Super Enhancer-Associated Circular RNA-CircKrt4 Regulates Hypoxic Pulmonary Artery Endothelial Cell Dysfunction in Mice

Since female sex hormone, for example, estrogen, can attenuate chronic hypoxia and SuHx-induced PH,^21,22 adult male C57BL/6 mice (6–9 weeks old) weighing 20 to 25 g from the Laboratory Animal Center of the Second Affiliated Hospital of Harbin Medical University were employed in this study. All experimental procedures were performed in accordance with the ethical standards in the 1964 Declaration of Helsinki and its later amendments and approved by the Ethics Committees of Harbin Medical University (HMUDQ20220317001). The corresponding target RNA cloning construction and serotype 5 adenovirus-associated virus (AAV5) were packaged by Genechem (Shanghai, China). Mice were randomly divided into different groups and infected with the AAV5 vector at 10¹¹ genome equivalents in 20 to 30 µL Hank balanced salt solution after isoflurane anesthesia, followed by nasal drops. After 7 to 14 days, the mice were assigned to normoxia (Fi, O₂ 0.21) and hypoxia (Fi, O₂ 0.10) as described in a previous study.^16,23 The normoxic group was kept in the same place adjacent to the hypoxic group, and the concentration of oxygen was monitored continuously using an oxygen analyzer (P110, BioSpherix New York) for 4 weeks. For SuHx-mediated PH model, SU5416 (S8442, Sigma-Aldrich, Darmstadt, Germany) was injected to mice subcutaneously at 20 mg/kg, and control mice were injected with the same volume of vehicle alone during the next 3 weeks in 10% oxygen and followed by reexposure to normoxia 2 weeks. Lung tissues were taken for corresponding experiments after anesthesia, and the right ventricular (RV) hypertrophy index (ratio of RV free wall weight over the sum of septum (S) plus left ventricular (LV) free wall weight), RV/(LV+S) was calculated.

At the end of hypoxic and SuHx treatment, mice were subjected to echocardiography using a Vevo2100 imaging system equipped with a 18- to 38-MHz (MS400, mouse cardiovascular) transducer probe (VisualSonics, Inc, Toronto, Ontario, Canada), as previously described.¹⁶ Mice were anesthetized with inhaled isoflurane, administered at 1% to 3% by an Advanced Table Top Anesthesia machine (HARVARD APPARATUS LIMITED). Mice were secured on a prewarmed (37 °C) imaging platform and heart and respiratory rates were observed continuously throughout the echocardiography procedure. Pulsed-wave Doppler was used to measure pulmonary artery acceleration time and the pulmonary artery velocity time integral. Pulmonary artery acceleration time was measured from the pulsed-wave Doppler flow velocity profile of the RV outflow tract in the parasternal short-axis view and was defined as the interval from the onset to the maximal velocity of forward flow. LV ejection fraction was measured at the parasternal short-axis view, the LV area was traced during end diastole and end systole, and the LV ejection fraction was calculated with: LV ejection fraction (%)=(LVEDV [left ventricular end-diastolic volume]−LVESV [left ventricular end-systolic volume])/LVEDV×100. All measurements were performed by the same experienced investigator. Off-line data analysis was performed using Vevo2100 software by investigator in a blind manner.

The RV systolic pressure was measured with PowerLab monitoring equipment (AD Instruments, Colorado Springs, CO). A 1.2 French Pressure Catheter (Scisense Inc) was inserted into the superior vena cava and finally into the RV vein, and the RV systolic pressure was continuously recorded for 20 to 40 minutes.

Hematoxylin and eosin staining and Masson trichrome staining were performed, as previously described.¹⁶ In brief, lung tissues of mice were immersed in 4% paraformaldehyde for 48 hours. After dehydration, clearing, and embedding in paraffin wax, the blocks were cut into 5-μm–thick sections and stained as appropriate. For immunohistochemistry, the sections were incubated with α-SMA (Boster, Wuhan, China). Then, the tissue sections were stained with 3,3-diaminobenzidine and restrained with hematoxylin. The total wall thickness and positive staining area in the vascular walls were quantified using a color-recognition algorithm in Image-Pro Plus 6.0 software.

