miRNAs are important small ncRNAs with a length of 20–22 nucleotides. Classically, miRNAs bind to Ago2 protein to form an RNA-induced silencing complex, which then recognizes and binds to the target mRNA by complementary base pairing, leading to the degradation of the target mRNA (Yang F. et al., 2017). Numerous studies have indicated the critical roles of miRNAs in glioma angiogenesis. For example, in gliomas relative to normal brains, miR-296 is elevated in endothelial cells and directly targets hepatocyte growth factor-regulated tyrosine kinase substrate (HGS). Because HGS mediates the degradation of VEGFR2 and PDGFRβ, miR-296 upregulates VEGFR2 and PDGFRβ to promote angiogenesis (Wuerdinger et al., 2008). Aside from miR-296, other miRNAs including miR-93 and miR-675-5p have been reported to promote glioma angiogenesis (Fang et al., 2011; Tchaicha et al., 2011; Lo Dico et al., 2016). In contrast, some miRNAs such as miR-124-3p, miRNA-205, and miR-128 may play inhibitory roles in regulating vascularization (Shi et al., 2012; Yue et al., 2012; Adlakha and Saini, 2014; Zhang et al., 2018). The miRNAs that have been reported to regulate angiogenesis in glioma are summarized in Table 1.

lncRNAs are transcripts longer than 200 nucleotides without open reading frames (Zhao W. et al., 2019). The abnormal expression of functional lncRNAs in gliomas suggests the involvement of lncRNAs in the occurrence, development, and other malignant phenotypes of gliomas (Peng et al., 2018). lncRNAs may function through the following four main mechanisms. First, lncRNAs bind to and regulate the modification, stability, localization, and interaction of target proteins. The NF-κB-interacting lncRNA NKILA is significantly upregulated in gliomas, and higher NKILA levels are correlated with poorer patient prognosis. NKILA functions by upregulating HIF-1α expression and the activity of the hypoxia pathway to enhance the Warburg effect and glioma angiogenesis (Chen et al., 2020). Second, lncRNAs bind to long-stranded RNA molecules, including mRNAs, lncRNAs, pre-mRNAs, and pre-miRNAs, to regulate the stability and translation of bound RNAs. For example, ANKHD1 and LINC00346 are elevated, whereas ZNF655 is reduced in glioma-associated endothelial cells. ANKHD1 binds to and enhances the stability of LINC00346, which in turn promotes the degradation of ZNF655 mRNA. ZNF655 functions to target the promoter of ANKHD1. Thus the ANKD1/LINC00346/ZNF655 feedback loop regulates glioma angiogenesis (Yang et al., 2020). Third, lncRNAs bind to miRNAs and relieve their inhibitory effects on downstream target genes. lncRNA H19 plays an important role in GBM by up-regulating the expression of the angiogenic factor VASH2 through inhibition of miR-29a (Jia et al., 2016; Jiang et al., 2016). In addition, H19 promotes angiogenesis through the miR-342/Wnt5a/β-catenin axis and the miR-138/HIF-1α axis (Liu Z. Z. et al., 2020; Zhou et al., 2020). Finally, lncRNAs bind to genomic DNAs to regulate gene transcription. For instance, SLC26A4-AS1 recruits NFKB1 to promote NPTX1 transcription, which exerts anti-angiogenic effects on glioma cells (Li et al., 2021). Whereas many studies have shown the promotion of angiogenesis by several lncRNAs (NEAT, HULC, SNHG16, linc00667, SNHG15, PVT1, etc.) through various downstream pathways (Jia et al., 2016; Zhu et al., 2016; Ma et al., 2017a,b; Zhang et al., 2018; Wang C. et al., 2019; Wang D. et al., 2019; Xu H. et al., 2019; Chen et al., 2020; Liu Z. Z. et al., 2020; Yang et al., 2020; Zhou et al., 2020), lncRNAs may negatively regulate angiogenesis. For example, lncRNA SLC26A4-AS1 suppresses angiogenesis by upregulating NPTX1 via NFKB1 transcriptional factor (Li et al., 2021), while LINC00320 inhibits angiogenesis by downregulating NFKB1-mediated AQP9. (Chang et al., 2020). Table 2 summarizes the lncRNAs participating in the regulation of glioma angiogenesis through different mechanisms.

