co-transfection of the miR323-5p inhibitor with siRNA targeting circGSE1 decreased the proliferation of OECs. Overexpressing miR-323-5p via mimics decreased the proliferation ability of YECs, but co-transfection with OE-circGSE1 increased their proliferation (Figure 5D).

Figure 5. CircGSE1 regulates the angiogenic ability of MAECs via the miR-323-5p. (A) Wound healing assay of OECs treated with vector, the miR-323-5p inhibitor or the miR-323-5p inhibitor + circGSE1 siRNA and YECs treated with the NC, miR-323-5p mimic or miR-323-5p mimic+OE-circGSE1. The graph indicates the quantification of the wounded area 24 hours after scratching (n=3, scale bar: 200 μm). (B) Tube formation assay of OECs treated with the vector, miR-323-5p inhibitor or miR-323-5p inhibitor+circGSE1 siRNA and YECs treated with the NC, miR-323-5p mimic or miR-323-5p mimic+OE-circGSE1 (n=3, scale bar, 200 μm). (C) Migration assay of OECs treated with the vector, miR-323-5p inhibitor or miR-323-5p inhibitor + circGSE1 siRNA and YECs treated with the NC, miR-323-5p mimic or miR-323-5p mimic+OE-circGSE1 (n=5, scale bar, 200 μm). (D) EdU assay of OECs treated with the vector, miR-323-5p inhibitor or miR-323-5p inhibitor + circGSE1 siRNA and YECs treated with the NC, miR-323-5p mimic or miR-323-5p mimic+OE-circGSE1 (n=5, scale bar, 200 μm). (The data are expressed as the mean ± SD, *P < 0.05 **P < 0.01 versus the negative control).

NRP1 is the downstream target of miR-323-5p

To continue searching for the downstream targets of circGSE1 and miR-323-5p and explore the specific mechanisms by which they promote angiogenesis, we used three databases (TargetScan, RNAhybrid and Miranda) to jointly predict the downstream targets of miR-323-5p, found 85 mRNAs possibly to be downstream of miR323-5p. Through bioinformatics, it was found that NRP1 had a strong binding with miR323-5p. At the same time, this protein has been reported to play a role in angiogenesis [25–28] and was thus selected for further experiments.

Through WB assay and PCR assay, we found that NRP1 expression was up-regulated in YEC compared with OEC (Figure 6A, 6B). The expression of NRP1 was increased after transfection of a miR-323-5p inhibitor into OECs, and its expression was decreased after transfection of miR-323-5p mimics into YECs, indicating that miR-323-5p can regulate the expression of NRP1 in ECs (Figure 5D). In addition, bioinformatic analysis revealed the sequence of miR-323-5p that may bind to NRP1 (Figure 6F). To verify the direct binding of miR-323-5p to NRP1, we synthesized NRP1-wt-psiCHECK2 and NRP1-mut-psiCHECK2 dual-luciferase reporter gene vectors. The NRP1 wt 3’UTR but not the mut 3’UTR was downregulated by the miR-323-5p mimics and upregulated by the miR-323-5p inhibitor, which indirectly proved that circGSE1 bind to miR-323-5p (Figure 6G). Then, qRT-PCR and western blot assays demonstrated that overexpression of circGSE1 significantly counteract the mRNA and protein level of NRP1, whereas the co-transfection of si-circGSE1 and miR323-5p inhibitor might eliminate effects in YECs. Similarly, it was also found that the knockdown of circGSE1 significantly decreased the mRNA and protein level of NRP1, whereas co-transfection of OE-circGSE1 and miR323-5p mimics might increase these effects in YECs (Figure 6H, 6I). In conclusion, these data demonstrated that circGSE1 acts as a sponge for miR-323-5p, thus enhancing NRP1expression in MAECs.

Figure 6. NRP1 is the downstream target of miR323-5p. (A, B) Western blot analysis of the NRP1 expression in OECs and YECs. (C) qRT-PCR analysis of the NRP1 mRNA expression in OECs and YECs. (D) Western blot analysis of the NRP1 expression in OECs transfected with negative control and miR323-5p inhibitor, YECs transfected with negative control and miR323-5p mimics. (E) qRT-PCR analysis of the NRP1 mRNA expression in YECs transfected with negative control and miR323-5p mimics, OECs transfected with negative control and miR323-5p inhibitor. (F) Schematic diagrams of the circGSE1-WT and circGSE1-mutant dual-luciferase reporter vectors. (G) Dual-luciferase reporter assays showed that the circGSE1 wt 3’UTR but not the mut 3’UTR was downregulated by miR-323-5p mimics or upregulated by the miR-323-5p inhibitor. (H, I) WB analyses of OECs treated with the vector, OE-circGSE1 or OE-circGSE1 + miR323-5p mimic and YECs treated with the NC, si-circGSE1 or si-circGSE1+miR323-5p inhibitor. (The data are expressed as the mean ± SD, *P < 0.05, **P < 0.01 versus the negative control).

