GlycoRNA complexed with heparan sulfate regulates VEGF-A signalling

Abstract

Heparan sulfate proteoglycans (HSPGs) have been recognized as key plasma membrane-tethered co-receptors for a broad range of growth factors and cytokines containing cationic heparan-binding domains^1,2. However, how HSPGs mechanistically mediate signalling at the cell surface—particularly in the context of cell surface RNA—remain poorly understood. During developmental and disease processes, vascular endothelial growth factor (VEGF-A), a heparan sulfate-binding factor, regulates endothelial cell growth and angiogenesis³. The regulatory paradigm for endothelial cell-mediated selectively of VEGF-A binding and activity has largely been focused on understanding the selective sulfation of the anionic heparan sulfate chains^4,5,6,7,8. Here we examine the organizational rules of a new class of anionic cell surface conjugates, glycoRNAs^9,10, and cell surface RNA-binding proteins (csRBPs^11,12). Leveraging genome-scale knockout screens, we discovered that heparan sulfate biosynthesis and specifically the 6-O-sulfated forms of heparan sulfate chains are critical for the assembly of clusters of glycoRNAs and csRBPs (cell surface ribonucleoproteins (csRNPs)). Mechanistically, we show that these clusters antagonize heparan sulfate-mediated activation of ERK signalling downstream of VEGF-A. We demonstrate that the heparan sulfate-binding domain of VEGF-A₁₆₅ is responsible for binding RNA, and that disrupting this interaction enhances ERK signalling and impairs vascular development both in vitro and in vivo and is conserved across species. Our study thus uncovers a previously unrecognized regulatory axis by which csRNPs negatively modulate heparan sulfate-mediated signalling in the context of angiogenesis driven by VEGF-A.

Main

The extracellular matrix is a key interface that actively regulates cell–cell interactions, host–pathogen interactions, and autocrine and paracrine signalling through growth factor binding (reviewed in refs. ^13,14). HSPGs are a central component of the extracellular matrix and consist of a core protein and one or more covalently linked heparan sulfate (HS) chains¹. The protein interactome of HSPGs is well established and includes chemokines, cytokines, growth factors, morphogens, cell-adhesion proteins and viral components such as the SARS-CoV-2 spike protein^{5,15,16,17,18}. HSPGs are conceptually considered as coreceptors that facilitate the formation of ligand–receptor complexes. Protein–HS interactions occur via sulfated domains of the chains, and high-level and/or site-specific sulfation of HS control the various activities ascribed to HSPGs¹⁶, which span development, physiology and disease.

In addition to HS, RNAs represent another class of negatively charged biopolymers on the cell surface (cell surface RNAs (csRNAs))^9,19,20,21. Among them are glycoRNAs⁹, which contain sialylated and fucosylated N-glycans on small RNAs and can be detected by the sialic acid-binding immunoglobulin-type lectins (Siglec) family of receptors⁹. A covalent linkage between small non-coding RNA and N-glycans has been identified as the modified RNA base 3-(3-amino-3-carboxypropyl)uridine (acp³U)¹⁰. Functionally, N-glycosylation of RNA at acp³U has been shown to chemically shield this RNA modification from sending by endosomal TLRs²². GlycoRNAs have also been found to interact with lectins (for example, P-selectin) and control neutrophil recruitment in vivo²³. In addition to glycoRNAs, the surface of cells also contains a selective suite of RBPs that are in physical proximity to glycoRNAs, form precise nanosized clusters and are sites of cell-penetrating peptide (for example, TAT) entry, regulated by RNA¹¹. These csRBPs have been demonstrated on many different cell types and certain csRBPs can be selectively presented on cancer cells, in the cases of csNPM1 (ref. ¹²) and csU5 snRNP200 (ref. ²⁴). Together, these reports highlight that csRNAs are not merely discarded passenger molecules but important biological players in cell surface biology. However, the specific roles of these csRNP clusters in cell signalling and development remain poorly understood.

Although insights into csRNPs are emerging, extracellular RNA (exRNA) and extracellular RNases have long been characterized, there is long-standing evidence that exRNA biology may contribute to angiogenesis. It is thought to modulate VEGF signalling; however, the precise mechanisms by which exRNA and its associated RNPs exert this regulatory effect remain unclear. For example, exRNA has been proposed to modulate VEGF activity by directly interacting with both the C terminus²⁵ and the N terminus²⁶ of the VEGF-A₁₆₅ isoform. This RNA-mediated process appears to be sensitive to RNase treatment, as it can abrogate the observed VEGF stimulation^25,27. Separately, angiogenin (ANG), an extracellular RNase, is known to promote angiogenesis (reviewed in ref. ²⁸), yet whether it targets exRNAs is poorly defined. Of note, pharmacological inhibition of ANG orthologues in zebrafish broadly impairs vascular development, suggesting that exRNA degradation itself may have an essential regulatory role in angiogenesis during development²⁹. Thus, despite decades of research generally on angiogenesis^30,31 and RNase-mediated or exRNA-mediated angiogenic pathways^27,28, it remains unresolved which specific RNA species are involved, through what receptors or molecular partners they act, and the exact mechanisms on the cell surface by which they modulate angiogenesis.

Here we uncovered a new layer of regulation of csRNPs as potential antagonist of HS-mediated signalling. We found that csRNP assembly on living cells depends on intact HS chains modified with N-sulfate and 6-O-sulfate. Functionally, VEGF-A₁₆₅ can directly bind to RNA on the surface of endothelial cells, which negatively regulates its intracellular signalling cascade. The heparan-binding C terminus of VEGF-A₁₆₅ can bind to glycoRNAs. Mutating key arginine residues to lysine in this domain decouples HS binding from glycoRNA binding and enabled on cell and in vivo experiments, revealing hyper-vascularization of the mouse retina and the developing vasculature in zebrafish embryos in this mutant. Together, our work nominates csRNPs as an antagonist of VEGF-A₁₆₅-mediated signal transduction in endothelial cells and during angiogenesis.

Siglec-11 binds to csRNA

We previously found that Siglec-11, a member of the Siglec receptor family³², which are cell surface receptors, binds to the surface of HeLa cells in an RNA-dependent manner⁹, suggesting that Siglec-11 is potentially a csRNA receptor. To explore whether other members of the Siglec family are also RNA-dependent cell surface binders, we used confocal imaging and screened the 13 commercially available Siglec–Fc fusion proteins for binding to suspension (MOLM-13) and adherent (U2OS) cell lines. Only Siglec-4, Siglec-7, Siglec-9 and Siglec-11 specifically bound both cell types and only Siglec-11 was reduced after live-cell RNase treatment¹¹ (RNase A and RNase III; Extended Data Fig. 1a–c). By contrast, although Siglec-7 and bulk cell surface sialic acids (periodate labelling³³) were lost after live-cell sialidase treatment, Siglec-11 binding was not impacted (MOLM-13) or only mildly impacted (U2OS; Extended Data Fig. 1d). To test whether Siglec-11 was indeed binding near RNA, we used an anti-double-stranded RNA (dsRNA) antibody (9D5) that detects csRNA¹¹. Co-staining of U2OS and MOLM-13 cells with 9D5 and Siglec-11 showed that 42% and 61% of 9D5 puncta overlapped with Siglec-11 on MOLM-13 and U2OS cells, respectively (Fig. 1a and Extended Data Fig. 2a,b), 34.5% and 24% of Siglec-11 puncta overlapped with 9D5 on MOLM-13 and U2OS cells, respectively. Co-staining 9D5 with Siglec-7 and separately with Siglec-9, both of which are not sensitive to RNases, showed that only 2.5% and 1.4% of 9D5 puncta overlapped with Siglec-7 or Siglec-9 in MOLM-13, and 0% and 2.8% of 9D5 puncta overlapped with Siglec-7 or Siglec-9 in U2OS, respectively. Finally, although the cell surface binding of Siglec-11 is sensitive to RNase, we next assessed whether Siglec-11 can directly interact with sialic acid-modified glycoRNAs. Using highly purified RNA (Methods), we evaluated enriched material from immunoprecipitations with Siglec-11 or an Fc control and chemically labelled for glycoRNAs (rPAL¹⁰). We found that Siglec-11 captured sialoglycoRNAs above the Fc and that this signal was sensitive to RNase treatment (Extended Data Fig. 2c). The well-correlated binding of 9D5 and Siglec-11 suggests that Siglec-11 ligands are near glycoRNA on the cell surface.

Fig. 1: csRNA and RBP clusters are dependent on HS biogenesis.

a, Representative images of U2OS cells co-stained with Siglec-11 (yellow) and 9D5 (magenta). An enlargement of the hatched box is shown, and a bright field (BF) is shown to represent the outline of the cell membrane. The nearest neighbour distance analysis of the Siglec pairs is also imaged (bottom). For each pair, the distance (µm) from that anchor (left side protein name in the figure key) to the other pair was calculated across the indicated number of cells. These values were plotted in a density histogram. b, Dot plot of genes identified in the genome-wide CRISPR-KO screen for the loss of Siglec-11 cell surface binding ranked by CRISPR score. The top 15 gene names are displayed, with cut-off shown. The inset illustrates how Siglec-11 could interact with a glycoRNA. c, Dot plot of genes identified in the genome-wide KO screen as in panel b, for the loss of 9D5 cell surface binding. The inset illustrates how 9D5 could interact with a glycoRNA. d, Upset plot analysis of genes with a score cut-off of −0.8 from the Siglec-11, 9D5 and MAA-I genome-wide KO screens. The common overlapping hits between 9D5 and Siglec-11 are highlighted in blue. The total number of hits for each intersection is noted. Statistical assessment was performed with a hypergeometric test and P values are shown without adjustment. e, Representative images of WT and KO U2OS cells stained live with 10E4, Siglec-11–Fc, 9D5 and Siglec-7 (all in red). Ab, antibody; dsRNA, double-stranded RNA; mE, mEmerald. f, Quantification of the indicated signal dot numbers and intensity per cell from panels e,g from three independent experiments. Mut, mutant. g, Representative images of the indicated cell lines stained live with anti-DDX21 and anti-hnRNP-U (both in red).

HSs organize csRNPs

To identify what mediates Siglec-11 and 9D5 binding, a genome-wide CRISPR–Cas9 gene-knockout (KO) approach³⁴ was developed based on flow cytometry of MOLM-13 cells using Siglec-11, 9D5 and the plant lectin MAA-I, which is well characterized to selectively bind to sialic acid³⁵ (Fig. 1b,c and Extended Data Fig. 3a). We identified 154, 187 and 246 hits in the Siglec-11, 9D5 and MAA-I screens, respectively (Fig. 1d and Supplementary Table 1). The top hits enriched in the MAA-I screen were related to sialic acid biosynthesis (for example, CMAS (#1) and NANS (#11)³⁶; Supplementary Table 1), confirming the specificity of the screen; however, both of these genes scored poorly in the Siglec-11 and 9D5 screen. Intersectional and pairwise analysis of these enriched genes showed a more robust overlap between 9D5 and Siglec-11 than MAA-I with either of the other two probes (Fig. 1d), consistent with Siglec-11 association to the cell surface in an RNA-dependent manner. Direct inspection of the 41 genes commonly scoring between 9D5 and Siglec-11 showed consistent robust scores for each hit, whereas these genes often scored poorly in the MAA-I screen (Extended Data Fig. 3b), highlighting the overlap and common effect of these genes on both 9D5 and Siglec-11 cell surface binding. Gene ontology analysis revealed terms enriched in membrane compartments and processes including heparin biosynthetic process (Extended Data Fig. 3c). Inspection of the top hits of Siglec-11 revealed EXT1, EXT2 and UXS1 as the top three genes, which are key enzymes of HS biogenesis (Fig. 1b). The 9D5 screen also revealed EXT1 as a highly scoring hit, whereas UXS1 and EXT2 were present but below the cut-off.

