Circular RNA Profiling Identifies circ5078 as a BMPR2-Derived Regulator of Endothelial Proliferation and Stress Responses

Abstract

Background: The BMPR2 gene encodes the BMPR-II (bone morphogenetic protein receptor type-II) and is a known regulator of endothelial proliferation, apoptosis, and translational stress responses. While these effects are generally attributed to the actions of BMPR-II protein, we used circular RNA profiling to identify circ3218 and circ5078 as new BMPR2-derived functional RNAs.

Methods: Circular RNAs were profiled by ultradeep RNA sequencing of human pulmonary artery endothelial cells. Novel BMPR2-derived circular RNAs were assessed for their effects on endothelial proliferation, apoptosis, and translational stress responses in human pulmonary artery endothelial cells and endothelial cells from patients with pulmonary arterial hypertension with heterozygous loss-of-function BMPR2 mutations. circ5078 to linear BMPR2 mRNA ratios were quantified in cultured lymphocytes from patients with pulmonary arterial hypertension with BMPR2 mutations versus unaffected mutation carriers.

Results: Depletion of circ3218 enhanced human pulmonary artery endothelial cell apoptosis, whereas circ5078 silencing increased proliferation. Enhanced proliferation with circ5078 depletion was eliminated by cosilencing of circ5078 with either linear BMPR2 mRNA or the stress granule protein, Caprin-1, indicating a potential interdependence of BMPR2 transcripts in the regulation of endothelial function. Patients with pulmonary arterial hypertension with BMPR2 mutations exhibited increased circ5078 to linear BMPR2 mRNA ratios, alongside impaired stress responses in patient endothelial cells that were deficient in linear BMPR2 transcripts but not circ5078. circ5078 depletion enhanced stress responses in human pulmonary artery endothelial cells and rescued stress granule formation in patient endothelial cells, independent of BMPR-II protein levels. Assessment of translational regulation by polysome profiling did not identify any impact of linear or circular BMPR2 transcript loss on global protein synthesis or stress-induced eIF2α (eukaryotic initiation factor 2α) phosphorylation but did identify the enhanced translational efficiency of select nuclear-encoded mitochondrial ribosome proteins with circ5078 depletion, offering a link between mitochondrial function and the circ5078-deficient endothelial phenotype.

Conclusions: The identification of circ3218 and circ5078 as novel BMPR2-derived gene products reveals interdependent roles for coding and noncoding BMPR2 transcripts as regulators of endothelial function.

Keywords: RNA, circular; endothelial cells; phosphorylation; pulmonary arterial hypertension; stress granules.

Figure 1.

Screening of an ultradeep RNA-sequencing data set to profile human pulmonary artery endothelial cell (HPAEC) circular RNA (circRNA) expression. A, Schematic of the bioinformatic analysis method used to identify circRNA-derived sequencing reads. B, Graphical representation and (C) summary of the number of circRNA annotations and reads per sample at each stage of analysis. D, circRNA annotations ranked by average transcripts per kilobase million (TPM) across 12 samples. An average TPM ≥0.1 was selected as the cutoff of abundantly expressed circRNA annotations, as indicated by the dashed line. E, Volcano plot demonstrating the impact of BMPR2 silencing on the expression of 425 abundantly expressed circRNAs (red circles in B). Annotations with P_adj<0.05 are summarized. F, RNase R validation of differentially expressed circRNA annotations alongside RPL19 as linear RNA control (n=7). F, 1-way ANOVA with the Dunnett post hoc test.

Figure 2.

