High-density lipoprotein regulates angiogenesis by long non-coding RNA HDRACA – Signal Transduction and Targeted Therapy

nHDL and dHDL cause different expression of HDRACA in endothelial cells

To identify lncRNAs involved in the nHDL and dHDL regulation of angiogenesis, we analyzed previous lncRNA sequencing results to select lncRNAs for their regulatory effects on angiogenesis.²⁶ We preferred transcripts longer than 200nt in the Ensembl genome browser database to obtain preliminary information on the transcripts. We searched for transcripts with higher expression levels in the dHDL than in the nHDL group (P value ≤ 0.05, Fold Change ≥ 1.5). We focused on transcripts with relatively high abundance in the dHDL group (RPKM ≥ 0.5).²⁷ Five candidate transcripts were identified (Fig. 1a, b). These five transcripts were knocked down using specific small interfering RNAs (siRNAs) and antisense oligonucleotides (ASONs) (Supplementary Fig. S1a) to evaluate their effects on angiogenesis. Knockdown of ENST00000562749.1 promoted both tube formation and proliferation of human umbilical vein endothelial cells (HUVECs), whereas knocking down the other four candidate transcripts had no effect (Fig. 1c, d; Supplementary Fig. S1b, d). Furthermore, knockdown of any of these five transcripts had no effect on endothelial cell migration (Supplementary Fig. S1c, e). Therefore, ENST00000562749.1 was chosen for further study.

In the Ensembl genome browser, ENST00000562749.1 is encoded by MRFAP1P1 gene that is located on the forward strand of Chromosome X. To determine the exact sequences of ENST00000562749.1, we performed 5ʹ and 3ʹ rapid amplification of complementary DNA ends (RACE) assays and found a 553-nt transcript (Fig. 1e; Supplementary Fig. S2a). We termed it HDRACA. Data from the coding potential assessment tool (CPAT) showed that HDRACA possessed very low coding potentials (Supplementary Fig. S2b). According to ORFfinder, HDRACA does not possess an open reading frame (ORF) longer than 300 nt (Supplementary Fig. S2c). Furthermore, we overexpressed these ORFs in HEK-293T cells and confirmed that none of the ORFs encoded peptides (Supplementary Fig. S2d). A fluorescence in situ hybridization (FISH) assay demonstrated that HDRACA was primarily localized in the cytoplasm of HUVECs (Fig. 1f), as verified by nuclear and cytoplasmic fractionation (Supplementary Fig. S2e). Real-time quantitative PCR (RT-qPCR) results indicated no significant changes in HDRACA levels at various time points during tube formation (Supplementary Fig. S2f). To examine the distribution of HDRACA in different types of vascular cells, droplet digital PCR (ddPCR) was performed to quantify the HDRACA levels in HUVECs, human aortic smooth muscle cells (HASMCs), human aortic fibroblasts (FBs), and THP-1 macrophages. We verified that HDRACA was expressed in various human blood vessel cells, but its expression was the highest in endothelial cells (Supplementary Fig. S2g). RT-qPCR verified the expression of HDRACA in endothelial cells stimulated with nHDL and dHDL. We found that nHDL reduced HDRACA expression in various types of endothelial cells, whereas dHDL was less effective than nHDL at downregulating HDRACA (Fig. 1g). We further verified this result using FISH assays (Supplementary Fig. S2h). These data suggest that HDRACA is a 553-nt lncRNA primarily localized in the cytoplasm of HUVECs, and nHDL can downregulate HDRACA expression levels in HUVECs, whereas dHDL is less effective than nHDL, which is consistent with the lncRNA-seq results.

HDL-bound S1P regulate the transcription of HDRACA in the endothelial cells by affecting the ubiquitination of KLF5

The differential regulation of HDRACA by nHDL and dHDL prompted us to explore its potential mechanism. As HDL usually regulates endothelial cell activities by interacting with cellular receptors, we designed siRNAs to knock down the five most common HDL receptors on endothelial cells (Supplementary Fig. S15a–e and S15l–p). Knockdown of ATP-binding cassette transporter A1 (ABCA1), ATP-binding cassette transporter G1 (ABCG1), SRB1, and S1P receptor 3 (S1P3) did not affect the regulatory effects of nHDL and dHDL on HDRACA (Supplementary Fig. S3a–d). However, when S1P1 was knocked down, both nHDL and dHDL failed to down regulate HDRACA expression (Supplementary Fig. S3e). W146, an S1P1 inhibitor, blocked the ability of nHDL and dHDL to decrease HDRACA levels (Fig. 2a).