To verify that HS chains are required for the tethering of Siglec-11 and RNA to the cell surface, we generated KO clones in U2OS cells. Within the Golgi, HS chains are initiated by xylosylation of proteoglycan core proteins, which depends on UDP-xylose formation catalysed by UXS1 (refs. ^1,37). This is subsequently elongated by a hetero-dimeric complex formed by EXT1 and EXT2 by adding alternating residues of D-glucuronic acid (GlcA) and GlcNAc to the nascent polymer^38,39,40, providing the eventual substrate for sulfation and HS-interacting proteins to bind. We stained KO cells (Extended Data Fig. 4a,b) with 10E4, a specific antibody recognizing mature HS^41,42, and verified that HS was not produced (Fig. 1e,f). Moreover, EXT1 KO, EXT2 KO and UXS1 KO cells all show complete loss of both Siglec-11 and 9D5 binding. The binding of Siglec-7 and Siglec-9 was not affected on EXT2 KO cells (Fig. 1e,f and Extended Data Fig. 4c,d). The loss of 9D5 signal was selective to the cell surface as intracellular staining revealed no change inside of cells upon EXT2 KO (Extended Data Fig. 4e). To understand whether this effect was due to the enzymatic activity of EXT2, EXT2 KO cells were transfected with wild-type (WT) EXT2 or a catalytically inactive (D517N/D573N⁴⁰) mutant cDNA (Extended Data Fig. 4f). Reintroduction of WT EXT2 rescued the loss of 9D5 and Siglec-11, whereas the catalytically inactive mutant did not (Fig. 1e–g). These effects were not restricted to EXT2 KO, as EXT1 KO and UXS1 KO also resulted in the loss of HS, 9D5 and Siglec-11 binding (Extended Data Fig. 4g–i). These results show the dependency of clustered csRNA on HS biosynthesis.

Given the colocalization of glycoRNA and csRBPs¹¹, we next tested whether csRBPs are also organized by HS. On WT U2OS cells, both cs-DDX21 and cs-hnRNP-U colocalize with 9D5 (Extended Data Fig. 4j). Consistently, no observable DDX21 and hnRNP-U antibody binding occurred on the cell surface in EXT2 KO cells (Fig. 1f,g). This loss was not due to bulk changes in DDX21 or hnRNP-U as whole-cell lysates had no changes between the WT and KO (Extended Data Fig. 4k). The formation of csRBP clusters is also dependent on EXT2 enzymatic activities as reintroduction of the WT, but not the catalytically inactive (D517N/D573N)⁴⁰, rescued the formation of csRBP clusters in EXT2 KO cells (Fig. 1f,g). These findings indicate that HS is a critical scaffold in modulating the recruitment and retention of RNA and their respective binding proteins (for example, csRBP), csRNPs, to the cell surface.

csRNPs depend on intact HS polymers

To determine whether intact HS chains on the cell surface facilitate csRNP clustering, we performed cell staining on U2OS cells after heparin lyase treatment (30 min, heparin lyases I/II/III). Treated cells lost the binding of 10E4, Siglec-11, 9D5, anti-DDX21 and anti-hnRNP-U, whereas Siglec-7 and Siglec-9 binding was not impacted (Extended Data Fig. 5a,b). Given the colocalization of 9D5 and Siglec-11 and sensitivity to heparin lyases, we sought to understand the spatial and temporal relationship between 9D5, Siglec-11 and HS. We therefore performed a three-colour co-staining experiment on U2OS cells and found that Siglec-11 puncta are highly correlated in localization with 10E4 puncta (Fig. 2a and Extended Data Fig. 5c,d, green line). This relationship was also true when examining 10E4, Siglec-11 and anti-DDX21 (Extended Data Fig. 5e), supporting the clustering model of csRNPs near to sites of HSPGs. We next assessed the temporal regulation on csRNP clustering by conducting a recovery experiment in cells after heparin lyase treatment, removal of the enzymes and incubation of the cells with fresh media for 0, 45, 90 or 180 min to access de novo HS production and cell surface presentation (Fig. 2a and Extended Data Fig. 5c,d). HS puncta reappeared 45 min after a recovery from heparinase treatment (9.4%). At this early time point, Siglec-11 and 9D5 clustering was not detected. After 90 min, however, 9D5 and Siglec-11 started to recover together at sites of large HS clusters, suggesting that a critical concentration of HS is needed for proper csRNP clustering. After 180 min, the levels of HS, Siglec-11 and 9D5 binding recovered to native levels, confirming that intact HS chains are vital for csRNP cluster formation on the live cell surface.

Fig. 2: csRNPs are dependent on intact HS chains.

a, Representative images of U2OS cells treated with a heparinase pool for 30 min, recovering for the indicated times and co-stained with 10E4 (cyan), Siglec-11 (yellow) and 9D5 (magenta). An enlargement of the hatched box is shown, and a bright field is shown to represent the outline of the cell membrane. Three independent experiments were performed. b, Schematic of a HSPG with the regions of activity for the various enzymes and HS oligos perturbed in the following experiments. 2S, 2-O-sulfation; 6S, 6-O-sulfation; NS, N-sulfation. c, Representative images of WT, NDST1-KO-treated, HS6ST1-KO-treated, HS2ST1-KO-treated, Sulf1 stably overexpressing (OE)-treated, Sulf2 stably OE-treated, Tega HS #09-treated and Tega HS #37-treated U2OS cells stained with 10E4, Siglec-11, 9D5, anti-DDX21 and anti-hnRNP-U (all in red) separately. d, Quantification of data in panel c with the number of cells counted from three independent experiments. Data are mean ± s.e.m.

6-O-sulfated HS facilitates csRNPs

The complex but precise sulfation of HS chains¹⁶ is initiated by one or more members of the NDST family of N-deacetylases-N-sulfotransferases (NDSTs) acting on a subset of GlcNAc residues (Fig. 2b). NDST1 KO in U2OS cells (Extended Data Fig. 6a–c) resulted in a 74% loss of N-sulfoglucosamine residues as measured by reduction of 10E4 staining (Fig. 2c,d). NDST1 deficiency caused a 77%, 72%, 62% and 63% loss of binding intensity per cell in Siglec-11, 9D5, cs-DDX21 and cs-hnRNP-U, respectively (Fig. 2c,d), suggesting that sulfation of HS chains is required for tethering RNA and associated proteins to the cell surface. To broadly and non-genetically inhibit sulfation, we treated cells with sodium chlorate, a metabolic inhibitor of sulfation⁴³, and found near complete loss of 10E4, Siglec-11, 9D5, cs-DDX21 and cs-hnRNP-U, whereas sodium chloride (control) had no effect (Extended Data Fig. 6d). Correspondingly, 3G10 antibody specific for cut HS⁴¹, only bound cells after heparin lyase treatment (Extended Data Fig. 6e).

Following N-sulfation, partial epimerization of adjacent GlcA residues to L-iduronic acid (IdoA) can lead to 2-O-sulfation (catalysed by Hs2st), 6-O-sulfation of GlcNAc and GlcNS residues (by HS6ST1–3)^37,44. We therefore generated HS2ST1-KO and HS6ST1-KO U2OS cells (Extended Data Fig. 6a–c). Loss of 2-O-sulfation via HS2ST1 KO did not significantly alter the binding of the antibody panel, whereas loss of 6-O-sulfation through HS6ST1 KO reduced Siglec-11, 9D5, cs-DDX21 and cs-hnRNP-U binding by 76%, 67%, 56% and 53%, respectively (Fig. 2c,d). To complement this result, we individually overexpressed Sulf1 or Sulf2, two extracellular sulfatases that can remove 6-O-sulfation from intact HS^45,46 (Fig. 2b and Extended Data Fig. 6f,g). Indeed, overexpression of these sulfatases removed total Siglec-11, 9D5 binding, cs-DDX21 and cs-hnRNP-U on U2OS cells (Fig. 2c,d). Finally, we performed a competition experiment by adding exogenous HS chains with high N-sulfation, 6-O-sulfation and 2-O-sulfation (rHS09) to the media. rHS09 caused a loss of clustering of Siglec-11, 9D5, cs-DDX21 and cs-hnRNP-U, whereas the addition of HS chains with only high N-sulfation and 6-O-sulfation (rHS37) increased the average level of binding twofold to threefold (Fig. 2b–d). Although both rHS37 and rHS09 are highly sulfated, rHS37 merely lacks the 2-O-sulfate groups. Combined with the genetic evidence, this indicates that N-sulfation and 6-O-sulfation of HS promotes csRNP formation.

csRNAs regulate VEGF-A₁₆₅ signalling

Functionally, HSPGs bind to extracellular growth factors, facilitating signal transduction by forming a growth factor–growth factor receptor complex on the cell surface¹⁶. VEGF-A is a prototypical pro-angiogenesis factor causing endothelial cell migration and proliferation, with certain forms dependent on 6-O-sulfated HS for cell binding^7,47. Upon activation by VEGF-A, the VEGF receptor (VEGFR), a receptor tyrosine kinases expressed on endothelial cells^48,49, signals through activation of RAS–RAF1–MEK, which in turn phosphorylates ERK1/2¹⁵. Given this, we tested the presence and enzyme sensitivity of csRNPs on human umbilical vein endothelial cells (HUVECs). HUVECs generally stain positively for the same ligands tested on cancer cells and heparin lyase treatment resulted in a complete loss of binding across the antibody panel (Fig. 3a and Extended Data Fig. 7a). Treatment of HUVECs with RNases reduced binding of Siglec-11, 9D5, cs-DDX21 and cs-hnRNP-U puncta by 73%, 83%, 96% and 93%, respectively, without affecting 10E4 staining. These features of csRNP assembly, including heparin lyase and RNase sensitive, were also consistent on primary cultures of keratinocytes (Extended Data Fig. 7b,c), confirming that control of csRNP clustering by HSPGs is conserved on primary cells such as endothelial cells and keratinocytes.