Identification of novel BMPR2-derived circular RNAs (circRNAs). A, Summary of all linear and circular BMPR2–derived transcripts expressed in human pulmonary artery endothelial cells (HPAECs). Regions targeted by small interfering RNAs (siRNAs) used in this study are indicated. B, Relative expression of linear BMPR2 transcripts and the 8 BMPR2-derived circRNA annotations that were detected from the RNA-sequencing screen in siControl-treated HPAECs (n=3). The dashed line indicates the expression cutoff of 0.1 transcripts per kilobase million (TPM). C, RNase R validation of BMPR2-derived circRNA annotations, alongside linear BMPR2 mRNA and RPL19 as a linear RNA control (n=6). D, Relative expression of circ5078 and (E) circ3218 in HPAECs treated with pooled siRNAs targeting BMPR2 (siBMPR2) or a nontargeting siRNA control pool (siControl; n=5). F, Quantification of human BMPR2 transcript equivalents in RNA-sequencing data from rat lungs (GSE159668; n=6). G, Graphical depiction of the introns surrounding exon 12 of BMPR2, the reverse complementary matches (RCMs) that they contain, and their conservation in rodent and nonhuman primate species. H, Summary of the alignments between the RCM sequences within intron 12 of BMPPR2 and the indicated segment of intron 11. C, One-way ANOVA with the Dunnett post hoc test. D and E, Paired t test.

Figure 3.

Functional characterization of BMPR2-derived circular RNAs (circRNAs). A, Impact of siLinear (n=4), si5078 (n=6), siLinear+5078 (n=5), and si3218 (n=6) on the abundance of generic linear BMPR2 and individual BMPR2 transcripts. The dashed line represents the expression of each transcript in the siControl sample; bars represent the expression of each transcript under the respective knockdown condition. B, Representative immunoblot and (C) quantification of BMPR-II (bone morphogenetic protein receptor type-II) expression under each knockdown condition (n=7). D, Representative flow plots of propidium iodide (PI) and FITC (fluorescein isothiocyanate)-annexin-V–stained human pulmonary artery endothelial cells (HPAECs) and (E) quantification of apoptotic Annexin-V⁺/PI⁻ apoptotic cells under each small interfering RNA (siRNA) condition, with and without a TNFα (tumor necrosis factor α) and cycloheximide apoptotic stimulus (n=7). F, Relative change in cell number over 96 hours for each siRNA treatment relative to siControl from experiments in A. A, Two-way ANOVA with the Bonferroni post hoc test. C, One-way ANOVA with the Dunnett post hoc test. E, Two-way ANOVA with the Dunnett post hoc test. F, Mixed-effects model (REML [restricted maximum likelihood] method, siRNA treatment fixed variable, and repeated measures random variable) with the Dunnett post hoc test.

Figure 4.

Regulation of proliferation by circ5078 is Caprin-1–dependent. A, Potential circ5078 protein binding partners identified by BEAM RNA Interaction Motifs (BRIO). B, RNA immunoprecipitation of linear BMPR2 mRNA and circ5078 with anti-Caprin-1 antibody or IgG control. C, Representative immunoblot demonstrating Caprin-1 protein abundance in cytosolic, cytoskeletal, and membrane-bound fractions following treatment with small interfering RNAs (siRNAs) targeting linear BMPR2 mRNA, circ5078, or both vs nontargeting siControl. D, Relative change in cell number over 96 hours of human pulmonary artery endothelial cells (HPAECs) following silencing of circ5078, CAPRIN1, or both, relative to siControl (n=8). E, Representative immunoblot and (F) quantification of cyclin-D1 protein following treatment with siControl, siLinear, si5078, or siLinear+5078 siRNAs (n=11). G, CCND1 gene expression under each siRNA condition (n=5). D, One-way ANOVA with the Bonferroni post hoc test. F and G, One-way ANOVA with the Dunnett post hoc test.

Figure 5.