CircRNAs are a new family of ncRNAs found in all eukaryotic cells. CircRNA covalently forms a closed loop structure with neither poly (A) tail nor polarity of 5′–3′. Because the structure of circRNA occludes endogenous biomolecules, circRNA is resistant against exonuclease Rnase R digestion and miRNA-mediated non-classical degradation. The biogenesis of circRNAs is regulated by specific cis-acting elements and trans-acting factors, resulting in a tissue-specific and cell-specific expression pattern (Kristensen et al., 2019). Classically, circRNAs specifically bind to and adsorb miRNAs, thereby relieving the inhibitory effects of miRNAs on downstream target genes. For example, down-regulation of circ_002136 expression significantly inhibits the survival, migration, and tube formation of glioma endothelial cells. Circ_002136 functionally targets miR-138-5p to upregulate SOX13, which as a target gene of miR-138-5p directly associates with and activates SPON2 to promote angiogenesis (He Z. et al., 2019). Table 3 summarizes the circRNAs (cZNF292, cir-DICER1, circ-SHKBP1, and circSCAF11) involved in the regulation of angiogenesis in the glioma microenvironment (Yang et al., 2016; He et al., 2018; He Q. et al., 2019; Meng et al., 2019). Of note, so far the studies about circRNAs in glioma angiogenesis mainly focus on the classical regulation of miRNAs. However, circRNAs have the capacity to directly bind to proteins and affect their functions. In addition, circRNAs may be associated with pre-mRNA cleavage and ribosomal RNA maturation. Future studies may reveal the involvement of circRNAs in angiogenesis through these non-classical mechanisms.

Numerous studies have emphasized the pivotal functions of soluble factors in angiogenesis, including the common proangiogenic factors VEGF, angiopoietin, and FGF, and the common angiogenic inhibitors arresten, endostatin, angiostatin, and MMP. Soluble factors are critical mediators of the functions of ncRNAs in regulating angiogenesis. For instance, the lncRNA PVT1 is able to bind and downregulate miR-26b and promote the expression of the soluble factors CTGF and ANGPT2 to boost vascularization (Zheng et al., 2018). CTGF is a member of the CCN family and other CCN proteins such as Cyr61 and NOV have similar functions in glioma (Le Mercier et al., 2008; Sin et al., 2008; Goodwin et al., 2010). Future studies may discover more ncRNAs for regulation of the CCN family. The most common proangiogenic factor VEGFA has been reported to be regulated by multiple ncRNAs including LINC01116, lncRNA CCAT2, and lncRNA TUG1 (Cai et al., 2017; Sun et al., 2020; Ye et al., 2020). It is reasonable that ncRNAs may regulate much more soluble factors to deliberately control angiogenesis (Yu et al., 2017; Yang et al., 2018). Table 4 summarizes the ncRNAs and their target soluble factors that regulate angiogenesis in the glioma microenvironment.

During glioma invasion and progression, GJIC is an important intercellular communication process based on gap junctions and channels between two adjacent cells that connect the cytoplasm of cells (Katakowski et al., 2010; Hong et al., 2015). Briefly, six connexin (Cx) proteins assemble on one cell to form hemichannels (linkers) that pair with homotypic/heterotypic hemichannels of adjacent cells to mediate GJIC (Laird, 2006). miRNAs can cross gap junction channels to establish intercellular communication and directly affect gene expression in recipient cells (Lim et al., 2011). Cx43 is the most abundant connexin isoform in astrocytes, and down-regulation of Cx43 impairs tumor cell motility and invasiveness probably through disrupting the GJIC between astrocytes and tumor cells (Bates et al., 2007). GJIC also plays a role in functional miRNA transfer between GBM cells and human microvascular endothelial cells (HMEC; Thuringer et al., 2016). It has been demonstrated that miR-145-5p expressed in HMEC and miR-5096 derived from GBM cells can transfer between the two types of cells through GJIC. MiR-5096 may promote GBM invasiveness and angiogenesis after transferring into HMEC, while miR-145-5p acts as a tumor suppressor after moving into GBM cells (Thuringer et al., 2017). Direct cell-to-cell communication via GJIC is superior to indirect communication in terms of specificity and efficiency. in vivo delivery of therapeutic miRNAs through GJIC may be more effective but of a shorter duration than other delivery approaches (Lemcke et al., 2015).