At least 4 different batches of PAECs (catalog no. CP-M011) from male mice were provided by Procell Life Science & Technology (Wuhan, China). Cells were maintained in endothelial cell medium (ScienCell, 1001, CA) containing 15% FBS and 1% penicillin streptomycin at 37 °C, 5% CO₂, and 100% relative humidity. A tri-gas incubator (Thermo Fisher, MA) with an atmosphere comprising 3% O₂, 5% CO₂, and 92% N₂ was used for hypoxic culture. All experiments were performed with cells less than passages 4. For the Cell Culture and Cell Counting Kit 8 assay, PAECs were seeded in 96-well plates and treated according to the different experimental groups for 24 to 48 hours. After adding 10 µL, Cell Culture and Cell Counting Kit 8 (Biosharp, Hefei, China) reagent to each well for 2 hours, the results were detected at 450 nm wavelength using a spectrophotometer.

PAECs were treated with 50 μmol/L 5-ethynyl-2’-deoxyuridine (C0071S, Beyotime, Shanghai, China) and incubated for 4 hours at 37 °C. The cells were fixed with 4% formaldehyde for 20 minutes and exposed to 0.5% Triton X-100 for 15 minutes. 4′,6-diamidino-2-phenylindole (5 μg/mL) was used to stain the DNA of cells in each well for 30 minutes. Then, the images were observed under a fluorescence microscope, and the results are shown as the percentage of proliferating cells (ethynyl-2’-deoxyuridine–positive cells) with 4′,6-diamidino-2-phenylindole. To assess the cell cycle distribution in the different experimental groups, a Cycle Test Plus DNA Reagent Kit (C1052, Beyotime, Shanghai, China) was used. Cells were fixed with 70% ethanol and exposed to 500 µL of propidium iodide. A BD FACSCalibur flow cytometer was used to measure DNA fluorescence.

Cell transfection was performed according to the manufacturer’s instructions for the Lipofectamine 2000 Reagent (11668019; Thermo Fisher, MA), and the procedure for transfection was described previously.¹⁶ Small interfering RNA (siRNA) against Pura, CEBPA, RBM25, and nontargeted control siRNA were designed and synthesized by GenePharma (Suzhou, China). The sequences are shown in Table S1. Glpk overexpression and mutant plasmids were constructed using the GV658 vector by GeneChem (Shanghai, China), and the vector alone was used as a negative control (NC).

Total RNA samples were extracted from cultured PAECs and lung tissues using TRIzol reagent (Thermo Fisher, MA) according to the manufacturer’s protocol. RNA was extracted from blood/plasma using a Blood/Plasma RNA Extraction Kit (HaiGene, Harbin, China) according to the manufacturer’s protocol. Cytoplasmic and nuclear RNAs were isolated and purified using Norgen’s Cytoplasmic & Nuclear RNA Purification Kit (Thorold, ON, Canada). According to the manufacturer’s instructions, PAECs were lysed with ice-cold lysis buffer, cytoplasmic and nuclear RNA were bound to the column, and the mixture was separated by RNA elution. For each sample, 500 ng of total RNA was converted to cDNA using a Superscript First-Strand cDNA Synthesis Kit (HaiGene, Harbin, China). Real-time polymerase chain reaction (PCR) was performed in a LightCycler 480 II real-time PCR system (Roche, Basel, Switzerland) with SYBR Green (TOYOBO, Osaka, Japan). The threshold cycle (Ct) was determined, and the data were analyzed using the 2^-ΔΔCT method. The primers (5′-3′) are shown in Table S2.

Protein samples (20–50 μg) extracted from lung tissues and PAECs were separated by SDS–PAGE (8%–12%) and transferred onto nitrocellulose membranes. The membranes were blocked in 5% nonfat dry milk for 1 hour and then incubated with the indicated antibodies (see Major Resources Table in the Supplemental Material for information about the antibodies) overnight at 4 °C. After washing with TBST buffer, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies and subjected to chemiluminescence reagent imaging.

For the transwell assay, resuspended PAECs were added to the upper chamber of a modified Boyden chamber (Costar, Corning, NY) with a polycarbonic membrane (6.5 mm in diameter and 8 µm pore size). The migrated cells were fixed with 4% paraformaldehyde and stained with 0.4% crystal violet, and the number of stained migrated cells was counted under an inverted microscope (Nikon, Japan). For the network formation assay, 96-well culture plates were coated with 30 μL growth factor-reduced Matrigel (BD Biosciences, NY) for 30 minutes at 37 °C. PAECs were trypsinized and resuspended, and 200 μL of cell suspension was added to solidified Matrigel. Network formation of PAECs was observed under an inverted microscope and measured using Image-Pro Plus 6.0.