Linear and circular BMPR2 transcripts differentially modulate Caprin-1 granule formation. A, Representative confocal (scale bar, 20 µm) and (B) stimulated emission depletion (STED) super-resolution (scale bar, 0.5 µm) images of human pulmonary artery endothelial cells (HPAECs) following 1 hour of treatment with 100-μM sodium arsenite–induced, stained with Caprin-1 (green) and the stress granule marker G3BP-1 (Ras GTPase-activating protein-binding protein 1; red) following siControl, siLinear, si5078, or siLinear+5078 treatment. C, Quantification of stress granules under each knockdown condition (n=9). D, Representative immunoblot and (E) quantification of phosphorylated eIF2α (eukaryotic initiation factor 2α) in HPAECs, relative to total eIF2α (n=3). F, Expression of circ5078 relative to linear BMPR2 RNAs in cultured lymphocytes from BMPR2-mutation positive individuals affected by pulmonary arterial hypertension (PAH; n=45) and unaffected mutation carriers (n=24). G, Relative expression of linear BMPR2, BMPR2a, BMPR2b, and circ5078 transcripts in blood outgrowth endothelial cells (BOECs) from patients with BMPR2 mutation–bearing PAH (n=5) and healthy controls (n=10). H, Quantification of cells with stress granules in BOECs from healthy controls (n=4) and patients with PAH with BMPR2 mutations (n=5) in response to sodium arsenite. I, Stress granule production in control BOECs treated with siControl (n=4) and PAH BOECs treated with siControl or si5078 (n=5). C and I, One-way ANOVA with the Dunnett post hoc test. E, Two-way ANOVA with the Dunnett post hoc test. F through H, Unpaired t test.

Figure 6.

BMPR2 transcripts regulate ribosomal assembly but not global protein synthesis. A, Sucrose gradient UV absorbance (A₂₆₀) profiles for human pulmonary artery endothelial cells (HPAECs) treated with siControl (black), siLinear (red), si5078 (blue), or siLinear+5078 (purple) small interfering RNAs (siRNAs). Insets are magnifications of the curves in fractions 6 to 12. B, Representative immunoblots of RPL7 (ribosomal protein large 7), RPS6 (ribosomal protein small 6), eIF3B (eukaryotic initiation factor 3B), eIF2α (eukaryotic initiation factor 2α), and Caprin-1 across gradient fractions. C, Overlaid A₂₆₀ profiles highlighting the 43/48S preinitiation complex (PIC), 60S ribosomal subunits, 80S monosomes, and polysome halfmers. D, Quantification of absorbance less background for the 43S/48S, (E) 60S, and (F) 80S peaks (n=3 each). G, Relative abundance of circ5078 RNA across all fractions in each siRNA condition. The data shown are the pooled average of 3 independent experiments. H, Quantification of global protein production by incorporation of O-propargyl-puromycin (OPP) into nascent peptides in HPAECs (n=6). D through F, One-way ANOVA with the Dunnett post hoc test. H, Paired t test between untreated and cycloheximide groups. Two-way ANOVA with the Dunnett post hoc test between all siRNA-treated groups.

Figure 7.

Loss of circ5078 in human pulmonary artery endothelial cells (HPAECs) alters mitochondrial function via altered translation of mitochondrial ribosomes. A, RNA sequencing was performed on total input RNA from unfractionated lysates and RNA from the actively translated fractions (8–13) of polysome profiles for HPAECs treated with siControl, siLinear, si5078, or siLinear+5078 small interfering RNAs (siRNAs; n=3 independent experiments). B, Scatter plots of the log₂-fold change vs siControl in both total RNA and polysome RNA fractions for each siRNA condition. Highlighted transcripts are those deemed to be translationally regulated (P_adj≥0.05 in input fraction and P_adj<0.05 in polysome fraction). C, Venn diagram and (D) heatmap of genes regulated by translation in at least 1 of the 3 conditions. E, Summary of pathways enriched among the transcripts in which translation was upregulated by circ5078 loss alone, as identified by Reactome. F, Relative expression of transcripts encoding components of the mitochondrial translation pathway in polysome fractions. G, HPAECs treated with siControl (n=14), siLinear (n=12), si5078 (n=12), or siLinear+5078 (n=11) were analyzed by the Seahorse XFe24 Extracellular Flux Analyzer. The oxygen consumption rate (OCR) was assessed across the course of the experiment following the addition of oligomycin, FCCP, (carbonyl cyanide p-trifluoromethoxyphenylhydrazone) rotenone, and antimycin A. H, Basal respiration, (I) coupling efficiency, and (J) spare capacity under each knockdown condition. F and H through J, One-way ANOVA with the Dunnett post hoc test.