Due to the lack of specific molecular markers, in 2018 the International Society for Extracellular Vesicles (ISEV) recommended stratifying extracellular vehicles (EV) into small vesicles <200 nm) and medium/large vesicles (>200 nm) according to their sizes (Thery et al., 2018). Several types of cells can secrete multiple types of EVs including exosomes (30–150 nm in diameter) derived from the inner vesicles, and microvesicles (100–1,000 nm in diameter) formed by outward budding of the cell membrane (Skog et al., 2008; Raposo and Stoorvogel, 2013). EVs play an important role in cell communication events such as direct cell-cell contact, plasma membrane fusion, and receptor-mediated endocytosis. Through transmitting the bioactive components of donor cells into recipient cells, EVs regulate the behaviors of recipient cells and affect the tumor microenvironment (Skog et al., 2008; Balaj et al., 2011; Montecalvo et al., 2012). The biomolecules as cargos of EVs include genomic DNA, cDNA, various RNAs, and proteins (cytoplasmic and membrane-bound), but ncRNAs are the most common and abundant contents of EVs that are transferred between cells with the strongest activity in regulating angiogenesis in the tumor microenvironment (Valadi et al., 2007; Van Der Vos et al., 2011; Rooj et al., 2016). A single GBM cell is able to secrete approximately 10,000 EVs in 48 h (Whitehead et al., 2020). Glioma cells express high levels of Lnc-HOTAIR that can be loaded into exosomes and transmitted to endothelial cells, which upregulates the pro-angiogenic factor VEGFA in vascular endothelial cells and ultimately enhances angiogenesis (Ma X. et al., 2017). Table 5 summarizes the ncRNAs enriched in EVs that regulate angiogenesis in the glioma microenvironment.

ncRNAs may come from multiple cellular origins in the heterogeneous glioma microenvironment as summarized in Table 5. Among glioma cells, GSCs or glioma initiating cells are a small proportion of stem-like tumor cells with self-renewal and multi-lineage differentiation capacity. GSCs are enriched in perivascular niches and function as key regulatory cells to promote angiogenesis (Calabrese et al., 2007; Cheng et al., 2020). GSC-derived exosomes have a central role in angiogenesis and contain plenty ncRNAs in addition to pro-angiogenic factors (VEGF, TGF-β1, and CXCR4, etc.; Treps et al., 2017; Quezada et al., 2018). Skog et al. (2008) firstly reported that miRNAs loaded in EVs can be taken up by cultured endothelial cells to promote angiogenesis in GBMs. Recently, Lucero et al. (2020) performed RNA-seq on human brain endothelial cells exposed to GSC-derived EVs along with in silico analysis of epigenetic profiles of GBMs in the TCGA database, and they found that GSC-derived EVs activated transcription of multiple genes in endothelial cells that are highly correlated to angiogenesis. Of note, whereas some reports specifically studied the GSC-derived ncRNAs such as miR-26 and miR-21 in promoting angiogenesis (Sun et al., 2017; Abels et al., 2019; Wang Z.-F. et al., 2019), most studies did not emphasize the hierarchy of glioma cells when it comes to the cellular origin of ncRNAs (Boccaccio and Comoglio, 2013; Gao et al., 2013; Lang et al., 2017; Lucero et al., 2020; Wang et al., 2020). It remains an open question whether the expressions of proangiogenic ncRNAs are under the control, at least in part, of the signaling pathways specifically activated in GSCs.