HDL-bound S1P regulates the transcription of HDRACA in the endothelial cells by affecting the ubiquitination of KLF5. a RT-qPCR showed the effect of W146 on HDRACA levels in HUVECs treated with nHDL or dHDL. b The S1P content in nHDL (n = 20) and dHDL (n = 20). c The levels of HDRACA in HUVECs treated with increasing concentrations of S1P (from 0.5 μM to 8 μM) were assayed by RT-qPCR. d The levels of HDRACA in HUVECs treated with S1P (1 μM), r-ApoM (1 μM), and r-ApoM-S1P (1 μM) were assayed by RT-qPCR. e The middle tracks display the online dataset of KLF5 ChIP-seq in GM12878 (blue; GSE127670), H3K4me3 ChIP-seq in HUVECs (red; GSM945181, GSM733673, GSE96250) and H3K27ac ChIP-seq in HUVECs (yellow; GSM733691). The purple arrow indicates the putative KLF5 binding site in HDRACA promoter based on the conserved KLF5 binding motif (shown on the top) and peaks. f The mRNA levels of KLF5 in HUVECs treated with nHDL or dHDL were assayed by RT-qPCR. g Immunoblot of KLF5 in HUVECs treated with nHDL or dHDL. The representative plots (left) and quantification (right) are shown. h KLF5 ubiquitination in HUVECs treated with nHDL or dHDL was assayed by immunoprecipitation (IP) and immunoblotting (IB). MG132 was added to inhibit KLF5 degradation. i RT-qPCR confirmed the mRNA levels of HDRACA in HUVECs transfected with negative control siRNA or KLF5-siRNAs. j ChIP-RT-qPCR analysis (right) for KLF5 binding to the HDRACA promoter in HUVECs. Normalized data are shown as percentages of the input controls. KLF5 binding to BECN1 promoter served as a positive control. The whole-cell lysate (INPUT) and immunoprecipitated proteins (IP) of each immunoprecipitation were analyzed by immunoblotting (left) for KLF5 or IgG. k Luciferase reporter assays for HUVECs with endogenous KLF5 expression transfected with pGL4.1 reporter plasmids containing deletion HDRACA promoter constructs. HUVECs transfected with a blank pGL4.1 plasmid served as a negative control. l Luciferase reporter assays for HUVECs with endogenous KLF5 expression transfected with pGL4.1 reporter plasmids containing wild type or mutated HDRACA promoters. HUVECs transfected with a blank pGL4.1 plasmid served as a negative control. Data are presented as the mean ± SD. For all the experiments, n = 6. **p < 0.01; ***p < 0.001; ****p < 0.0001; ns not significant

S1P is an endogenous agonist of S1P1, and S1P is primarily carried by HDL particles in the plasma.²⁸ Both S1P and HDL plasma levels were lower in patients with CAD than in healthy individuals (Supplementary Fig. S3f, g). However, when we extracted HDL from the plasma to measure its S1P content, dHDL from patients with CAD contained significantly lower S1P content than did nHDL from healthy individuals (Fig. 2b), which is consistent with a previous report.²⁹ Plasma HDRACA levels were not significantly different between CAD patients and healthy individuals (Supplementary Fig. S3h). RT-qPCR results showed that S1P decreased HDRACA expression in HUVECs in a dose-dependent manner (Fig. 2c). Most S1P is bound to ApoM in HDL.³⁰ ApoM levels were not significantly different between nHDL and dHDL (Supplementary Fig. S3i). Recombinant human ApoM (r-ApoM)-bound S1P (r-ApoM-S1P) showed a stronger inhibitory effect than free S1P on HDRACA in HUVECs (Fig. 2d). We administered nHDL and dHDL along with exogenous S1P. The results suggested that both nHDL and dHDL could efficiently take up S1P and S1P-loading could rescue the inhibitory effect of dHDL on HDRACA (Supplementary Fig. S3j, k). Additionally, we examined whether nHDL or dHDL could carry and deliver HDRACA into endothelial cells. We extracted RNA from nHDL and dHDL for lncRNA microarray analysis. The results showed that nHDL and dHDL carried numerous lncRNAs, including HDRACA (Supplementary Fig. S3l). Further, RT-qPCR indicated that there was no significant difference in the HDRACA content between nHDL and dHDL (Supplementary Fig. S3m). We used W146 to block the inhibitory effect of HDL-bound S1P on HDRACA expression in HUVECs, followed by treatment with α-amanitin, a specific inhibitor of RNA polymerases II and III, to inhibit endogenous RNA transcription. Neither nHDL nor dHDL affected HDRACA levels in HUVECs (Supplementary Fig. S3n). Collectively, nHDL and dHDL affected the levels of HDRACA by activating S1P1 in endothelial cells via S1P, rather than delivering HDRACA into endothelial cells.