Fig. 3: Intact csRNPs repress activity of VEGF-A₁₆₅ on endothelial cells.

a, Quantification of 10E4, Siglec-11, 9D5, cs-DDX21 and cs-hnRNP-U intensity per cell from three independent staining experiments on HUVECs. Data are mean ± s.e.m. a.u., arbitrary units. b, Western blot analysis of whole-cell lysates isolated from HUVECs after starvation and treatment with an RNase pool followed by 3 ng ml⁻¹ VEGF-A₁₆₅, VEGF-A₁₂₁ or EGF stimulation (top). Quantification of the ratio of pERK to total ERK was also calculated across the biological triplicates (bottom). Statistical assessment was performed with a two-sided Student’s t-test and P values are shown. Data are mean ± s.e.m. c, Representative images of starved HUVECs treated with an RNase pool, and subsequently with VEGF-A₁₆₅ or VEGF-A₁₂₁, finally stained with anti-VEGF-A₁₆₅ (red) or anti-VEGF-A (red). Statistical assessment was performed with a Student’s t-test and P values are shown. Data are mean ± s.e.m. NS, not significant. d, Representative images of HUVECs treated with an RNase pool and stained with anti-VEGFR2 (red). Statistical assessment was performed with a Student’s t-test and P values are shown. Data are mean ± s.e.m. e, Representative images of starved HUVECs treated with VEGF-A₁₆₅ and then co-stained with anti-VEGF-A₁₆₅ (purple) and Siglec-11 (yellow). Three independent experiments were performed. f, Schematic of the microfluidic chip (top left) used to grow HUVECs without (top middle) and with (top right) RNase A for 6 days. Representative images of BFP expressed in the HUVECs. Statistical assessment of the total migration area was performed using an unpaired two-sided Student’s t-test (bottom). Data are mean ± s.e.m. Four independent experiments were performed. g, Representative image (maximum z-projection view (left) and z-projection slice view (right)) of sprouts from a +RNase A device for the cells (blue), F-actin (red) and PECAM1 (green). Four independent experiments were performed.

Various proteoforms of VEGF-A, including VEGF-A₁₂₁ and VEGF-A₁₆₅, bind to the VEGFR via their N-terminal domains; however, VEGF-A₁₆₅ contains an extended C terminus that enables interactions with HS and neuropilin-1 (a VEGFR coreceptor)^50,51. We hypothesized that this domain could also interact with glycoRNAs. To test this, we serum starved HUVECs and used western blotting to detect total and phosphorylated ERK (pERK) after VEGF-A stimulation. Upon stimulation with 3 ng ml⁻¹ of VEGF-A₁₆₅ or VEGF-A₁₂₁ for 5 min, we observed an increase in pERK levels in serum-starved HUVECs (Fig. 3b, brown bars). Pre-treatment of HUVECs with RNases before the addition of VEGF-A₁₆₅ resulted in a 3-fold and 2.1-fold increase in pERK levels compared with cells without RNase treatment (Fig. 3b, green bars), and VEGF-A₁₂₁ activity was RNase insensitive. The RNase-dependent effect of VEGF-A₁₆₅ (and the insensitivity of VEGF-A₁₂₁) is also seen at 25 ng ml⁻¹ (Extended Data Fig. 8a–c). The RNase-dependent enhancement occurred through VEGFR2 signalling (at tyrosine 1175), as we observed detectable pVEGFR2 that was enhanced by 2.69-fold upon RNase treatment of the HUVECs (Extended Data Fig. 8d). Consistently, in vitro binding analysis with microscale thermophoresis showed that only VEGF-A₁₆₅, but not VEGF-A₁₂₁, was capable of interacting with small RNA (Extended Data Fig. 8e). These findings suggest that csRNA antagonizes VEGF-A₁₆₅ activation of ERK signalling at least in part through VEGFR2. This effect was specific to VEGF-A₁₆₅ as epidermal growth factor (EGF), which also signals through the ERK pathway but does not have a HS-binding domain⁵², was able to fully induce pERK in the presence of RNase (Fig. 3b). These data suggest that growth factors bearing a HS-binding domain are probably subject to RNase modulation.

We next assessed whether VEGF-A₁₆₅ is recruited to csRNP clusters and whether they impact cell surface association of VEGF-A₁₆₅. Serum starvation and pre-treatment with the RNases led to approximately two times more binding of VEGF-A₁₆₅ to the cell surface of HUVECs, whereas VEGF-A₁₂₁ was unchanged (Fig. 3c). The observed increase in VEGF-A₁₆₅ cell binding was not due to changes in its receptor VEGFR2 (ref. ³), as we found no changes in the receptor abundance on the cell surface after RNase treatment (Fig. 3d). Co-staining of VEGF-A₁₆₅ and Siglec-11 on HUVECs revealed 78% of Siglec-11 puncta overlapped with VEGF-A₁₆₅ and 40% of VEGF-A₁₆₅ puncta overlapped with Siglec-11 (Fig. 3e and Extended Data Fig. 8f,g). These data support the binding of VEGF-A₁₆₅ at assembled csRNPs and that RNase treatment selectively increases VEGF-A₁₆₅ cell binding.

RNase-dependent changes in pERK occur quickly (within 5 min) after VEGF-A₁₆₅ addition, suggesting that downstream gene expression and cell activation could be impacted in the HUVECs. To test this, we leveraged a multi-channel microfluidic chip to assess the effect of long-term RNase A treatment on HUVECs. Blue fluorescent protein (BFP)-labelled HUVECs were pre-treated with or without RNase A and loaded into the left channel with polymerized collagen in the central channel (Fig. 3f, left). Growth into the cell-free collagen region was evaluated after 6 days (fresh RNase A added daily). RNase A treatment significantly enhanced the total migration area (BFP signal) of the HUVECs into the 3D matrix (Fig. 3f and Extended Data Fig. 8h). Staining the extended HUVEC protrusions for F-actin (red) and PECAM1 (green) demonstrated cytoskeleton and cell–cell junctions, respectively, highlighting productive expansion and growth into the collagen (Fig. 3g and Extended Data Fig. 8i). 3D reconstructions of the finger-like protrusions revealed hollow, lumen-like, tubular formations, providing evidence that the RNase A-treated HUVECs can generate structures reminiscent of vessels. The enhanced growth with RNase A treatment observed in 3D further suggests that csRNA is antagonistic to VEGF-A activity.

VEGF-A₁₆₅ selects RNA using arginines

Although previous reports have suggested that exRNA could interact with either the C terminus²⁵ or N terzminus²⁶ of VEGF-A₁₆₅, our data (Fig. 3) point to the C-terminal HS-binding domain (Fig. 4a). To test whether VEGF-A₁₆₅ directly interacts with RNA on the surface of cells, we starved HUVECs for 4 h, added VEGF-A₁₆₅ and UV-C (254 nm) crosslinked the culture to produce covalent bonds between any RNA and VEGF-A₁₆₅. We then immunoprecipitated the VEGF-A₁₆₅ with a validated anti-VEGF-A antibody and found that without UV-C, no RNA was crosslinked to VEGF-A₁₆₅; however, upon crosslinking, the native migrating band disappears (Extended Data Fig. 9a,b). We predicted that this was due to covalent UV crosslinking of VEGF-A₁₆₅ to RNA; consistent with this, RNase treatment of the lysate after UV exposure restored the native migrating VEGF-A₁₆₅ band (Extended Data Fig. 9b). Next, we performed RNA immunoprecipitation sequencing using VEGF-A₁₆₅ to determine the specificity of RNA binding. Using small RNAs from HUVECs and capturing with the anti-VEGF-A antibody, we found that VEGF-A₁₆₅ enriched 235 RNAs over the input and a control omitting the VEGF-A₁₆₅ (Extended Data Fig. 9c,d and Supplementary Table 2). Small nuclear RNA, rRNA, small nucleolar RNA and tRNAs were most abundant in the VEGF-A₁₆₅ immunoprecipitate (Extended Data Fig. 9e). Comparing these hits to the initial set of glycoRNAs (defined in HeLa cells), we found significant overlap of 34 RNAs (Extended Data Fig. 9f and Supplementary Table 2). These data demonstrate specific preference for VEGF-A₁₆₅ binding to certain RNAs, which overlap with known glycoRNA transcripts.

Fig. 4: In vivo control of angiogenesis by RNA-binding defective VEGF-A₁₆₅.

a, Protein structure of VEGF-A₁₆₅ (left), and the amino acid sequence of VEGF-A₁₆₅ WT (top right) and the arginine-to-lysine mutant (bottom right). b, Western blot analysis of whole-cell lysates isolated from HUVECs after starvation and stimulated by 25 ng ml⁻¹ VEGF-A₁₆₅HS WT or VEGF-A₁₆₅HS(R/K) (left). Quantification of the ratio of pERK to total ERK was also calculated across the biological triplicates (right). Statistical assessment was performed with a two-sided Student’s t-test and P values are shown. Data are mean ± s.e.m. c, Western blot analysis of whole-cell lysates isolated from HUVECs after starvation and stimulated by 25 ng ml⁻¹ VEGF-A₁₆₅HS WT or VEGF-A₁₆₅HS(R/K) with or without RNase treatment (top). Quantification of the ratio of pERK to total ERK was also calculated across the biological triplicates (bottom). Statistical assessment was performed with a two-sided Student’s t-test and P values are shown. Data are mean ± s.e.m. d, Representative images of postnatal day 6 retinas of mice 4 h post-intraocular administration of human VEGF-A₁₆₅HS WT (100 ng per retina), VEGF-A₁₆₅HS(R/K) (100 ng per retina) or vehicle (PBS) were immunostained with anti-ERG (red) and IB4 (white). e, Quantification of ERG numbers (left) and IB4 intensity (right) in the vascular front regions (blue dashed line area in panel d). Statistical analysis was performed using an unpaired two-sided Student’s t-test and P values are shown without adjustment. Data are mean ± s.e.m. f, Representative images of fli1a–eGFP zebrafish larvae at 36 h post-fertilization with the indicated injection of mRNA. g, Quantification of the normalized vessel length in panel f was calculated across three independent experiments. The boxplots show the median, the 25th and 75th percentiles (bounds of the box), and minimum and maximum values (whiskers). Statistical assessment was performed with a two-sided Student’s t-test and P values are shown. h, Schematic of HS–RNP complex regulation of VEGF-A signalling and angiogenesis. The schematic was created in BioRender. Flynn, R. (2025) https://BioRender.com/m98s373.

Source Data

We next examined more closely the HS-binding domain of VEGF-A₁₆₅, which is an arginine-rich protein domain (HS WT; Fig. 4a). Arginine-rich domains are well established to facilitate interactions with HS¹⁶ and RNA⁵³. By introducing mutations converting arginine residues to lysine residues (R-to-K mutations) within the HS-binding domain of VEGF-A₁₆₅, we hypothesized that we could selectively perturb RNA binding while preserving the overall positive charge of the domain. To this end, we generated an R/K mutant VEGF-A₁₆₅ HS domain by changing all eight arginine residues in the VEGF-A₁₆₅ HS domain to lysine (HS(R/K); Fig. 4a and Extended Data Fig. 9g). We expressed a WT and R/K version of VEGF-A₁₆₅ and evaluated their ability to stimulate pERK after serum starvation of HUVECs. The addition of the same amount of VEGF-A₁₆₅ HS(R/K) resulted in approximately twofold more pERK activation than the HS WT (Fig. 4b) as well as 2.67-fold more spots per cell (Extended Data Fig. 9h). A similar effect was seen when examining the levels of pVEGFR2 after HS WT or HS(R/K) addition to starved HUVECs (Extended Data Fig. 9i). This increase in pERK and pVEGFR2 levels is reminiscent of the twofold increase in VEGF-A₁₆₅ activation after RNase treatment (Fig. 3b), suggesting that HS(R/K) has reduced detection of RNA.