Glioma-associated mesenchymal stem cells (gbMSCs) are stromal cells associated with the malignant development of gliomas. gbMSCs consist of CD90^high and CD90^low subgroups with different roles in glioma progression. In particular, CD90^high gbMSCs function to increase proliferation and migration of glioma cells, while CD90^low gbMSCs have the capacity to transform into pericytes and alter miRNA expression profiles of vascular endothelial cells to promote angiogenesis (Yi et al., 2018). Intracranial xenografts derived from GSCs co-cultured with gbMSCs showed increased CD31 expression, suggesting that gbMSCs enhanced microvascular density (Kong et al., 2013). Because of the functional importance of MSCs in glioma angiogenesis, they may be both origins and targets of ncRNAs for the regulation of angiogenesis. So far there are very limited studies concerning ncRNAs in MSCs (Liu L. et al., 2020). Future studies about the ncRNAs in regulation of MSC phenotypes or the impact of MSC-derived ncRNAs on other components in tumor vasculature will certainly inspire the development of new anti-angiogenic strategies for glioma treatment.

The glioma tissues contain multiple types of immune cells, and tumor-associated macrophage (TAM) and microglia as the major glioma-infiltrating immune cells account for more than 30% of the GBM bulk cell population (Hambardzumyan et al., 2016). TAM is thought to be closer to macrophages of the M2 type and contributes to the resistance against anti-VEGF therapy through promoting angiogenesis and upregulating Treg cells (Zhu et al., 2017; Powell et al., 2018; Long et al., 2020). ncRNAs delivered by EVs and received by immune cells may be involved in angiogenesis. For example, miR-21 as a cargo of EV is delivered to and ingested by microglia, resulting in regulation of specific downstream mRNA targets and thus promoting angiogenesis (Abels et al., 2019). Future studies may discover more functions of glioma-infiltrating immune cells as the source cells and recipient cells of ncRNAs to regulate angiogenesis in the tumor microenvironment.

RNA therapy refers to RNA-based targeted therapy either aiming at a specific RNA sequence or utilizing RNA-based molecules as drugs. RNA-based drugs can be grossly divided into siRNA/shRNA, miRNA, antisense oligonucleotide (ASO), and aptamer in terms of molecular structures. Relative to traditional small molecular chemical drugs and large molecular biopharmaceuticals, RNA therapy is more quantitative, dynamic, and flexible.

The majority of siRNA, miRNAs, and ASOs function by promoting the degradation or blocking the translation of target mRNAs. SiRNA, or small interfering RNA, is a class of double-stranded RNA with a length of 20–25 base pairs. By intraperitoneal injection of plasmids encoding siRNAs, Gondi et al. (2004, 2007) showed that siRNAs against uPA, the uPA receptor, and MMP-9 led to a significant reduction in glioma angiogenesis and tumor growth in preclinical mouse models. Likewise, Niola et al. (2006) showed that VEGF-targeted siRNA decreased GBM angiogenesis in a xenograft mouse model. Alternatively, Kargiotis et al. (2008) used adenovirus to deliver siRNA against MMP-2, resulting in decreased invasiveness and inhibition of angiogenesis. In addition to siRNAs, miRNAs and ASOs are being developed for cancer treatment. Both miRNA antagonists and miRNA mimics can work as miRNA-based drugs. As to glioma treatment, the miR-296 antagonist reduces angiogenesis in tumor xenografts in vivo (Wuerdinger et al., 2008). ASOs are small-sized single-stranded nucleic acids that form RNA-DNA hybrids with complementary RNA targets to promote RNase-H-mediated degradation of target genes. ASO had been used to target lncRNA TUG1 and inhibit the self-renewal of GSC (Katsushima et al., 2016).