According to JASPAR, a database of transcription factor binding profiles, KLF5, a transcription factor closely related to angiogenesis,³¹ was predicted to bind to the promoter of HDRACA with a canonical binding motif (Fig. 2e). An online chromatin immunoprecipitation (ChIP) sequencing dataset (GSE127670) revealed the peaks of KLF5 at the HDRACA promoter in GM12878 cells (Fig. 2e). Moreover, several online histone ChIP-seq datasets (GSM945181, GSM733673, GSE96250, and GSM733691) showed that H3K4me3 and H3K27ac peaked in HUVECs in the same region as the two KLF5 peaks in the HDRACA promoter in GM12878 cells (Fig. 2e). Therefore, we evaluated whether nHDL and dHDL differentially regulated KLF5 activity to affect HDRACA levels. Although neither nHDL nor dHDL affected the levels of KLF5 mRNA in HUVECs (Fig. 2f), nHDL effectively downregulated KLF5 protein levels, whereas dHDL had less of an effect (Fig. 2g). In addition, r-ApoM-S1P did not affect KLF5 mRNA levels in HUVECs but significantly downregulated KLF5 protein levels (Supplementary Fig. S4a, b). Furthermore, S1P1 knockdown blocked the ability of nHDL, dHDL, and r-ApoM-S1P to downregulate KLF5 protein levels without affecting KLF5 mRNA levels (Supplementary Fig. S4c–f). Previous studies have reported that KLF5 is an unstable protein that can be ubiquitinated by multiple ubiquitin-protein ligases and degraded by the ubiquitin-proteasome system.³² Treatment of HUVECs with the proteasome inhibitor MG132 increased KLF5 protein levels and attenuated the decrease in KLF5 protein levels after nHDL stimulation (Supplementary Fig. S4g). As shown in Fig. 2h, nHDL increased the ubiquitination of KLF5 in HUVECs, while dHDL had much less effect, suggesting that nHDL and dHDL differentially regulate the ubiquitination of KLF5 to affect its degradation. r-ApoM-S1P promoted KLF5 ubiquitination (Supplementary Fig. S5a).

WWP2, an E3 ubiquitin ligase, can facilitate multiple types of ubiquitination of its substrates, leading to proteolysis or changes in activity, depending on the type of E2 ubiquitin-conjugating enzyme it associates with.^33,34,35 S1P1 is constitutively associated with WWP2, and WWP2 is able to catalyze KLF5 ubiquitination.^36,37 A recent study also indicated the important role of WWP2 in protecting vessels from oxidative stress-associated endothelial injury.³⁸ Therefore, we investigated whether S1P activates WWP2 to induce KLF5 ubiquitination in HUVECs. WWP2 knockdown effectively attenuated r-ApoM-S1P-induced ubiquitination and downregulation of KLF5 (Supplementary Fig. S5b, c; Supplementary Fig. S15f, q). The WW2-WW3 linker (2,3-linker) is critical for the autoinhibition of WWP2 enzymatic activity.³⁹ However, phosphorylation at two tyrosine sites, Tyr369 and Tyr392, in the 2,3-linker region can disrupt the inhibitory effects of 2,3-linker on WWP2 catalysis.³⁹ S1P-S1P1 signaling promotes the phosphorylation of protein tyrosine sites by activating receptor protein tyrosine kinases or non-receptor tyrosine kinase.^40,41 To investigate whether S1P-S1P1 signaling activated WWP2 by promoting the phosphorylation of its tyrosine sites, we transduced HUVECs with lentiviruses encoding FLAG-tagged wild-type WWP2 or mutant WWP2, in which Tyr369 or Tyr392 was replaced with phenylalanine (Y369F or Y392F) (Supplementary Fig. S5d). Supplementary Fig. S5e shows that r-ApoM-S1P promoted the tyrosine phosphorylation of WWP2, which was attenuated by the Y369F or Y392F mutation. The Y369F and Y392F mutations in WWP2 decreased r-ApoM-S1P induced KLF5 ubiquitination (Supplementary Fig. S5f). WWP2 interacts with the PY2 motif of KLF5 to promote KLF5 ubiquitination.³⁷ PY2 has been demonstrated to directly degraded KLF5.⁴² Transducing HUVECs with lentiviruses encoding KLF5 with a PY2 deletion (KLF5-ΔPY2) effectively upregulated KLF5 levels, even after nHDL treatment (Supplementary Fig. S5g; Supplementary Fig. S15v). Knocking down KLF5 using siRNAs downregulated HDRACA levels in HUVECs, while overexpressing KLF5-ΔPY2 significantly upregulated HDRACA levels and antagonized the inhibition of nHDL on HDRACA (Fig. 2i; Supplementary Fig. S5h, i; Supplementary Fig. S15g, r).

Next, we searched for the binding sites for KLF5 based on the conserved binding motif and online ChIP-seq datasets. We found a canonical-GCCCCGCCCC-binding sequence in the common region of KLF5 and histone peaks in the promoter region of HDRACA [+73 to +82] (Fig. 2e). To validate the association between KLF5 and the predicted transcription factor-binding site, we performed ChIP-RT-qPCR and confirmed the interaction between KLF5 and the HDRACA promoter region (Fig. 2j). We used the BECN1 promoter as a positive control, which has been previously shown to bind to KLF5.⁴³ Additionally, KLF5 knockdown attenuated KLF5 enrichment in the HDRACA promoter (Supplementary Fig. S5j). We further transfected HUVECs with a series of luciferase reporter constructs carrying sequential deletions of the 5ʹ-flanking regions of HDRACA. We observed enhanced luciferase activity with sequence deletion from −1115 to −65 base pairs upstream of the HDRACA transcription start site (Fig. 2k). Additionally, a nucleotide mutation at the KLF5 binding site [+73 to +82] abrogated luciferase activity following transfection of the pGL4.1 reporter plasmid (Fig. 2l).