We next evaluated the sensitivity of the VEGF-A₁₆₅ HS(R/K) mutant to RNase treatment. The VEGF-A₁₆₅ HS(R/K) was less sensitive to RNase treatment and showed high pERK levels regardless of the addition of RNase (Fig. 4c; R/K was 51% less sensitive to RNase than WT). To test whether VEGF-A interacts with sialoglycoRNA and whether this is defective in the HS(R/K) form, we performed an RNA immunoprecipitation followed by rPAL analysis. HS WT was able to enrich for the sialoglycoRNA signal that was RNase sensitive, whereas the HS(R/K) was not (Extended Data Fig. 9j). Microscale thermophoresis analysis of small RNA treated with and without sialidase also revealed a reduction (approximately 35%) in interaction with HS WT after the removal of sialic acids (Extended Data Fig. 9k). To establish whether the HS(R/K) still binds to HS, we selected rHS29, which has N-sulfation, 2-O-sulfation and 6-O-sulfation; on the basis of expression data in HUVECs⁵⁴, NDST1, HS2ST1 and HS6ST1 are expressed (Extended Data Fig. 9l) and VEGF-A₁₆₅ has been reported to interact with N-sulfated, 2-O-sulfated and 6-O-sulfated HS^4,6. In vitro, the VEGF-A₁₆₅ HS(R/K) has a similar interaction with rHS29 as the HS WT VEGF-A₁₆₅ (Extended Data Fig. 9m), in line with the model that the R/K mutation has an RNA-selective phenotype.

VEGF-A₁₆₅ (R/K) mutant enhances angiogenesis

In the neonatal mouse retina, intravitreal injection of exogenous VEGF-A₁₆₅ leads to excessive endothelial cell accumulation, loss of vascular branching and aberrant tip cell morphology⁵⁵. Using this model, we performed intravitreal injection of recombinant VEGF-A₁₆₅HS WT and VEGF-A165 HS(R/K) into the eyes of postnatal day 6 mice. Retinal vasculature was assessed 4 h post-injection using isolectin B4 (IB4) staining to visualize vessels and the endothelial nuclear marker ETS-related gene (ERG) immunostaining to quantify endothelial cells. Consistent with previous findings⁵⁵, both forms of VEGF-A₁₆₅ significantly reduced the number of filopodia in tip cells compared with PBS controls, indicating altered tip cell morphology (Extended Data Fig. 10a). As expected, VEGF-A₁₆₅HS WT induced an increase in IB4 coverage and ERG + endothelial cell abundance at the vascular front (Fig. 4d,e). Injection of HS(R/K) elicited a further enhancement of VEGF activity, marked by significantly elevated IB4 intensity and increased endothelial cell density in the vascular front (Fig. 4d,e). Staining with anti-VEGF-A in the mouse retina across the three examined conditions showed similar patterns of VEGF-A expression confirming that the intravitreal protein injection did not generally disrupt the normal expression pattens of VEGF-A (Extended Data Fig. 10b). These in vivo results are consistent with our HUVEC data where HS(R/K) potentiates VEGF-driven angiogenesis.

We next investigated the conservation of this effect in vivo using zebrafish (Danio rerio) embryos. The role of Vegfa during zebrafish development has been extensively studied (reviewed in ref. ⁵⁶). Vegfa is expressed and secreted by the somites^57,58, and binds to it receptors Kdrl in endothelial cells, inducing its phosphorylation and activating the Mapk cascade, which leads to Erk phosphorylation in specific endothelial cells⁵⁹. We therefore leveraged Tg(fli1a–eGFP), a well-validated zebrafish transgenic model of vascular development^60,61, to assess vascular effects during development. We injected in vitro-transcribed mRNA encoding either haemagglutinin (HA)–vegfaHS WT or HA–vegfaHS(R/K) at the one cell stage of Tg(fli1a–eGFP) embryos and monitored vasculature formation by eGFP fluorescence after 24 h post-fertilization. We verified that both mRNAs were expressed at similar levels by western blotting against the HA tag from protein extracts derived from the injected embryos (Extended Data Fig. 11a). Although HA–vegfaHS WT-injected embryos showed no vascularization defects in the trunk, the vegfaHS(R/K) had a significant defect in the vasculature in the zebrafish trunk displaying underdeveloped trunk vessels (Fig. 4f,g and Extended Data Fig. 11b). We also found that the vegfaHS(R/K) significantly increased the number of endothelial cells in the dorsal aorta and posterior cardinal vein region, whereas vegfaHS WT slightly reduced the number in the same area (Extended Data Fig. 11c).

Finally, to assess whether the activity of the VEGF-A₁₆₅ HS domain is sufficient to alter growth factor signalling, we designed a fusion system between the HS domain and Wnt3a. Wnt3a is a signalling protein and mitogen important for embryonic development among many other processes^62,63. We hypothesized that fusing Wnt3a to the VEGF-A₁₆₅HS WT (Wnt3a–HS WT) would target it to csRNPs and therefore quench its bioactivity, whereas a native Wnt3a or Wnt3a–HS(R/K) fusion would result in reduced anterior development and overall axial shortening of the embryo, as previously reported^64,65. As compared with uninjected embryos, Wnt3a-injected embryos had significantly shorter axial length due to hyperactivation of WNT signalling in the embryo^64,65 (Extended Data Fig. 11d,e). This phenotype was ameliorated if the embryos were injected with an mRNA encoding Wnt3a–HS WT, but the axial length was shortened again after injecting Wnt3a–HS(R/K) (Extended Data Fig. 11d,e); each fusion protein was expressed (Extended Data Fig. 11f). These developmental phenotypes were accompanied by upregulation of WNT signalling downstream targets⁶⁶, with the levels of 6 out of 8 targets elevated in Wnt3a and Wnt3a–HS(R/K), but not in Wnt3a–HS WT (Extended Data Fig. 11g). These data provide evidence that the enhanced activity of HS(R/K) (RNA-binding deficient) is sufficient to drive cell surface ligand binding to enable cellular and organismal phenotypes.

Discussion

HSPGs are widely appreciated to govern critical processes at the cell surface in health and disease. Here we have began to refine this view with the characterization of how glycoRNAs and csRBPs assemble into clustered RNPs nucleated around HSPGs. Through genetic and enzymatic perturbation, we have demonstrated that HS chains are important for csRNP clustering and if removed, can repopulate the cell surface within hours. Molecularly, the sulfation of the HS chains is critical, with N-sulfation and 6-O-sulfation responsible for promoting the csRNP clustering. Finally, we established a new mechanistic pathway for information to be transferred from outside to inside of a cell: through the direct binding and regulation of growth factors such as VEGF-A₁₆₅ to RNAs including glycoRNAs at csRNPs on endothelial cells as demonstrated with 2D, 3D and in vivo models.

Mechanistically, we propose a new role for RNA binding to VEGF-A on the cell surface. VEGFA is composed of 8 exons and has 16 distinct mRNA isoforms reported^67,68,69. VEGF_165a was the first characterized and is the prototypical pro-angiogenic isoform⁷⁰. Amino acids encoded in exons 6a and 7 enable HS binding and are important for bioavailability of the VEGF-A isoforms^71,72,73. Lacking the HS domain (for example, 111 and 121) results in free diffusible forms, whereas incorporation of the HS-binding domain results in extracellular matrix tethering^73,74. Although previous reports have suggested that exRNA operates to positively influence VEGF-A activity^25,26,27, our data have demonstrated that cell-bound csRNPs are negative regulators of VEGF-A through the HS-binding C-terminal domain, and that this domain can interact with sialoglycoRNAs. This nominates a hypothesis: that features of endothelial cells themselves (which present specific types of csRNPs) can differentially regulate and precisely control the activity of VEGF-A to enable physiological angiogenesis. A central molecular paradigm that cells use to communicate with the extracellular environment is the extracellular matrix and, in particular, the assembly of complexes on HS-modified proteins. It will be of great interest to examine the mechanistic and functional role of csRNPs in regulating other well-studied factors with HS-binding domains. Consistent with our data, another recent study has also found a genetic dependency of csRBPs and csRNA (not assayed for glycans) on HSs for cell surface organization⁷⁵. Furthermore, the examination of these mechanisms in more complex and diverse tissue and microenvironment settings could provide new views on intercellular communication.

To contextualize the physiological role of csRNP–VEGF-A association and move away from bulk destruction of csRNA with live-cell RNase treatment, we developed a mutant form of the HS domain of VEGF-A₁₆₅ (R/K) that maintains its cationic nature and binding to HS but reduces its responsiveness to RNase-mediated destruction of csRNPs as well as losing specific interactions with sialoglycoRNA. This enabled the examination of csRNP-dependent and csRNP-independent signalling of VEGF-A on cells, in mice and in zebrafish. In all cases, the R/K mutant mirrored the RNase effect, suggesting a concerted negative action of presented csRNPs on VEGF-A₁₆₅ activity. These data are also consistent with other work described in zebrafish, which have demonstrated that inhibiting secreted RNase activity resulted in defects in trunk vasculature (and spinal neuron aberrations)²⁹. This highlights csRNPs as a ‘counterbalance’ domain that can modulate the magnitude of activity by VEGF-A.

Finally, the role of RNA in regulating biological processes is expansive and spans its catalytic, scaffolding and information-carrying capabilities^76,77,78; however, these functions have traditionally been restricted to intracellular pathways. Our data have demonstrated that canonical signal transduction by VEGF-A₁₆₅ across the plasma membrane is modulated by the presence of csRNPs. Together, our data establish a new type of RNA regulation where the abundance or organization of topologically csRNPs can directly modulate intracellular signalling cascades through selective interaction with growth factors, impacting cellular decision making. Understanding the evolution of HS-binding proteins, glycoRNAs and csRBPs could enable better modelling of developmental and homeostatic processes in complex cellular environments.