Distinct from other types of RNA molecular drugs, aptamers are oligonucleotide or peptide molecules that bind to specific target molecules, also known as “chemical antibodies”. They are functionally comparable to traditional antibodies but have several advantages including small physical size, flexible structure, multifunctional chemical modification, high stability, and lack of immunogenicity. In addition, aptamers can be internalized after binding to certain receptors and self-cleaved in the presence of ribozyme and target molecules, making them useful as targeted delivery agents. Pegaptanib, approved by the FDA in 2004, is an RNA aptamer that selectively binds VEGF165 to prevent the interaction between VEGF165 and its receptors (VEGF-R1, VEGF-R2, Npn-1) and block VEGF165-mediated signaling, thereby inhibiting choroidal neovascularization and vascular leakage (Gragoudas et al., 2004; Stein and Castanotto, 2017).

Besides the aforementioned RNA-based drugs, targeting the physiological interaction between lncRNAs and miRNAs may provide new approaches for anti-angiogenic therapies (Zhao J. et al., 2019; Teppan et al., 2020). A large number of preclinical and clinical studies have demonstrated a bright future of RNA therapy with a broad spectrum of potential targets in tumor angiogenesis.

EVs, especially exosomes, can work as privileged tools for the delivery of RNA cargos in the tumor microenvironment for angiogenic therapy of gliomas (Spinelli et al., 2019; Yekula et al., 2020). Compared with liposomal, metal, and polymeric nanomaterials, exosomes have better bioavailability but less cytotoxicity and immunogenicity. The membrane structure of EVs provides a stable and appropriate environment for the cargo molecules, and the signaling molecules on the surface of EV membranes may guide precise delivery to specific target cells. The ability of EVs to cross the blood-brain barrier is especially important for the treatment of brain cancers. Because of these advantages, EVs have strong potentials to be transformed into effective ncRNA delivery systems. A recent study showed that implantation of the miR-302–367-expressing donor cells to produce EVs containing miR-302–367 in the tumor microenvironment efficiently suppressed GBM development in mouse brain (Fareh et al., 2017; Nair et al., 2018). In addition, the EV donor cells may be genetically engineered to produce EVs with modified surfaces that could lead to more effective delivery to target cells. For example, the donor cells were engineered to express EGFR binding peptides on EV surface, and the modified EVs efficiently delivered let-7a specifically to EGFR-expressing tumor tissues (Ohno et al., 2013). Therefore, EVs may work as effective delivery tools for RNA therapy to express tumor-suppressive RNAs in the tumor microenvironment or specific tumor cells.

Despite the abovementioned advantages, the utilization of EVs in RNA therapy has been obstructed by several problems. There is no universally accepted standard for the isolation, purification, and preservation of exosomes. The mechanisms regulating the sorting and uptake of EVs in the tumor microenvironment remain unclear. Studies from different labs are somewhat incomparable because of the heterogeneity of EVs and the lack of a standardized quantification method. The application of EVs in RNA therapy requires a consensus on the methodology.

More than one hundred RNA modifications have been reported so far. A well-known case is the N 6-methyladenosine (m6A) modification that refers to the reversible addition of a methyl group to the N element at position 6 of the A base in RNAs, which is the most abundant and well-characterized internal modification in mRNAs. Such RNA modifications occur also in regulatory ncRNAs, which may affect the protein-binding of lncRNAs, maturation of miRNAs, and translation of circRNAs to regulate their biogenesis and functions. The dynamic modifications may endow the ncRNAs with the flexibility to adapt to the constantly changing microenvironment and promote tumor progression.

Many research groups have observed the m6A modification in lncRNAs during the study of the polyA-enriched RNAs (Dominissini et al., 2012; Meyer et al., 2012). The m6A modification may affect the interaction between lncRNAs and their partner proteins. The lncRNA MALAT1 has m6A hypermethylation modification on multiple sites. Two of these m6A residues prevent the formation of the secondary structure of MALAT1 through the “m6A switch” mechanism and enhance the recognition and binding of hnRNPC to U5 channels in the MALAT1 hairpin (Dominissini et al., 2012; Liu et al., 2013, 2015). Interestingly, lncRNAs may also function as an m6A regulator. FOXM1-AS, a lncRNA antisense to FOXM1, promotes the interaction between the m6A demethylase ALKBH5 and the FOXM1 nascent transcripts, leading to the removal of m6A from FOXM1 transcripts to enhance FOXM1 expression in glioblastoma (Zhang et al., 2017).