Taken together, these data suggest that nHDL-bound S1P interacts with S1P1 to activate WWP2, promotes the ubiquitination-mediated degradation of KLF5, and inhibits the transcription of HDRACA, whereas dHDL does not significantly affect KLF5 ubiquitination and degradation because of its lower S1P content, resulting in less effective regulation of HDRACA transcription.

nHDL and dHDL differently regulate HDRACA expression to affect endothelial cell proliferation and tube formation

To further address the role of HDRACA in angiogenesis, we silenced HDRACA using a specific mixture of siRNAs and ASONs (lncRNA Smart Silencer), and overexpressed HDRACA by transducing HUVECs with lentiviral vectors. The lncRNA Smart Silencer efficiently silenced HDRACA (downregulated more than 80%), and the lentivirus carrying the full-length sequence of HDRACA significantly increased HDRACA levels in HUVECs (Supplementary Fig. S6a, b). After silencing HDRACA in HUVECs, we performed mRNA sequencing and subjected the differentially expressed genes to GO and KEGG pathway analyses (Fig. 3a; Supplementary Fig. S6c–e). These results indicate that the cell cycle was the most regulated pathway upon HDRACA silencing in HUVECs (Supplementary Fig. S6d). Flow cytometry analysis revealed that silencing HDRACA increased the fraction of HUVECs in the S/G2/M phase (Fig. 3b). In contrast, transfection of HUVECs with a lentivirus encoding HDRACA decreased the proportion of HUVECs in S/G2/M (Fig. 3c). We evaluated the role of HDRACA in HUVECs apoptosis using Annexin V-FITC/PI staining and TdT-mediated dUTP nick-end labelling (TUNEL) assays. Silencing or overexpression of HDRACA did not change the proportion of apoptotic cells (Supplementary Fig. S6f, g). These data suggest that HDRACA negatively regulates the endothelial cell cycle, but does not affect endothelial cell apoptosis.

nHDL and dHDL differently regulate HDRACA expression to affect endothelial cell proliferation and tube formation. a Volcano plot shows differentially expressed genes in HUVECs after treatment with lncRNA Smart Silencer HDRACA (Silencer HDRACA) compared with lncRNA smart silencer negative control (Silencer Ctrl), n = 3. b, c Flow cytometric analysis of HUVECs showing the cell cycle at 30 min after addition of PI/RNAse and 48 h after lncRNA Smart Silencer transfection (b) or 72 h after lentivirus vector transfection (c). The representative plots (left and middle) and quantification (right) are shown. d, e The representative images (left) and quantification (right) of EdU incorporation assay in HUVECs treated with VEGF or nHDL or dHDL after silencing (d) or overexpressing (e) HDRACA. The proliferative HUVECs were labeled with EdU (red) and the nuclei were stained with Hoechst 33342 (blue). Scale bars, 100 μm. f, g The representative images (left) and quantification (right) of tube formation assay in HUVECs treated with VEGF or nHDL or dHDL after silencing (f) or overexpressing (g) HDRACA. Scale bars, 500 μm. Data are presented as the mean ± SD. For (b–g), n = 6. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

We performed Cell Counting Kit 8 assays and 5-Ethynyl-2ʹ-deoxyuridine (EDU) incorporation assays to verify the potential effect of HDRACA on endothelial cell proliferation. nHDL and r-ApoM-S1P promoted HUVECs proliferation; however, dHDL were less effective than nHDL (Fig. 3d, e; Supplementary Fig. S7a–d). Silencing of HDRACA promoted HUVECs proliferation (Fig. 3d; Supplementary Fig. S7a). Overexpression of HDRACA not only significantly inhibited HUVECs proliferation but also attenuated the promoting effect of nHDL and r-ApoM-S1P (Fig. 3e; Supplementary Fig. S7b–d). These data support the notion of an important role in regulating endothelial cell proliferation. nHDL promoted HUVEC migration; however, dHDL were less effective (Supplementary Fig. S7e, f). Unlike proliferation, the silencing or overexpression of HDRACA did not affect HUVEC migration (Supplementary Fig. S7e, f). Furthermore, we determined the effect of HDRACA on tube formation regulated by nHDL, dHDL, and r-ApoM-S1P, which was consistent with our findings regarding cell proliferation (Fig. 3f, g; Supplementary Fig. S7g). Moreover, knocking down KLF5 or overexpressing KLF5-ΔPY2 showed similar effects to those of HDRACA knockdown or overexpression on HUVECs proliferation and tube formation (Supplementary Fig. S8a–d). Taken together, our data showed that nHDL and r-ApoM-S1P promoted cell proliferation and tube formation by downregulating HDRACA in endothelial cells, whereas dHDL had less of an effect on decreasing HDRACA levels, resulting in less effective promotion of endothelial proliferation and tube formation.