PAECs were fixed with 4% paraformaldehyde for 15 minutes and permeabilized with 0.01% Triton X-100. After blocking with 5% BSA, the cells were incubated with antibodies against N-cadherin (1:100), Pura (1:100), Glpk (1:100), and anti-DNA/RNA Damage (8-hydroxy-2 deoxyguanosine, 1:100) at 4 °C. The next day, the cells were incubated with CY3 (cyanine-3)/FITC (fluorescein isothiocyanate)–conjugated secondary antibody (1:100) at 37 °C for 2 hours and 4′,6-diamidino-2-phenylindole away from light for 10 minutes. The results of immunofluorescence staining were recorded by a live cell workstation (AF6000; Leica, Wetzlar, Germany). Fluorescent in situ hybridization (FISH) experiments for PAECs were performed using a Fluorescent In Situ Kit following the manufacturer’s instructions, and Cy3-conjugated circKrt4 probes were synthesized by RiboBio (Guangzhou, China). After treatment with agents according to the different experimental groups, PAECs were washed, fixed with 4% paraformaldehyde, and permeabilized with 0.3% Triton X-100. After the cells were blocked with prehybridization solution, they were incubated with hybridization solution containing circKrt4, 18S, and U6 probes followed by 4′,6-diamidino-2-phenylindole incubation. FISH experiments for tissues samples were performed using a circRNA FISH Probe Mix following the manufacturer’s instructions synthesized by EXONBIO (Guangzhou, China). Briefly, 5-μm-thick lung tissue sections were pretreated in PBS and incubated with digoxigenin-labeled denature circKrt4 probes in hybridization buffer at 37 °C overnight. Following incubation with CY3-labeled anti-digoxigenin antibodies in blocking buffer and 4′,6-diamidino-2-phenylindole, results of FISH staining were recorded by a live cell workstation (AF6000; Leica, Wetzlar, Germany).

For the protein coimmunoprecipitation assay, lysed PAECs were collected and centrifuged at 15 000 revolutions per minute for 30 minutes at 4 °C to collect the supernatant, followed by the addition of 5 μg of target antibody or IgG and incubation at 4 °C for 6 hours. Protein A+G agarose beads were added overnight in a 4 °C shaker. The antibody-protein complexes were washed, and the pellet was resuspended in protein loading buffer (2×) followed by Western blotting. The RNA immunoprecipitation (RIP) assay was performed according to the manufacturer’s instructions (Bes5101; BersinBioTM, Guangzhou, China). Briefly, 1×10⁷ PAECs were collected and lysed with RIP buffer. Then, the samples were incubated with protein A/G bead–conjugated anti-Pura (5 μg, GTX00901; GeneTex, CA), anti-Glpk (4 μg, 13360-1-AP; Proteintech, IL) and anti-RBM25 (1.5 μg, 25297-1-AP, Proteintech, IL) antibodies overnight at 4 °C. After extracting the RNA from immunoprecipitation and input samples, 1000 ng of the total RNA was converted to cDNA and subjected to quantitative real-time polymerase chain reaction (qRT-PCR) assays. The primers (5′–3′) are shown in Table S2.

The oxygen consumption rate (OCR, a hallmark of mitochondrial oxidative phosphorylation) and extracellular acidification rate (a hallmark of glycolysis) were measured using a 24-well extracellular flux analyzer (Agilent Technologies Co, Ltd) following the specific manufacturer’s protocol of the Seahorse XF Cell Mito Stress Test Kit and Seahorse XF Glycolysis Stress Test Kit (103015-100 and 103020-100; Agilent, CA). For OCR analysis, PAECs plated in an XFp cell culture miniplate (102340-100; Agilent, CA) and ATP synthase inhibitor oligomycin (trifluoromethoxy carbonylcyanide phenylhydrazone; 1 μM), which decreases OCR levels, followed by the electron transport chain accelerator FCCP (1 μM), which causes maximal respiration, and finally rotenone+antimycin A (both 1 μM), which are mitochondrial complex I and III inhibitors, were sequentially injected into cells. The results were calculated from the OCR values in picomoles per minute of oxygen consumed and were normalized with total protein concentration. For the glycolysis assay, cells were glucose starved in XF assay medium in a CO₂-free XF prep station and then treated with glucose (2 mg/mL), oligomycin (1 μM), and 2-deoxy-D-glucose (100 mmol/L). All measurements were normalized with total protein concentration on each well and differences in glycolytic capacity were represented as mpH/min per protein.