CircGSE1 regulates the angiogenic ability of mouse ECs via the miR-323-5p/NRP1 axis

To further study the interaction among circGSE1, miR-323-5p and NRP1, we conducted a series of functional experiments. Wound healing experiments showed that the OE-NRP1 enhanced the migration ability of OECs, but siRNA targeting circGSE1 suppressed this effect. Moreover, the siRNA targeting NRP1 decreased the migration ability of YECs, but this inhibitory effect was reversed by co-transfection of OE-circGSE1 (Figure 7A). Tube formation experiments showed that the OE-NRP1 enhanced the migration ability of OECs, but siRNA targeting circGSE1 suppressed this effect. Moreover, the siRNA targeting NRP1 decreased the migration ability of YECs, but this inhibitory effect was reversed by co-transfection of OE-circGSE1 (Figure 7B). Migration experiments showed that overexpressing NRP1 enhanced the migration ability of OECs, but co-transfection of the OE-NRP1 plasmid with the siRNA targeting circGSE1, decreased their migration ability. Inhibiting NRP1 via siRNA decreased the migration ability of YECs, while co-transfection with OE-circGSE1 increased their migration ability (Figure 7C). EdU experiments showed that overexpression NRP1 enhanced the proliferation of OECs, while co-transfection of the siRNA targeting circGSE1 decreased the proliferation of OECs. Inhibiting circGSE1via siRNA decreased the proliferation ability of YECs, but co-transfection with OE-circGSE1 increased their proliferation (Figure 7D).

Figure 7. CircGSE1 regulates the angiogenic ability of MAECs via the NRP1. (A) Wound healing assay of OECs treated with vector, the OE-NRP1 or the OE-NRP1 + circGSE1 siRNA and YECs treated with the NC, NRP1-siRNA or NRP1-siRNA +OE-circGSE1. The graph indicates the quantification of the wounded area 24 hours after scratching (n=3, scale bar: 200 μm). (B) Tube formation assay of OECs treated with the vector, the OE-NRP1 or the OE-NRP1 + circGSE1 siRNA and YECs treated with the NC, NRP1-siRNA or NRP1-siRNA +OE-circGSE1 (n=3, scale bar, 200 μm). (C) Migration assay of OECs treated with the vector, the OE-NRP1 or the OE-NRP1 + circGSE1 siRNA and YECs treated with the NC, NRP1-siRNA or NRP1-siRNA +OE-circGSE1 (n=5, scale bar, 200 μm). (D) EdU assay of OECs treated with the vector, the OE-NRP1 or the OE-NRP1 + circGSE1 siRNA and YECs treated with the NC, NRP1-siRNA or NRP1-siRNA +OE-circGSE1 (n=5, scale bar, 200 μm). (The data are expressed as the mean ± SD, *P < 0.05 **P < 0.01 versus the negative control).

CircGSE1 can promote blood flow recovery and angiogenesis in the ischemic limbs of ageing mice

To further study the role of circGSE1, we successfully established acute lower limb ischemia models in ageing mice (C57BL/6, male,18 months old) and young mice (C57BL/6, male,4 weeks old). The ageing mice were randomly divided into 3 groups (PBS, negative control (NC), and overexpression (OE)), and a young mouse group were included (see the methods section for details). On 7th, 14th, 21st, and 28th days after the operation, the blood flow recovery in the lower limbs of the young mouse group and the OE-circGSE1 ageing mouse group was better than that in the NC ageing mouse group and PBS group, demonstrating that circGSE1 can promote angiogenesis in the ischemic lower limbs of ageing mice (Figure 8A, 8B). We also used qRT-PCR to verify the expression of circGSE1 and miR-323-5p in the ischemic lower limbs gastrocnemius muscles of mice in the different groups, revealing that circGSE1 was overexpressed in young and old mice of the circGSE1 group. The level of miR-323-5p in the OE-circGSE1 group was significantly higher than that in the NC and PBS groups in ageing mice, while miR-323-5p was expressed at significantly lower level in the young mouse and OE-circGSE1 ageing mouse group than in the NC and PBS ageing mouse group (Figure 8C). To assess the pathological changes in the ischemic tissue, we prepared tissue sections of the ischemic lower limb gastrocnemius muscles of different groups of mice 28 days after the operation and stained them with anti-CD31 and anti-NRP1. The expression levels of CD31 and NRP1 in the young mouse and circGSE1 OE groups were significantly higher than those in the NC and PBS groups of ageing mice (Figure 8D, 8E).

Figure 8. Overexpression of circGSE1 can promote blood flow recovery and angiogenesis in ageing mice. (A) Imaging of blood flow through the mouse femoral artery 3, 7, 14, 21, and 28 days after surgery (n=5). (B) The blood flow through the mouse ischaemic limbs was detected, and the results are expressed as the ratio of the perfusion of the left limb (ischaemic) to that of the right limb (control, nonischaemic) on the 7th, 14th, and 21st days after the surgery (n=5). (C) qRT-PCR analysis of circGSE1 and mi323-5p expression in the ischaemic left lower limb gastrocnemius muscles in the young mouse group, old mouse circGSE1 overexpression group and old mouse PBS group compared to the old mouse NC group; expression levels were normalized to that of GAPDH. (D, E) Representative images of HE, anti-CD31 and anti-NRP1 immunohistochemical staining of mouse left lower limb gastrocnemius muscles. The CD31 and NRP1 expression levels in the young mouse group, old mouse circGSE1 overexpression group and old mouse PBS group were compared to those in the old mouse NC group (n=5, scale bar, 100 μm) (The data are expressed as the mean ± SD, *P < 0.05, **P < 0.01). (F) Schematic diagram of the mechanism by which circGSE1 promotes angiogenesis in ageing mice.

The above experiments proved that the circGSE1/miR-323-5p/NRP1 axis exists in MAECs. Overall, these results suggest that circGSE1 acts as a miR-323-5p sponge to promote NRP1 expression and thus promote angiogenesis (Figure 8F).