HDRACA hinders the interaction between RAIN and vigilin in endothelial cells

lncRNAs are known to regulate the expression of neighboring protein-coding genes. Many protein-coding genes are located in the region [−200 to +200 bp] around MRFAP1P1. However, the transfection of HUVECs with HDRACA-siRNAs did not affect the levels of these genes (Supplementary Fig. S9; Supplementary Fig. S15h). These data suggest that HDRACA does not affect the transcription of nearby protein-coding genes in cis.

Next, we examined the trans role of HDRACA. Previous studies have reported that the regulatory roles of lncRNAs are closely related to its subcellular localization.⁴⁴ HDRACA is located in the cytoplasm of HUVECs; thus, it may interact with cytoplasmic RNAs and proteins to perform its biological functions. As the abundance of HDRACA in HUVECs was not high enough to sponge microRNAs, we did not consider it to be a competitive endogenous RNA (ceRNA).⁴⁵ RNA pull-down assays were performed using biotinylated HDRACA, followed by mass spectrometry. Among the proteins specifically associated with HDRACA, RAIN was one of the most enriched proteins (Fig. 4a; Supplementary Table S3). Previous studies have reported that RAIN plays an essential role in endothelial cell morphogenesis and blood vessel tubulogenesis.^46,47 The interaction between RAIN and HDRACA was confirmed using RNA pull-down and RNA immunoprecipitation (RIP) assays (Fig. 4b, c). In addition, HDRACA FISH and RAIN immunofluorescence demonstrated the co-localization of HDRACA with RAIN in the cytoplasm of HUVECs (Fig. 4d). EdU incorporation assay results verified that the knockdown of RAIN inhibited nHDL-induced HUVECs proliferation (Supplementary Fig. S10a). To further determine the segment of HDRACA that interacts with RAIN, we constructed a series of truncated versions of HDRACA according to the secondary structure and found that nucleotides 304–358 of HDRACA form a stem-loop structure and are the only motifs that bind to RAIN (Fig. 4e, f; Supplementary Fig. S10b). Deleting nucleotides 304–358 of HDRACA abrogated the interaction between HDRACA and RAIN (Fig. 4g), suggesting that this segment participates in binding to RAIN.

HDRACA hinders the interaction between RAIN and vigilin in endothelial cells. a Silver staining of proteins bound to HDRACA or antisense for HDRACA in HUVECs after RNA pull-down. RAIN (arrow) was identified by mass spectrometry. AS-HDRACA means antisense for HDRACA. b Immunoblotting for RAIN on protein lysate from HUVECs after RNA pull-down. c RIP assays demonstrated the interaction of HDRACA and RAIN in HUVECs. ACTB was used as a negative control. d Confocal images showing the colocalization of HDRACA (green) and RAIN (red) in HUVECs. The nuclei were stained with DAPI (blue). Scale bars, 10 μm. e Predicted secondary structure of HDRACA and HDRACA^304–358 by Mfold software. f RNA pull-down assays determined the interaction between HDRACA domain constructs and RAIN. g RNA pull-down analysis of biotinylated HDRACA nucleotides 304–358 deletion construct and RAIN. h, i Immunoprecipitation analysis of the interaction of RAIN and vigilin in HUVECs treated with nHDL or dHDL after silencing (h) or overexpressing (i) HDRACA. Representative plots (up) and quantitation (down) are shown. Data are presented as the mean ± SD. For (b–i), n = 6. ***p < 0.001; ****p < 0.0001

However, the silencing or overexpression of HDRACA did not affect RAIN protein levels (Supplementary Fig. S10c, d). It has been reported that lncRNAs affect the interactions between proteins.^48,49 After silencing HDRACA, we performed co-immunoprecipitation (Co-IP) with RAIN, followed by mass spectrometry (Supplementary Fig. S10e; Supplementary Tables S4 and S5). Among significantly different proteins (log₂(normalized ratio Silencer HDRACA/Silencer Ctrl) > log₂(1.5) or <−log₂(1.5) and unique peptides ≥2), vigilin and keratin type I cytoskeletal 9 (KRT9) were the top two enriched proteins (ranked based on MS2 spectral counts). KRT9 is involved in keratin filament assembly; however, little is known about its other functions. Vigilin, also termed HDL-binding protein (HDLBP), was upregulated after silencing HDRACA (Supplementary Table S4). Previous studies have indicated that vigilin can regulate post-transcriptional gene expression with RNA-binding potential.^50,51 It is likely that vigilin is involved in the regulation of cell cycle-related gene expression. Co-IP assays confirmed that vigilin was bound to RAIN (Supplementary Fig. S10f). nHDL enhanced the interaction between RAIN and vigilin; however, dHDL was less effective than nHDL (Fig. 4h, i). Silencing HDRACA promoted RAIN binding to vigilin (Fig. 4h), whereas overexpression of HDRACA significantly hindered the interaction between RAIN and vigilin and attenuated the ability of nHDL to enhance the interaction between RAIN and vigilin (Fig. 4i). Moreover, knocking down KLF5 or overexpressing KLF5-ΔPY2 showed similar effects as HDRACA knockdown or overexpression on the interaction between RAIN and vigilin (Supplementary Fig. S10g, h).