The m6A modification participates in miRNA maturation. miRNA primary transcripts (pri-miRNA) transcribed from DNA undergo a series of cleavages to get into hairpin miRNA precursor (pre-miRNA) and finally mature miRNAs. However, the canonical m6A motif GGAC is abundant on pri-miRNAs but rarely seen on pre-miRNAs and mature miRNAs. The m6A writer METTL3 binds to pri-miRNAs to execute m6A modification. The m6A-bearing pri-RNAs are then recognized by hnRNPA2B1, which in turn interacts with the microRNA Microprocessor complex protein DGCR8 to promote pri-RNA processing. Therefore, alteration of METTL3 expression may impact the overall m6A levels and the expression of mature miRNAs (Alarcon et al., 2015a,b).

The m6A modification is also prevalent in circRNAs and its read-write mechanisms are similar to those in mRNAs. However, the functions of m6A modification in circRNAs may be different from those in mRNAs. The m6A reader YTHDF2 sequesters methylated circRNA to prevent the activation of the RNA pattern recognition receptor RIG-I, which is essential for the suppression of innate immunity (Zhou et al., 2017; Chen et al., 2019). Meanwhile, the m6A modifications in circRNAs seem to promote the translation initiation from circRNAs through YTHDF3 and the initiation factor eIF4G2 (Yang Y. et al., 2017).

Previous studies have shown that the m6A RNA methylase METTL3 promotes glioma progression by stabilizing the expression of SOX2, suggesting an important role of m6A modifications in glioma (Somasundaram, 2018; Visvanathan et al., 2018). To date, most studies have been focusing on the identification of m6A-modified mRNAs, while little is known about the m6A modification on ncRNAs. Future investigations about the impact of m6A modifications on the production, cellular location, target selection, and other features of ncRNAs will certainly inspire more explorations about the regulation of angiogenesis by ncRNAs in the glioma microenvironment.

RNA-binding proteins (RBP) are a class of proteins that bind to RNAs, including mRNAs and ncRNAs, to regulate their biogenesis and expression levels (Xia et al., 2012; Xu Y. et al., 2019). The READDB¹ database contains the information of 1,344 RBPs and the related diseases (Hashemikhabir et al., 2015). In addition, the catRAPID² and the starBase v2.0³ databases provide useful tools to predict the interactions between RBPs and ncRNAs (Agostini et al., 2013; Li et al., 2014). The binding of RBP to ncRNAs participates in the regulation of target mRNAs from multiple aspects, including mRNA processing, maturation, transport, localization, and translation (Gerstberger et al., 2014; Janakiraman et al., 2018). In the meantime, the interaction between RBP and ncRNAs affects the expression levels of RBP and/or ncRNAs. Consequently, RBP-ncRNA interaction is involved in the development of neurodegenerative diseases, metabolic diseases, and various cancers (Kim et al., 2017).

RBP regulates not only the generation of miRNAs but also their functions. RBP may have antagonistic or facilitatory effects on miRNAs to synergistically regulate the translation of target genes (Van Kouwenhove et al., 2011; Ho and Marsden, 2014). RBP also affects the stability of lncRNAs. For example, the RBP IGF2BP1 binds to and destabilizes the lncRNA HULC in human hepatocellular carcinoma (Haemmerle et al., 2013). During the biosynthesis of circRNA, RBPs bind to introns near splicing sites and contribute to the production of circRNAs through the RBP-driven cyclization mechanism (Conn et al., 2015; Lyu and Huang, 2017). In addition, RBPs interact with circRNAs to regulate circRNA splicing, processing, folding, stabilization, and localization (Dudekulay et al., 2016). The roles of RBPs as upstream regulators of ncRNAs in glioma microenvironment and angiogenesis are largely unknown.

So far there is no available literature about ncRNAs other than miRNAs, lncRNAs, and circRNAs that have a role in glioma angiogenesis. However, accumulating reports emphasize the biological importance of the diverse types of ncRNAs, which may be star molecules for future studies on the glioma microenvironment.