Methods

Cell culture

MOLM-13 cells were cultured in Roswell Park Memorial Institute 1640 Medium (Thermo Fisher Scientific). U2OS (American Type Culture Collection (ATCC)) were cultured in McCoy’s 5A Medium (Thermo Fisher Scientific). HEK293T cells were cultured in DMEM media (Thermo Fisher Scientific). MOLM-13, U2OS and HEK293T cells were supplemented with 10% fetal bovine serum (Thermo Fisher Scientific) and 1× penicillin–streptomycin (Thermo Fisher Scientific). Primary umbilical vein endothelial cells (HUVECs; ATCC) were cultured in vascular cell basal medium (ATCC) supplemented with endothelial cell growth kit-VEGF (ATCC). Keratinocyte cells (ATCC) were cultured in dermal cell basal medium (ATCC) supplemented with keratinocyte growth kit (ATCC). All the experiments on HUVECs are performed before passage 10. All cells were cultured at 37 °C with 5% CO₂ and maintained as mycoplasma negative. U2OS cells were transfected with Avalanche-Omni Transfection Reagent (EZ Biosystems). Immortalized HUVECs expressing BFP (Im-HUVEC-BFP; a gift from R. Kamm at MIT)⁷⁹ were cultured in VascuLife VEGF-Mv endothelial medium (Lifeline Cell Technology) and passage p15 was used for experiments.

Live-cell enzyme and chemical treatments

For RNase treatment, RNase A (Sigma) and ShortCut RNase III (New England Biolabs (NEB)) were added directly to the cell culture at a final concentration of 18 µM and 100 U ml⁻¹, respectively. MOLM-13, U2OS, HUVECs and keratinocytes were all treated for 45 min.

For heparinase treatment, heparinase I (NEB), heparinase II (NEB) and heparinase III (NEB) were added directly to the cell culture at a final concentration of 4 units per millilitre (each) for 30 min.

For sodium chloride (NaCl) and sodium chlorate (NaClO₃) treatment, NaCl and NaClO₃ were added to the cell culture at a final concentration of 50 mM for 24 h.

For exogenous Tega HS chain treatment, rHS09 (TEGA Therapeutics) and rHS37 (TEGA Therapeutics) were added directly to the cell culture at a final concentration of 4 µM for 60 min.

For serum starvation, HUVECs were cultured in vascular cell basal medium (PCS100030, Bioresource Center) without endothelial cell growth kit-VEGF (Bioresource Center) after being washed briefly three times with phosphate buffer saline (PBS). VEGF-A₁₆₅ (Thermo Fisher Scientific), VEGF-A₁₂₁ (Thermo Fisher Scientific), EGF (Thermo Fisher Scientific), VEGF-A₁₆₅HS WT (Supplementary Table 3, cell-free Escherichia coli protein production by Liberum Biotech) and VEGF-A₁₆₅ HS(R/K) (Supplementary Table 3, cell-free E. coli protein production by Liberum Biotech) were then added to starved HUVECs at a final concentration of 3 ng ml⁻¹ and 25 ng ml⁻¹ for 5 min, separately. The WT and R/K proteoforms differ by approximately 400 Da (or approximately 1% total mass) so equal mass was added in each experiment, which is approximately equal moles.

Live-cell labelling and microscopy analysis

Adherent cells were cultured on glass coverslips #1.5 (Bioscience Tools) 24 h before labelling. MOLM-13 cells were counted and then blocked as per the manufacturer’s protocol with Human TruStain FcX (Fc block, BioLegend) for 15 min on ice before labelling. For Siglecs staining in live cells, 1 µg ml⁻¹ of recombinant human IgG1 Fc (R&D Systems), Siglec-1–Fc chimera protein (R&D Systems), Siglec-2–Fc chimera protein (R&D Systems), Siglec-3–Fc chimera protein (R&D Systems), Siglec-4–Fc chimera protein (R&D Systems), Siglec-5–Fc chimera protein (R&D Systems), Siglec-6–Fc chimera protein (R&D Systems), Siglec-7–Fc chimera protein (R&D Systems), Siglec-8–Fc chimera protein (R&D Systems), Siglec-9–Fc chimera protein (R&D Systems), Siglec-10–Fc chimera protein (R&D Systems), Siglec-11–Fc chimera protein (R&D Systems), Siglec-14–Fc chimera protein (R&D Systems) and Siglec-15–Fc chimera protein (R&D Systems) were precomplexed with 0.5 µg ml⁻¹ of donkey anti-human IgG AF647 (ImmunoResearch) secondary antibody in FACS buffer (0.5% BSA (Sigma) in 1× PBS) for 45 min on ice. For 9D5, 10E4 and 3G10 staining, 2.5 µg ml⁻¹ 9D5 (Absolute Antibody), 1 µg ml⁻¹ anti-HS 10E4 (Amsbio), 1 µg ml⁻¹ anti-HS 3G10 (Amsbio) and anti-VEGF (R&D Systems) were precomplexed with 1.25 µg ml⁻¹ of goat anti-rabbit AF647 secondary antibody (Thermo Fisher Scientific), 0.5 µg ml⁻¹ of goat anti-mouse AF647 secondary antibody (Thermo Fisher Scientific) or donkey anti-goat IgG AF647 (Thermo Fisher Scientific) secondary antibodies in FACS buffer for 45 min on ice, separately. For VEGF-A₁₆₅, VEGF-A₁₂₁ and VEGFR2 staining, 1 µg ml⁻¹ of anti-VEGF-A₁₆₅ (R&D Systems), 1 µg ml⁻¹ of anti-VEGF-A (Proteintech), and 1 µg ml⁻¹ anti-VEGFR2 (R&D Systems) were precomplexed with 0.5 µg ml⁻¹ of goat anti-rabbit AF647 secondary antibody and donkey anti-goat AF647 secondary antibody (Thermo Fisher Scientific). Precomplexed antibodies were then incubated with cells for 45 min on ice. For DDX21 and hnRNP-U staining, 2.5 µg ml⁻¹ of anti-DDX21 (Novus Biological) and anti-hnRNP-U (Proteintech) were incubated with cells for 45 min on ice. Cells were gently washed twice by FACS buffer and then stained with 2.5 µg ml⁻¹ of goat anti-rabbit AF647 for 30 min on ice.

For 9D5 and Siglecs co-staining, recombinant human Siglec–Fc chimera proteins were precomplexed with donkey anti-human IgG AF488 (ImmunoResearch), 9D5 was precomplexed with goat anti-rabbit AF647 secondary antibody. For 9D5, Siglec-11 and 10E4 co-staining, 10E4 was precomplexed with goat anti-mouse AF568 (Thermo Fisher Scientific). For 10E4, Siglec-11 and DDX21 co-staining, 10E4 was precomplexed with goat anti-mouse AF568 (Thermo Fisher Scientific), Siglec-11 was precomplexed with donkey anti-human IgG AF488 (ImmunoResearch) and anti-DDX21 was incubated with precomplexed 10E4 and Siglec-11 on cells for 45 min on ice. Cells were gently washed twice by FACS buffer and then stained with goat anti-rabbit AF647 for 30 min on ice. For VEGF-A₁₆₅ and Siglec-11 co-staining, anti-VEGF-A₁₆₅ was precomplexed with donkey anti-goat AF647 secondary antibody, recombinant human Siglec-11–Fc chimera protein was precomplexed with donkey anti-human IgG AF488. For 9D5 and DDX21 or hnRNP-U co-staining, mouse host 9D5 (Absolute Antibody) was precomplexed with goat anti-mouse AF488 secondary antibody (Thermo Fisher Scientific). Sequential labelling was used for cs-DDX21 and hnRNP-U labelling as described above. Antibodies or regents were precomplexed in separate tubes with the same concentration as used in single-channel staining and then mixed before adding to cells. After staining, cells were washed three times with ice-cold PBS and a fixation was performed with 3.7% formaldehyde for 15 min at room temperature in the dark. Nuclei were stained with 0.1 µg ml⁻¹ DAPI in PBS. MOLM-13 cells were applied to glass slides using a CytoSpin centrifuge (Thermo Fisher Scientific) at 500g for 5 min.

For mEmerald-positive U2OS cell imaging, cells were fixed with 3.7% formaldehyde for 15 min at room temperature and then stained with DAPI. Finally, all samples above were mounted in ProLong Diamond Antifade Mountant (Thermo Fisher Scientific) and a coverglass was sealed over the samples with nail polish. All samples were then imaged on a Leica SP8 STED ONE microscope with ×63 oil lens. Images were acquired using Leica LAS X software. The DAPI channel was acquired with a PMT detector while all other channels were imaged using hybrid detectors.

For quantification and statistical analysis, at least three random regions of interest (ROIs) from three independent samples were acquired across one or more z-slices. To analyse the colocalization, images were processed using Imaris Microscopy Image Analysis software (Oxford Instruments). A single z-slice from each ROI was taken, selected to be near the middle of the cells (with respect to their z-thickness), and the spot-finder function was used to identify spots of roughly 0.5 µm. In-software background subtraction was used as the default settings, and spots were selected by thresholding spot quality at the elbow of the distribution. This resulted in a series of x and y positions for each spot from each channel, which were then exported for quantitative analysis. Colocalization of spots from paired channels were analysed by implementing a custom Python script (https://github.com/FlynnLab/jonperr) to identify the nearest neighbours of each spot (in nm) with a k-d tree algorithm (scipy.spatial.KDTree). Then, the distances between nearest neighbours were calculated for each pair of targets across all ROIs and plotted in a histogram. To assess the relative fraction of each spot type (channel #1) within the spots of the other pair (channel #2), we calculated a Manders’ colocalization coefficient using the aforementioned Python script. We performed this calculation in both directions: spots of channel #1 in total channel #2 spots, and the reverse.

To quantify and compare the intensities of spots on the cell surface, Leica LAS X software was used to identify ROIs throughout the entire 4× slice z-stack. To quantify and compare the spot numbers of spots on the cell surface, Imaris was used to identify spots throughout the entire 4× slice z-stack. The mean intensities and numbers of ROIs were then divided by the cell numbers and compared across groups. Statistical analysis and data plotting were performed using GraphPad Prism 10.

Live-cell flow cytometry

MOLM-13 cells were directly counted. U2OS cells were gently lifted with Accutase (Sigma-Aldrich) for 3 min at 37 °C, quenched with growth media and then counted. For each condition, 50,000 cells were used. For antibody staining, MOLM-13 cells were blocked as per the manufacturer’s protocol with Human TruStain FcX in FACS buffer for 15 min on ice. Of recombinant human IgG1–Fc, Siglec-7–Fc chimera protein and Siglec-11–Fc chimera protein, 1 µg ml⁻¹ was precomplexed with 0.5 µg ml⁻¹ of donkey anti-human IgG AF647 secondary antibody in FACS buffer for 45 min on ice. Precomplexed antibodies were then added to bind cells on ice for 45 min. For live-cell periodate labelling of cell surface glycans³³, cells were washed twice with cold PBS + Ca + Mg and then incubated at 4 °C in cold PBS + 1 mM sodium periodate for 5 min at 1 million cells per millilitre. Cells were then quenched with 1 mM glycerol added to the PBS, and then cells were washed twice with cold PBS. Cells were then incubated at 4 °C in cold FACS buffer + 25 µM aminooxy-biotin (Cayman Chemical) + 10 mM aniline for 30 min at 1 million cells per millilitre. Cells were mixed via pipetting halfway through the incubation. Cells were then washed once with cold 1× PBS and blocked as per the manufacturer’s protocol with Human TruStain FcX in FACS buffer for 15 min on ice. Cells were then stained for 30 min on ice with streptavidin–AF647 at 1 µg ml⁻¹. After staining, all cells were spun at 4 °C for 3 min at 400g and supernatant was discarded. Cells were washed once with 150 µl of FACS buffer, spun under the same conditions, and finally resuspended in FACS buffer containing 0.1 µg ml⁻¹ DAPI. Data collection occurred on a BD Biosciences LSRFortessa 3, and a gating strategy was used to isolate live, single cells, to examine antibody binding using FlowJo Software (FlowJo LLC).