To investigate the mechanism by which HDRACA disrupts the interaction between RAIN and vigilin, we identified the binding sites between RAIN with HDRACA and vigilin. RAIN contains an N-terminal Ras-associating (RA) domain and a C-terminal diluting (DIL) domain. We transduced HUVECs with lentiviruses encoding FLAG-tagged full-length or truncated RAIN (F1:1-259; F2:260-599; F3:600-963) (Supplementary Fig. S11a). RIP and Co-IP assays showed that both HDRACA and vigilin bound to the F3 segment of RAIN (Supplementary Fig. S11b, c), which contains the DIL domain that binds to cargo, such as protein complexes and RNA.⁵² Therefore, to evaluate whether HDRACA and vigilin bind to the DIL domain of RAIN, we transduced HUVECs with lentiviruses encoding the FLAG-tagged DIL domain of RAIN. RIP assays verified that HDRACA bound to the DIL domain of RAIN (Supplementary Fig. S11d), and co-IP assays indicated that vigilin also bound to the DIL domain of RAIN (Supplementary Fig. S11e). Moreover, silencing of HDRACA increased the interaction between the DIL domain of RAIN and vigilin, whereas overexpression of HDRACA hindered their interaction (Supplementary Fig. S11f, g). This finding indicated that HDRACA bound to the DIL domain of RAIN to interrupt the interaction between the DIL domains of RAIN and vigilin.

HDRACA increases the inhibitory effect of vigilin on PCNA expression in endothelial cells

Studies have revealed that vigilin has 14 or 15 RNA-binding hnRNP K homologous (KH) domains and functions in chromosome maintenance, mRNA stability, regulation of mRNA location, and regulation of protein translation.^50,53 Accordingly, we speculated that the HDRACA-RAIN-vigilin interaction regulated the expression of genes involved in the cell cycle, according to the KEGG analysis (Fig. 5a). To further explore the effect of the HDRACA-RAIN-vigilin interaction on these genes, we performed a series of RT-qPCR assays. HDRACA knockdown elevated the mRNA levels of the cell cycle-related genes CCNA2, CCNB1, CCNB2, CDC6, CDC25A, TFDP1, MYC, YWHAH, MAD2L1, PCNA, and PKMYT1 (Fig. 5b). We selected the genes whose expression increased by more than 50% after HDRACA knockdown for further studies. Among the upregulated genes, RAIN knockdown downregulated the mRNA levels of CCNA2, CCNB1, CDC6, TFDP1, YWHAH, PCNA, and PKMYT1 (Fig. 5c; Supplementary Fig. S15i, s), whereas knockdown of vigilin upregulated the mRNA levels of CDC6, PCNA, TFDP1, PKMYT1, and YWHAH (Fig. 5d; Supplementary Fig. S15j, t).

HDRACA increases the inhibitory effect of vigilin on PCNA expression in endothelial cells. a Heatmap shows the differentially expressed cell cycle genes after transfection with Silencer Ctrl or Silencer HDRACA. b Cell cycle genes expression in HUVECs after transfection with negative control siRNA or HDRACA-siRNAs were determined by RT-qPCR. c RT-qPCR determined the expression of cell cycle genes upregulated>1.5 Fold in panel (b) after transfection with negative control siRNA or RAIN-siRNAs. d The expression of cell cycle genes differentially expressed in panel (c) after transfection with negative control siRNA or Vigilin-siRNAs was determined by RT-qPCR. Data in panel (b–d) were normalized to the negative control siRNA group (siCtrl), and the dotted lines represent the value of the siCtrl. e The binding of vigilin protein with PCNA, TFDP1, YWHAH, PKMYT1, or CDC6 mRNAs in HUVECs was determined by RIP assays. f, g The representative plots (left) and quantitation (right) of immunoblot analysis of PCNA in HUVECs treated with nHDL or dHDL after silencing (f) or overexpressing (g) HDRACA. h The representative images (left) and quantification (right) of EdU incorporation assay in HUVECs treated with VEGF or nHDL or dHDL after transfection with negative control siRNA or PCNA-siRNA (siPCNA-2). The proliferative HUVECs were labeled with EdU (red) and the nuclei were stained with Hoechst 33342 (blue). Scale bars, 100 μm. i The representative images (left) and quantification (right) of tube formation assay in HUVECs treated with VEGF or nHDL or dHDL after transfection with negative control siRNA or PCNA-siRNA (siPCNA-2). Scale bars, 500 μm. Data are presented as the mean ± SD. For (b–i), n = 6. For (b–d), *p < 0.05 compared to siCtrl. For (e–i), **p < 0.01; ***p < 0.001; ****p < 0.0001

To date, the most reported functional mechanism of vigilin involves binding to mRNAs and regulating mRNA stability and translation. RIP assays confirmed that the vigilin protein bound to PCNA mRNA but not to HDRACA in HUVECs (Fig. 5e; Supplementary Fig. S12a). Knockdown of KLF5 inhibited the interaction between vigilin and PCNA mRNA in HUVECs, whereas overexpression of HDRACA promoted their interaction (Supplementary Fig. S12b, c). We overexpressed RAIN by transducing HUVECs with RAIN-overexpressing lentivirus (Supplementary Fig. S15w). Overexpression of RAIN in HUVECs inhibited the interaction between vigilin and PCNA mRNA (Supplementary Fig. S12d).