A dual-luciferase reporter assay was performed according to the protocol of the Dual-Lumi Luciferase Reporter Gene Assay Kit (RG088S, Beyotime, Shanghai, China). The Krt4 promoter fragment containing the CEBPA binding site was cloned into the GV238 plasmid expressing luciferase (GenePharma, Suzhou, China). PAECs were cotransfected with the CEBPA expression plasmid and Krt4 plasmid expressing luciferase with Lipofectamine 2000 for 48 hours. Then, the luciferase activities were measured by the dual-luciferase reporter assay system (GloMax Multi Detection System, Promega, Madison).

For intracellular reactive oxygen species (ROS) assay, fresh media containing 10 μM 2′,7′-dichlorodihydroflfluorescein diacetate probes (S0033S, Beyotime, Shanghai, China) was treated for cells in a 37 °C humidified incubator. Green fluorescence was captured (488 nm excitation/525 nm emission) from ≥3 optical fields. Mitochondrial ROS activity was measured with a redox-sensitive fluorescent probe, mitochondrially targeted superoxide indicator (MitoSOX Red; M36008; Thermo Fisher, MA). PAECs were incubated with 5 µmol/L MitoSOX probe for 20 minutes. The cells were washed with PBS, and red fluorescence was captured (514 nm excitation/585 nm emission) from ≥3 optical fields. Total cellular ROS and mitochondrial ROS in the fluorescence intensity of treated cells was quantified by Image J.

SOD (superoxide dismutase) activity was measured using a colorimetric CuZn/Mn-SOD activity assay kit (S0103, Beyotime, Shanghai, China). Briefly, 20 µL of the sample was mixed with WST working solution and enzyme working solution, and activities of SOD were calculated by measuring the absorbance at 450 nm following incubation at 33 °C for 20 minutes. GPx (glutathione peroxidase) activity was measured using a GPx activity assay kit (S0058, Beyotime, Shanghai, China). The collected cells were washed and lysed with 200 µL cold Assay Buffer. Following centrifugation 15 minutes at 4 °C at 10 000×g to remove any insoluble material, 10 µL sample was added to the mixture reaction system (40 mmol/L NADPH solution, glutathione reductase, and glutathione). The concentration of NADPH was measured at 340 nm at 0 and 5 minutes after cumene hydroperoxide addition, and GPx activities were calculated from the changes in NADPH concentrations.

The specific biotinylated probe containing the Krt4 promoter (−2000/−1580) fragment was synthesized by GenePharma (Suzhou, China). The Electrophoretic Mobility Shift Assay was performed following the instructions of a chemiluminescent Electrophoretic Mobility Shift Assay Kit (GS009, Beyotime, Shanghai, China). The probe and the extracted nuclear protein were reacted in binding reaction buffer for 20 minutes. Then, the sample was added to a 4% nondenatured polyacrylamide gel for electrophoresis purposes and transferred onto a nylon membrane, followed by UV cross-linking at 245 nm. Finally, the membrane was incubated with streptavidin-horseradish peroxidase conjugate, followed by enhanced chemiluminescent reagent imaging.

A chromatin immunoprecipitation (ChIP) assay was performed using a ChIP assay kit (P2078, Beyotime, Shanghai, China). Briefly, PAECs were cross-linked with a final concentration of 1% formaldehyde and quenched by a final concentration of 125 mmol/L glycine. After the cells were lysed with SDS supplemented with PMSF (phenylmethanesulfonyl fluoride), they were lysed by ultrasound to generate chromatin fragments of 100 to 1000 base pair. Antibodies against acetyl H3K27 (H3K27ac [acetyl-histone 3 lysine 27]; 1:100, A7253, ABclonal, Wuhan, China), monomethyl H3K4 (H3K4me1 [methyl-histone 3 lysine 4]; 1:100, A2355, ABclonal, Wuhan, China), CEBPA (3.75 μg, 18311-1-AP, Proteintech, IL), Pura (5 μg, GTX00901, GeneTex, SA), and normal rabbit IgG were used for immunoprecipitation. The cross-linking between histone and DNA was then removed. A DNA purification kit (D0033; Beyotime, Shanghai, China) was used for DNA purification. Finally, qRT-PCR was performed to measure the enrichment of H3k27ac, H3K4me1, CEBPA, and Pura to DNA molecules of interest. The primers (5′–3′) are shown in Table S2.