Western blot

Cells were quickly rinsed with ice-cold PBS and directly lysed with sample buffer (150 mM NaCl, 50 mM Tris and 0.5% Triton X-100, pH 7.4) containing phosphatase inhibitor cocktail (Cell Signaling Technology) on ice for 15 min. Zebrafish samples were sonicated in samples buffer. After centrifugation at 12,000g for 15 min at 4 °C, lysates were heated at 95 °C for 10 min in 1× NuPAGE LDS loading buffer (Thermo Fisher Scientific) containing 5 mM dithiothreitol. Samples were then resolved by SDS–PAGE using AnyKD Criterion TGX Precast Midi Protein Gels (Bio-Rad Laboratories) and transferred to nitrocellulose membranes. Membranes were blocked in blocking buffer and incubated with primary antibodies (diluted in blocking buffer) at 4 °C overnight. After washing three times for 3 min each in 1× PBS with 0.1% Tween-20 (PBST), membranes were incubated with secondary antibodies at room temperature for 45 min, followed by the same three times PBST washing. Membranes were finally rinsed in 1× PBS and scanned on a LI-COR Odyssey CLx scanner. Images and intensity of bands were acquired using LI-COR Image Studio software.

The primary antibodies used were: mouse monoclonal anti-EXT2 (sc-514092, Santa Cruz; immunoblot 1:1,000), anti-EXT1 (sc-515144, Santa Cruz; immunoblot 1:1,000), mouse monoclonal anti-GAPDH (sc-47724, Santa Cruz; immunoblot 1:1,000), rabbit polyclonal anti-GFP (A11122, Invitrogen; immunoblot 1:1,000), mouse monoclonal anti-NDST1 (sc100790, Santa Cruz; immunoblot 1:1,000), mouse monoclonal anti-HS6ST1 (sc-398231, Santa Cruz; immunoblot 1:1,000), mouse monoclonal anti-HS2ST1 (sc-376530, Santa Cruz; immunoblot 1:1,000), rabbit monoclonal anti-phospho-p44/42 MAPK (Erk1/2(Thr202/Tyr204); 4370S, Cell Signaling Technology, immunoblot 1:500), mouse monoclonal anti-p44/42 MAPK (Erk1/2; 4696S, Cell Signaling Technology; immunoblot 1:500), rabbit monoclonal anti-phospho-VEGFR2(Tyr1175) (2478S, Cell Signaling Technology; immunoblot 1:500), rabbit polyclonal anti-VEGFR2 (26415-AP, Proteintech; immunoblot 1:500), mouse monoclonal anti-HA (sc-7392, Santa Cruz; immunoblot 1:1,000), rabbit polyclonal anti-DDX21 (NB100-1718, Novus Biologicals; immunoblot 1:500), rabbit polyclonal anti-hnRNP-U (14599-1-AP, Proteintech; immunoblot 1:500) and rabbit polyclonal anti-WNT3A (26744-1-AP, Proteintech; immunoblot 1:500). The secondary antibodies used were: IRDye 800CW goat anti-rabbit IgG secondary antibody (926032211, LI-COR Biosciences; immunoblot 1:1,000) and IRDye 800CW goat anti-mouse IgG secondary antibody (926-32210, LI-COR Biosciences; immunoblot 1:1,000).

Genome-wide CRISPR–Cas9 screening

MOLM-13 cells expressing Cas9 under blasticidin selection²³ were grown as above and selected with 10 µg ml⁻¹ blasticidin for 3 days to ensure a homogenous starting population. Starting after selection on day 0, 30 million cells were infected with 1:150 the genome-wide single-guide RNA (sgRNA) lentivirus²³ in 60 ml of fresh media with 8 µg ml⁻¹ polybrene. On day 3, cells were spun down and resuspended in 70 ml of fresh media with 1 µg ml⁻¹ puromycin to select for sgRNA-infected cells. On day 5, the media were exchanged for fresh media with 1 µg ml⁻¹ puromycin. On day 6, the cells were switched to normal media without puromycin for expansion. From days 7 to 18, the cells were counted and passaged as needed in fresh media to maintain a cell density between 750,000 and 2,000,000 cells per millilitre. On day 18, cells were counted: 40 million cells were saved before sorting for an input population reference and 200 million cells for each Siglec-11, 9D5 and MAA-I were saved for staining and sorting. To stain, 1,000 µg Siglec-11 was precomplexed with 270 µg secondary antibody, 450 µg 9D5 was precomplexed with 225 µg secondary antibody, and 266 µg of MAA-I was precomplexed with 266 µg streptavidin–AF647. Live-cell staining was performed as noted above with Fc blocking; however, before FACS sorting, cells were filtered over a 40-µm strainer (Corning) and stained with DAPI. Cells were selected for DAPI negative (live) and then the bottom 5% intensity of cells stained with each Siglec-11, 9D5 or MAA-I were sorted into tubes. Cells were collected in FACS buffer after sorting and spun down into pellets at 500g for 4 min at room temperature; input cells were processed in a similar manner to obtain a cell pellet. After removing the supernatant, cell pellets were frozen at −80 °C for later processing. Cells were then processed by resuspending in 200 µl 1× PBS + 5 µl RNaseA + 20 µl proteinase K and incubated at 25 °C for 5 min. Then, 200 µl of Buffer AL (Qiagen) was added and samples were carefully vortexed to mix without shearing genomic DNA (gDNA). The samples were then headed to 56 °C for 60 min. After heating, 200 µl of 100% ethanol was added, gentle vortexing was again used, and then the material was purified over Zymo columns. All spins were performed at 6,000g for 20 s; after spinning the sample through, the columns were washed twice with 80% ethanol and then the DNA was eluted with 2 × 15 µl water. For input samples the initial volumes were scaled up from 200 µl to 1,000 µl of digestion and Qiagen buffers.

To amplify the sgRNAs out of the gDNA, we performed real-time PCR and for each sample performed 12 parallel reactions each with 1 µg of input gDNA. The PCRs were 50 µl final with 200 nM forward and reverse primers with 1× Q5 PCR Master Mix (NEB); 23 cycles of 98 °C for 20 s, 65 °C for 20 s and 72 °C for 90 s were completed. After PCR, the 12 reactions were pooled and purified over a Zymo column following the manufacturer’s recommended protocol for PCR DNA. To add a final index primer for sequencing, a final round of PCR was performed by taking 100 ng of amplified sgRNA library, and five cycles of the above PCR program was run followed by Zymo column clean up. The finally indexed libraries were assessed for size and concentration on a BioAnalyzer High Sensitivity DNA Chip (Agilent). Libraries were pooled equimolar and then sequencing on the NextSeq platform (Illumina) with a 19-bp read 1 and two 8-bp index reads. For the read 1, a custom sequencing primer 5′-TCTTCCGATCTCTTGTGGAAAGGACGAAACACCG-3′ was used. Enrichment of guides and genes were analysed using the MAGeCK statistical package⁸⁰ by comparing read counts from each cell line with counts from matching plasmid as the initial population.

Generation of KO and stably expressing cell lines

All CRISPR–Cas9 KO assays used PX459 (ref. ⁸¹). The target oligonucleotides used are listed in Supplementary Table 3. U2OS cells were transfected with gDNA vectors. Two days later, puromycin (ant-pr-1, Invivogen) was added to the cell culture at a final concentration of 2 µg ml⁻¹ and the live cells were selected by flow cytometry (FACS Calibur 2, BD Science) for isolation of single clones. The expanded individual clones were screened by gDNA sequencing and western blot analysis.

cDNA for human EXT2 was amplified from the cDNA library (Takara Bio); cDNAs for Sulf1 and Sulf2 were gifts from S. Rosen (Addgene plasmid)⁴⁶. All three cDNA were inserted into mEmerald-C1 (Addgene). All plasmids were verified by DNA sequencing. To generate U2OS cells stably expressing mEmerald–EXT2, mEmerald–EXT2(D517N/D573N), mEmerald–Sulf1 or mEmerald–Sulf2, cells were transfected with the indicated plasmids and selected using 200–1,000 µg µl⁻¹ (gradually increasing) G418 (Invivogen) for 2 weeks; green-positive cells were sorted into mono-clones by flow cytometry and cultured in the presence of 200 µg µl⁻¹ G418 for 2 weeks. Proliferated clones were verified by immunoblotting and fluorescence imaging.

UV crosslinking

25 ng ml⁻¹ of VEGF-A₁₆₅ was added to starved HUVECs for 5 min. Cells were treated by UV (60,000 µJ, approximately 2 min) on ice and then directly lysed with samples buffer. For RNase treatment, RNase A and RNase III were added to the samples at a final concentration of 1,000 ng ml⁻¹ and 20 U ml⁻¹, resepectively. Samples were all incubated at 37 °C for 10 min and then lysed on ice for another 10 min. After centrifugation at 12,000g for 15 min at 4 °C, lysates were incubated with 5 µl Protein-G bead (Thermo Fisher Scientific) pre-conjugated with 1 µg of anti-VEGF-A (Proteintech) at 4 °C overnight. The beads were washed three times with PBS and heated at 95 °C for 10 min in 1× NuPAGE LDS loading buffer containing 5 mM dithiothreitol. Samples were then analysed by western blot described above. Anti-VEGF-A₁₆₅ (AF293-NA, R&D Systems; immunoblot 1:1,000) and donkey anti-goat IgG secondary antibody (92632214, LI-COR Biosciences; immunoblot 1:1,000) were used as primary and secondary antibodies, separately.

Microfluidic device setup

IdenTx3 microfluidic chips were purchased from AIM Biotech (#DAX-1). Each chip consists of a central 3D gel channel flanked by two media channels. Type I collagen gel was prepared using a collagen kit (Nitta Gelatin) consisting of Cellmatrix type I-A (3 mg ml⁻¹), 10× Ham’s F12 medium and reconstitution buffer in an 8:1:1 ratio, resulting in a final collagen concentration of 2.4 mg ml⁻¹. The chips were cultured using vascular cell basal medium supplemented with endothelial cell growth kit-VEGF, with or without 10 µM RNase A (0.121 mg ml⁻¹ final concentration).

The central channel of the devices was first filled with 18 µl of collagen. After collagen polymerization for 30 min at 37 °C, one of the medium channels was injected with 15 µl of Im-HUVEC-BFP cell suspension (±RNase A supplementation) at a concentration of 8 million cells per millilitre. The other medium channel was filled with 15 µl of medium ± RNase A. The microfluidic chip was then tilted by 90° for 1–2 min to allow cells to adhere to the collagen gel surface. Subsequently, the four reservoirs in each chip were filled with 50 µl of medium ± RNase A. After 2 h, the medium ± RNase A in the reservoirs was replaced. The devices were cultured for 6 days, and all four reservoirs were replenished daily with 50 µl of medium ± RNase A.