We evaluated the regulatory effects of vigilin on PCNA mRNA stability and translation. We initially transfected HUVECs with negative control siRNA or viglin siRNAs. We then treated the cells with the transcription inhibitor α-amanitin. As shown in Supplementary Fig. S12e, PCNA mRNA levels were gradually decreased by inhibiting transcription; however, knockdown of vigilin effectively decelerated PCNA mRNA decay. Next, we treated the transfected cells with the translation inhibitor cycloheximide and sedimented RNAs at a high velocity through a sucrose gradient (Supplementary Fig. S12f). The majority of PCNA mRNA was associated with polysomes, and the knockdown of vigilin did not affect the relative distribution (Supplementary Fig. S12g). These data suggest that vigilin regulates the expression of PCNA mRNA by decreasing its stability instead of affecting its translation.

nHDL increased the expression of PCNA protein, but dHDL were less effective than nHDL (Fig. 5f, g). Silencing HDRACA effectively enhanced the expression of PCNA protein (Fig. 5f), whereas overexpression of HDRACA downregulated PCNA protein expression and attenuated the ability of nHDL to increase PCNA protein expression (Fig. 5g). We further performed EdU incorporation and tube formation assays to determine the role of PCNA in nHDL- and dHDL-regulated endothelial cell proliferation and tube formation. PCNA knockdown inhibited HUVEC proliferation and tube formation and attenuated the promoting effect of nHDL (Fig. 5h, i; Supplementary Fig. S15k, u).

These data suggest that nHDL enhances the interaction between RAIN and vigilin, contributing to the promotion of PCNA expression and PCNA-mediated angiogenesis, but dHDL is less effective in enhancing the interaction between RAIN and vigilin, which resultes in less effective promotion of PCNA expression and PCNA-mediated angiogenesis.

Ectopic expression of human HDRACA in mouse endothelial cells inhibits angiogenesis

HDRACA is poorly conserved across species. We could not find a clear orthologous counterpart in the mouse genome via BLAST (Basic Local Alignment Search Tool). Therefore, we ectopically expressed human HDRACA in mouse femoral artery endothelial cells (mFAECs) to verify its regulatory effect on angiogenesis. We successfully expressed human HDRACA in mFAECs by using HDRACA-overexpressing lentiviruses (Fig. 6a). HDRACA overexpression inhibited mFAEC proliferation and tube formation (Fig. 6b, c). Because the protein sequence of RAIN is highly conserved and identical between humans and mice, we examined whether human HDRACA could bind to RAIN in mFAECs. RNA pull-down and RIP assays confirmed that human HDRACA could also bind to mouse RAIN in mFAECs (Fig. 6d, e). These results indicated that HDRACA inhibited endothelial angiogenesis in humans and mice.

Ectopic expression of human HDRACA in mouse endothelial cells inhibits angiogenesis. a RT-PCR indicated the presence of human HDRACA in mFAECs transfected with lentiviruses. b The representative images (up) and quantification (down) of EdU incorporation assays in mFAECs treated with or without VEGF after overexpressing HDRACA. The proliferative mFAECs were labeled with EdU (red) and the nuclei were stained with Hoechst 33342 (blue). Scale bars, 100 μm. c The representative images (up) and quantification (down) of tube formation assays in mFAECs treated with or without VEGF after overexpressing HDRACA. Scale bars, 500 μm. d Immunoblotting for RAIN on protein lysate from mFAECs after RNA pull-down with biotin-labeled HDRACA or HDRACA antisense (AS-HDRACA). e RIP assays demonstrated the interaction between HDRACA and RAIN in mFAECs transfected with HDRACA-overexpressing lentiviruses. Gapdh was used as a negative control. Data are presented as the mean ± SD. n = 6. ****p < 0.0001

HDRACA inhibits angiogenesis in vivo

To verify the regulatory effects of HDRACA in vivo, blood flow in the left hindlimbs of C57BL/6 mice was surgically reduced and we ectopically expressed human HDRACA in these ischemic hindlimbs (Fig. 7a, b). The hindlimbs of the C57BL/6 mice successfully expressed human HDRACA (Supplementary Fig. S13a–c). Laser Doppler perfusion imaging showed that transduction of mice with adenoviral vectors carrying HDRACA (AdV-HDRACA) inhibited blood flow recovery compared with the control vector (AdV-Ctrl) or normal saline (NS) groups (Fig. 7c). Micro-computed tomography (micro-CT) analysis showed that the ectopic expression of HDRACA significantly decreased collateral vessel density in ischemic hindlimbs (Fig. 7d). Immunofluorescence staining of CD31 showed that the ectopic expression of HDRACA decreased capillary density (Fig. 7e). Ectopically expressing HDRACA downregulated PCNA levels in endothelial cells in ischemic hindlimbs without affecting RAIN and vigilin levels (Fig. 7f; Supplementary Fig. S13d–e).