Use and analysis of microfluidic devices

Cells in the idenTx3 chips were fixed with 4% paraformaldehyde in 1× PBS for 20 min and permeabilized with 0.2% Triton X-100–PBS for 5 min. Then, the cells were incubated with 1% BSA in 0.2% Triton X-100–PBS for 30 min to block nonspecific binding. The cells were rinsed with 0.2% Triton X-100–PBS two times for 5 min each and incubated overnight at 4 °C with Alexa Fluor 488-conjugated mouse anti-human platelet/endothelial cell adhesion molecule-1 (PECAM-1; Cell Signaling Technology) at a 1:300 dilution in 0.2% Triton X-100–PBS. Afterwards, the cells were rinsed with 0.2% Triton X-100–PBS three times for 5 min each and incubated with Alexa Fluor Plus 555 Phalloidin (Thermo Fisher Scientific) at a 1:400 dilution in PBS for 20 min at room temperature. Finally, the cells were rinsed three times with PBS and stored in PBS for immunofluorescence imaging.

Images were acquired using a Nikon AXR point scanning confocal microscope with ×10 and ×20 objectives (Nikon), and z-stack images were processed using the Nikon Denoise.ai software and ImageJ. The total migration area was determined by manually measuring the total area of endothelial migration starting from the original gel interface in ImageJ. Statistical analysis (unpaired t-test) was performed in GraphPad Prism 10.

In vitro immunoprecipitation and rPAL

For the small RNA fraction used for immunoprecipitation, TRIzol extractions were performed as previously described in detail⁹. For specifics: TRIzol was added to cultured HEK293T cells. After the initial lysis, the TRIzol’ed samples were incubated at 37 °C to further denature non-covalent interactions. Phase separation was initiated by adding 0.2× volumes of 100% chloroform, vortexing to mix, and finally spinning at 12,000g for 15 min at 4 °C. The aqueous phase was carefully removed, transferred to a fresh tube and mixed with 2× volumes of 100% ethanol. The RNA was then purified over a Zymo column. For all column cleanups, we followed the following protocol. First, 350 µl of pure water was added to each column and spun at 10,000g for 30 s, and the flowthrough was discarded. This step was repeated until all the precipitated RNA was passed over the column once. The column was then washed three times total: once using 400 µl of RNA Prep Buffer (3 M GuHCl in 80% ethanol) and twice with 400 µl of 80% ethanol. The RNA was then treated with proteinase K (Ambion) on the column. Proteinase K was diluted 1:19 in water and added directly to the column matrix and then allowed to incubate on the column at 37 °C for 45 min. Next, eluted RNA was spun out into fresh tubes and a second elution with water was performed. To the eluate, 1.5 µg of the mucinase StcE (Sigma-Aldrich) was added for every 50 µl of RNA and placed at 37 °C for 30 min to digest. The RNA was then cleaned up again using the Zymo column. Of RNA-binding buffer, 2× volumes were added and vortexed, and then 2× (samples + buffer) of 100% ethanol were added and vortexed. After total RNA extraction, the RNA can be further processed to fractionate small RNA (approximately 17–200 nt) and large RNA (more than 200 nt) using Zymo columns, as recommended by the manufacturer. First, an adjusted RNA-binding buffer was made by mixing equal volumes of RNA-binding buffer and 100% ethanol. Two volumes of the adjusted buffer were added to the total RNA and were vortexed thoroughly to mix. The sample was then bound to the column, but the flow through (small RNA fraction) was saved. One volume of 100% ethanol was added to the small RNA and vortexed to mix. The small RNA was then bound to new columns, cleaned up using 2 × 400 µl 80% ethanol, and finally eluted using water. Lyophilized RNA was resuspended in the size-exclusion chromatography buffer (6 M guanidine hydrochloride (Sigma) in 1× PBS) and loaded onto a superose 6 (5/150, Cytiva) column. The sample was run at a flow rate of 0.3 ml min⁻¹ and finally collected into 150-µl fractions. The RNA fractions were cleaned on Zymo column before immunoprecipitation.

For Siglec-11 immunoprecipitation, 5 µl Protein-A bead was pre-conjugated with 10 pmol of IgG–Fc or Siglec-11 in samples buffer for 1 h at room temperature. After washing three times with samples buffer, beads were incubated with 1 µg of small RNA for 3 h at 4 °C. After washing three times with samples buffer, the beads were suspended in 100 µl RNA-binding buffer and heated for 5 min at 55 °C. The beads were removed and the RNA extract solution was transferred to a new tube. Here, 50 µl of pure water was added and vortexed for 10 s, and then 300 µl of 100% ethanol was added and vortexed for 10 s. The RNAs were purified over a Zymo column.

For VEGF immunoprecipitation, 5 µl Protein-G bead was pre-conjugated with 1 µg of anti-VEGF-A (Proteintech) antibodies in samples buffer for 1 h at 4 °C. After washing three times with samples buffer, beads were incubated with 5 µg of VEGF-A₁₆₅HS WT and VEGF-A₁₆₅HS(R/K) immunoprecipitation for 2 h at 4 °C. The beads were washed three times with samples buffer and then incubated with small RNA for 2 h at 4 °C. After washing three times with samples buffer, the beads were suspended in 50 µl RNA-binding buffer and heated for 5 min at 55 °C. The beads were removed and the RNA extract solution was transferred to a new tube. Here, 100 µl of pure water was added and vortexed for 10 s, and then 300 µl of 100% ethanol was added and vortexed for 10 s. The RNAs were purified over a Zymo column.

For rPAL labelling, experiments were performed as previously described¹⁰. Specifically, lyophilized RNAs were suspend with 28 µl blocking buffer (1 µl 16 mM mPEG3-Ald (BroadPharm), 15 µl 1 M MgSO₄ and 12 µl 1 M NH₄OAc pH5 (with HCl)) and then incubated for 45 min at 37 °C. One microlitre of 30 mM aldehyde reactive probe (ARP/aminooxy biotin, Cayman Chemicals) was added first, then 2 µl of 2 mM NaIO₄ (periodate) was added. The periodate was allowed to perform oxidation for exactly 10 min at room temperature in the dark. The periodate was then quenched by adding 3 µl of 22 mM sodium sulfite. The reaction was allowed to proceed for 5 min at 25 °C and then moved to 35 °C for 90 min. The reaction was then cleaned up by the Zymo column. The RNAs were eluted from the column using 2 × 6.2 µl water. For in vitro RNase treatment, the RNase cocktail enzyme mix (Thermo Scientific) was added directly to RNA samples with RNase buffer (200 mM Tris pH 7.5, 1 M KCl and 1 mM MgCl₂) at a final concentration of 100 U ml⁻¹ and 4,000 U ml⁻¹, separately. The digestion was incubated at 37 °C for 1 h. After RNase digestion, the samples (digested or mock treated) were directly combined with Gel Loading Buffer II (95% formamide, 18 mM EDTA and 0.025% SDS) with a final concentration of 1× SybrGold (Thermo Fisher Scientific) and denatured at 55 °C for 10 min. Immediately after this incubation, the RNA was placed on ice for at least 2 min. The samples were then loaded into a 1% agarose, 0.75% formaldehyde and 1.5× MOPS buffer (Lonza) denaturing gel. RNA was electrophoresed in 1× MOPS at 115 V for 34–45 min, depending on the length of the gel. Subsequently, the RNA was visualized on a UV gel imager. The RNA was transferred with 3 M NaCl pH 1 (with HCl) to a nitrocellulose membrane for 90 min at 25 °C. Post-transfer, the membrane was rinsed in 1× PBS and dried on Whatman Paper (GE Healthcare). Dried membranes were rehydrated in Intercept Protein-Free Blocking buffer (LI-COR Biosciences) for 30 min at room temperature. After the blocking, the membranes were stained using IRDye IR800 streptavidin for 30 min at 25 °C. Excess streptavidin–IR800 was washed from the membranes using three washes with 0.1% Tween-20 in 1× PBS for 3 min each at 25 °C. The membranes were scanned on a LI-COR Odyssey CLx scanner (LI-COR Biosciences). Images were acquired using LI-COR Image Studio software.

VEGF immunoprecipitation sequencing

Of Protein-G beads, 5 µl was pre-conjugated with 1 µg of anti-VEGF-A antibody (Proteintech) in samples buffer for 1 h at 4 °C. After washing three times with samples buffer, beads were incubated with 1 µg of VEGF-A₁₆₅ (Thermo Fisher Scientific) for 2 h at 4 °C. The beads were then incubated with 1 µg of biological triplicates small RNA extracted from HUVECs for 2 h at 4 °C. After washing three times with samples buffer, the beads were suspended in 50 µl RNA-binding buffer and heated for 5 min at 55 °C. The beads were removed and the RNA extract solution was transferred to a new tube. Of pure water, 100 µl was added and vortexed, and then 300 µl of 100% ethanol was added and vortexed for 10 s. The RNAs were purified over a Zymo column and then lyophilized overnight. Of HUVEC small RNA, 100 ng served as input. Sequencing libraries were generated largely as reported in ref. ⁸², with the following new 3′ ligation linker, RT primer and cDNA ligation linker ordered from IDT (Supplementary Table 3). Libraries were sequenced on the Illumina NovaSeq X Plus in single-end mode (read 1, 100 bp). Raw reads were demultiplexed, unique molecular identifiers (UMI) extracted and adapter trimmed using Cutadapt (v4.9)⁸³. Reads were aligned to a custom reference of human small non-coding RNAs⁸⁴ (https://github.com/y9c/m6A-SACseq/tree/main/db) using Bowtie2 (v2.5.4)⁸⁵ with -k 10 to retain up to 10 alignments per read. Aligned reads were deduplicated by UMI and position using a published Python script⁸⁶. In brief, reads with the same UMI and alignment position were grouped; within each group, if one alignment had the highest score, it was retained, otherwise one of the tied top-scoring alignments was selected at random.

Deduplicated transcript-level UMI counts were used for differential analysis in DESeq2 (v1.42.1). Two DESeq2 models were built: one comparing immunoprecipitation versus input (pI), and one comparing control versus input (cI). Size factors were estimated from all transcripts excluding long rRNAs (18S, 28S and 45S). To detect transcripts specifically enriched in the immunoprecipitation condition, we defined a centred Z-score:

$${z}_{\mathrm{centred}}=\frac{({\text{log}}_{2}{\mathrm{FC}}_{\mathrm{pl}}-{\text{log}}_{2}{\mathrm{FC}}_{\mathrm{cl}}+{\Delta }_{\mathrm{SF}})}{\sqrt{{{\mathrm{SE}}_{\mathrm{pI}}}^{2}+{{\mathrm{SE}}_{\mathrm{cI}}}^{2}}}$$

Where

$${\Delta }_{\mathrm{SF}}={\log }_{2}\frac{\mathrm{size}\,{\mathrm{factor}}_{\mathrm{cI}}}{\mathrm{size}\,{\mathrm{factor}}_{\mathrm{pI}}}$$

is the average size-factor correction across replicate-matched samples. An empirical null was estimated from Z-scores with ∣Z∣ < 0.5, assuming a Gaussian distribution. Wald tests were performed, and P values were adjusted using the Benjamini–Hochberg procedure.

RIP hits were defined by the following criteria:

• DESeq2 filter: adjused P_pI < 0.05 and log₂FC_pI  > 0.5

• True positives: passing the DESeq2 filter and adjused P_centred < 0.05 and Z_centred > 0

Plots were generated using R (v4.3.3).

Electrophoretic mobility shift assay

Of rHS29 (TEGA Therapeutics), 2 µM was incubated with 1 µg of VEGF-A₁₆₅ HS WT or VEGF-A₁₆₅ HS(R/K) in PBS for 30 min at 25 °C. Of 50% glycerol, 4 µl was added to 10 µl of sample and the mix was then loaded into anyKD PAGE gel without SDS in the running buffer. Electrophoresis was conducted at constant voltage (150 V) at 4 °C. Gel was stained with Bulldog blue dye for 30 min at room temperature, washed three times with pure water for 5 min each, and finally scanned on a LI-COR Odyssey CLx scanner.

Microscale thermophoresis assay

Fluorescent HUVEC small RNA served as the target and VEGF-A served as the ligand in the microscale thermophoresis assay. For small RNA end repair, 10 µg of small RNA was incubated with 2.5 µl of FastAP (Thermo Fisher Scientific), 2.5 µl of T4 PNK (NEB) and 10 µl of reaction buffer (NEB) for 45 min at 37 °C. After 3′ end repair, the RNAs were purified over a Zymo column and eluted in 2 × 10 µl of pure water. For ligation, 2.5 µg of small RNA was incubated with 1.5 µl of 1 mM ATP, 3 µl of DMSO, 4.86 µl of 10 µM AF647 pCp (Jena Bioscience), 2 µl of 30 U T4 RNA ligase (NEB), 9 µl of 15% PEG8000 and 3 µl of 10× reaction buffer (NEB) overnight at 16 °C. The RNAs were then purified over a Zymo column. The binding of labelled small RNA to VEGF-A was measured on a Monolith NT.115 machine (NanoTemper) with the software MO.Control. Small RNA concentration was set as 5 nM (estimated from the average molecular weight of the small RNA pool). The VEGF-A₁₆₅ gradient was set from 5.5 µM to 0.000336 µM; the VEGF-A₁₂₁ gradient was set from 15 µM to 0.000458 µM. Excitation power was 4% and microscale thermophoresis power was 40% at 25 °C. Data from two independent biological replicates were analysed by GraphPad Prism.

Animal models

Zebrafish (D. rerio) lines were housed in AAALAC-accredited facilities and maintained in accordance with protocols approved by the Massachusetts Institute of Technology’s Committee on Animal Care. All experiments were conducted using zebrafish of the AB/Tübingen (TAB5/14) genetic background. The experiments involved zebrafish embryos at stages ranging from 1 to 5 days post-fertilization. As sex determination in zebrafish occurs after these stages, sex was not considered a variable in this study. Embryos were raised in clutches under standard growth conditions, without allocation into specific experimental groups or isolation.

Mouse work was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the US National Institutes of the Health. All of the animals were handled according to the approved institutional animal care and use committee protocols (#00001771) of the Boston Children’s Hospital. Sex was not considered a variable in this study. Every effort was made to minimize suffering.

Blinding strategy was applied in all animal experiments. Animals were all picked randomly.

Intraocular injection and retinal analysis

Intraocular injection of human recombinant WT VEGF-A₁₆₅ HS WT and VEGF-A₁₆₅ HS(R/K) was performed in postnatal day 6 C57BL/6 pups based on previous publications with minor modifications^55,87. Pups were first anaesthetized by hypothermia. The eyelids were carefully separated by a 30-G needle and eyeballs were lifted by a forceps for injection. Of VEGF-A₁₆₅ HS WT and VEGF-A₁₆₅ HS(R/K) both at a concentration of 100 ng μl⁻¹ or saline (vehicle control), 1 µl was injected into the vitreous humor using a Nanofil syringe attached with a 35-G needle (World Precision Instrument). Pups were placed on a heating pad and returned to the mother’s care. After 4 h, eyeballs were collected and processed for retina immunohistochemistry.

Retina immunohistochemistry is based on a publication with minor modifications⁸⁷. In brief, isolated eyeballs from postnatal day 6 pups were fixed in 4% paraformaldehyde in PBS at room temperature for 30 min. Retinas were isolated under dissection microscopy, permeabilized in 0.1% Triton X-100 and 0.1% PBST at room temperature for 30 min, and blocked with 1% BSA in PBST at room temperature for 30 min. Primary antibodies were added to the sections with indicated dilution in PBST (anti-ERG from Abcam, dilution 1:500; anti-mVEGF₁₆₅ from R&D system, dilution 1:500) and Alexa Fluor 647 isolectin GS-IB4 conjugate (Invitrogen, dilution 1:500) at 4 °C overnight. Retinas were further washed three times with PBST for 10 min at room temperature and incubated with fluorescent-conjugated secondary antibodies (Alexa 546 donkey anti-rabbit IgG, Alexa 488 donkey anti-goat IgG) with 1:500 dilution in PBST for 1 h at room temperature. Stained retinas were washed three times with PBST for 10 min at room temperature and mounted using Fluoromount-G slide mounting medium (Southern Biotech) on microscope slides (Thermo Fisher Scientific). Flat mount retina slides were proceeded to image with confocal microscopy.

Quantification of retinal images

Quantifications were based on previous publications^87,88. Images were acquired using a LSM810 or LSM910 confocal microscope (Zeiss) equipped with an EC Plan-Neofluar ×10/0.3, a Plan-Apochromat ×20/0.8 or a Plan-Apochromat ×63/1.4 Oil DIC objective. Images were taken using Zen2.1 software (Zeiss) and processed and quantified with Fiji (NIH).

For vascular front and endothelial cell number quantification, tile-scanned images of flat-mounted retinas stained with IB4 and ERG were used. Vascular front areas were manually lassoed based on the first three rows of ERG⁺ cell number and IB4 signal. Numbers of ERG⁺ cells were manually counted. A value of average signals from a single retina is represented as n = 1.

For filopodia number quantification, tile-scanned images of flat-mounted retinas stained with IB4 were used. Filopodia numbers were manually counted from images taken with ×10, ×20 or ×63 objective. A value of average signals from a single retina is represented as n = 1.

Wnt3a–VEGF-A–HS fusions

gBlocks (double-strand DNA) of Wnt3a, Wnt3a–HS WT and Wnt3a–HS(R/K) (Supplementary Table 3) were purchased from IDT, and then used as templates to perform PCR with primers (GGATCCTAATACGACTCACTATAGGGAGA served as the forward primer for all three DNAs; GAATTCTTACTTGCAGGTGTGCACGT served as the reverse primer for Wnt3a; GAATTCTTACCGCCTCGGCTTGTCACATC served as the reverse primer for Wnt3a–HS WT; and GGATTCTTATTTTTTCGGCTTGTCACATT served as the reverse primer for Wnt3a–HS(R/K)). All PCR products were digested by EcoRI (R3101M, NEB) and BamHI (R3136M, NEB), and then ligated into pcDNA3.1 (Invitrogen). All plasmids were verified by DNA sequencing. Double-strand DNAs of Wnt3a, Wnt3a–HS WT and Wnt3a–HS(R/K) were generated through EcoRI/BamHI digestion, agarose gel separation and purification. For the in vitro transcription reaction, 500 ng and 300 ng of linearized, digested DNA were used in an overnight (2-h) setup with the mMESSAGE mMACHINE T7 ULTRA Transcription Kit (AM1345, Invitrogen). The resulting in vitro transcription product was treated with TURBO DNase, included in the same kit, for 15 min. Following DNase treatment, the samples underwent poly-A tailing using the tailing reaction provided by the mMESSAGE mMACHINE Kit. Before adding the ePAP (NEB) enzyme, 2 μl of the sample was set aside for later analysis via RNA agarose gel to confirm band size and serve as a control for the post-poly-A tailing reaction. The samples were then purified using the Zymo RNA Clean and Concentrator kit (Zymo Research).

Zebrafish embryo analysis

Tg(fli1a–eGFP) zebrafish line and WT zebrafish were applied for Vegfa and Wnt3a injection, separately. Zebrafish were crossed by placing a male and female in a shared container, keeping them physically separated until the following morning when they were combined for fertilization. Embryos were collected for injection 30 min after combining the males and females. The injection mix, consisting of 35 ng μl⁻¹ in vitro-transcribed RNA encoding Wnt3a or 17.5 ng µl⁻¹ in vitro-transcribed RNA encoding VEGF-A₁₆₅ and 0.07% Phenol Red solution (Sigma), was prepared and loaded into needles made from glass capillary tubing (64-0766). Each embryo was injected with 1 nl of the injection mix using a picolitre injector (Warner Instruments). Injected embryos, along with uninjected controls, were incubated in fish water at 28 °C until collection at 24–36 h post-fertilization.

Zebrafish embryos at 24 h post-fertilization were collected, dechorionated using ethanol-sterilized forceps, and transferred to fresh fish water. Individual embryos were anaesthetized by transferring them into 1X Tricaine solution, prepared by diluting a 25X Tricaine stock solution (pH 7, containing 4 mg ml⁻¹ Tricaine-S (Syndel) and 20 mM Tris) to 1X in fish water. Following anaesthesia, embryos were fixed in 4% paraformaldehyde in fish water. Tg(fli1a–eGFP) zebrafish samples were permeabilized in Triton X-100 and then immunolabelled with anti-GFP antibody. For imaging, embryos were embedded in 1.5% low-melt agarose (50100, Lonza) and transferred to a clear dish suspended in low-melt agarose. Images were captured using a Leica M205 FCA fluorescence stereo microscope at ×5.5 magnification, with the same settings applied to all embryos. For vessel quantification in fli1a–eGFP line, the longitudinal length of all vessels (termed ‘L’) along axial with the same length (termed ‘l’ = 200 μm, anterior to posterior) was measured in ImageJ. The ratio L:l was quantified.

qRT–PCR

Quantitative reverse transcriptase PCR (qRT–PCR) was performed using Brilliant II SYBR Green qRT–PCR one-step Master Mix kit (600825, Agilent) on a CFX Connect Real-Time System (Bio-Rad). Of RNA, 60 ng was used in the reaction with 1.25 µM primers. The data were analysed using the ^ΔΔCt method. 18S rRNA was used as a control to normalize the expression of target genes⁸⁹. The primer sequences used for RT–PCR in this study are listed in Supplementary Table 3.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.