HDRACA inhibits angiogenesis in vivo. a Schematic graphic of the ischemia model used in this study. b Overview of experimental process in vivo. c Representative images (up) and quantification (down) of laser doppler flow before and at various time points after femoral artery ligation in mice treated with normal saline (NS), control adenovirus vector (AdV-Ctrl), or adenovirus vector carrying HDRACA (AdV-HDRACA). R (right) and L (left) foot. d Representative images (up) and quantification (down) of micro-CT analysis at day 14 after femoral artery ligation in mice treated with NS or AdV-Ctrl or AdV-HDRACA. e Representative images (left) and quantification (right) of CD31 staining (green) at day 14 after femoral artery ligation in mice treated with NS, AdV-Ctrl or AdV-HDRACA. The nuclei were stained with DAPI (blue). Scale bars, 100 μm. f Representative images (left) and quantification (right) of CD31 (green) and PCNA (magenta) co-staining at day 14 after femoral artery ligation in mice treated with NS, AdV-Ctrl or AdV-HDRACA. The nuclei were stained with DAPI (blue). Scale bars, 50 μm. Data are presented as the mean ± SD. For (c–f), n = 7. ***p < 0.001; ****p < 0.0001; ns not significant

We further confirmed the functional contribution of HDRACA in the regulation of angiogenesis using a Matrigel plug assay. Matrigel mixed with normal saline (NS), control adenovirus vector (AdV-Ctrl), or Adenovirus vector carrying HDRACA (AdV-HDRACA) was subcutaneously injected into the abdomen of mice and removed on day 7 for histological examination (Supplementary Fig. S14a). Matrigel plugs that incorporated AdV-HDRACA showed reduced neovascularization compared with AdV-Ctrl plugs, which was further confirmed by hematoxylin-eosin (HE) staining and CD31 immunohistochemistry (IHC) (Supplementary Fig. S14b–d).

HDRACA is upregulated in lower extremity artery intima of patients with arteriosclerosis obliterans

Arteriosclerosis obliterans (ASO) and CAD are manifestations of atherosclerosis that cause arterial stenosis in different parts of the body with similar risk factors and metabolic disorders. To evaluate the potential translational relevance of HDRACA in human tissues, we marked endothelial cells in the lower extremity artery using CD31 immunofluorescence assays, followed by co-staining with KLF5, HDRACA, or PCNA. The endothelial cells of the lower limb arteries in patients with ASO showed upregulation of KLF5 and HDRACA, and downregulation of PCNA (Fig. 8a, b). Furthermore, the S1P content in HDL was inversely correlated with HDRACA levels in the endothelial cells of the lower limb arteries (Fig. 8c). This finding preliminarily confirmed the change in S1P-KLF5-HDRACA-PCNA signaling in ASO.

HDRACA is upregulated in lower extremity artery intima of patients with ASO. a Representative images of co-stain for CD31 (green) with KLF5 (red; up), HDRACA (red; middle) and PCNA (red; down) in artery tissues from donors for organ transplantation (Ctrl; n = 10) and patients with ASO (n = 10). The nuclei were stained with DAPI (blue). Scale bars, 50 μm. b Quantification of KLF5, HDRACA, or PCNA mean fluorescence intensity in CD31 positive cells of artery tissues from Ctrl and ASO groups. c Correlation between S1P content in HDL and HDRACA mean fluorescence intensity in CD31 positive cells of artery tissues. n = 20. d Graphical illustration of HDL-HDRACA regulatory mechanism. (1) nHDL-bound S1P interacts with S1P1, leading to the phosphorylation and activation of WWP2. The activated WWP2 catalyzes the ubiquitination and subsequent degradation of KLF5, resulting in reduced KLF5-induced HDRACA transcription. Consequently, non-HDRACA-bound RAIN increases and interacts with vigilin to inhibit its binding to PCNA mRNA, leading to increased stability of PCNA mRNA, increased PCNA expression, and increased PCNA-induced endothelial cells proliferation and tube formation, ultimately promoting angiogenesis. (2) dHDL does not inhibit KLF5-induced HDRACA transcription via S1P1-WWP2 signaling pathway, due to the loss of S1P. HDRACA binds to RAIN, which disrupts the interaction between RAIN and vigilin, resulting in the binding of vigilin and PCNA mRNA, thereby decreasing the stability of PCNA mRNA and PCNA expression. As a result, dHDL fails to promote angiogenesis by increasing the expression of PCNA and PCNA-induced proliferation and tube formation. Data are presented as the mean ± SD. ***p < 0.001; ****p < 0.0001; ns not significant