Biotinylated RNA pull-down assays were performed using the BersinBioTM RNA pull-down Kit (Bes5102, Guangzhou, China) according to the manufacturer’s protocol. A total of 1×10⁷ collected cells were lysed, and nucleic acids were removed by incubating with DNase at 25 °C. After adding streptavidin beads to prewash, RNA probes with streptavidin beads were added after washing to form probe and streptavidin beads complexes. Finally, the eluent was added to a shaker at 37 °C for 2 hours, and 30 μL collected protein samples bound to RNA were used for Western blotting.

Statistical analysis was performed with GraphPad Prism 8 software. Data were checked for normal distribution and equal variance (F test) before statistical testing. Student t test (unpaired) was used for 2-group analysis with equal variance, and Welch correction test was used for 2-group analysis with unequal variance. One-way ANOVA with Tukey post hoc test was used to compare multiple groups with equal variance, and Brown-Forsythe and Welch ANOVA with Tamhane T2 post hoc test was used to compare multiple groups with unequal variance. Nonparametric analyses, including the Mann-Whitney U test for 2 groups or Kruskal-Wallis test followed by Dunn post test for >2 groups, were performed for nonnormally distributed data. Data are presented as the means±SEM, and P<0.05 was considered statistically significant.

CircKrt4 was identified by circRNA sequencing from our previous study²⁴ with log2 (fold change=7.28) and P<0.001. CircKrt4 was generated from the Krt4 gene and located on mouse chromosome 15 (101920551-101922975), and the circular structure of circKrt4 is illustrated in Figure 1A. Sanger sequencing further confirmed the head-to-tail junction, which was consistent with mmu_circKrt4 annotation (Figure 1B). We then revealed that circKrt4 amplified by divergent primers was detectable in cDNA rather than genomic DNA (Figure 1C). As shown in Figure 1D, compared with linear Krt4, circKrt4 was more resistant to RNase R digestion, confirming that circKrt4 was a bona fide circRNA. qRT-PCR assays verified that the expression of circKrt4 was higher in hypoxic lung tissues and plasma than in normoxia (Figure 1E). Next, we performed FISH analysis to determine the distribution of circKrt4 in lung tissues and found that circKrt4 was expressed in pulmonary arteries (Figure 1F). Furthermore, we found that circKrt4 was expressed in both PASMCs and PAECs, with significantly increased expression in PAECs exposed to hypoxia (Figure 1G). By subcellular fractionation and RNA FISH analysis, we clarified the expression of circKrt4 both in the cytoplasm and nucleus, which was upregulated by hypoxia exposure (Figure 1H and 1I).

To assess the function of circKrt4 in hypoxic PH, we designed loss-of-function assays by silencing circKrt4 in PAECs via circKrt4 ASO (antisense oligonucleotide; Figure S1A). The qRT-PCR results showed that ASO-circKrt4 effectively inhibited circKrt4 expression in PAECs (Figure S1B). Cytoplasmic and nuclear RNA isolation further confirmed the decreased level of circKrt4 by ASO-circKrt4 both in the cytoplasm and nucleus (Figure S1C). We found that linear Krt4 was upregulated under hypoxic conditions and that ASO-circKrt4 had no significant effect on the expression of Krt4 at the mRNA and protein levels (Figure S1D and S1E). Due to the essential role of HIFs (hypoxia-inducible factors) in PH, especially HIF-2 constitutes a universal mechanism of endothelium adaptation to prolonged hypoxia.^25,26 We detect the role of circKrt4 in the regulation of HIF-2α in PAECs and found that the expression of HIF-2α was not affected by ASO-circKrt4 (Figure S1F). CCK-8 (Figure 2A) and ethynyl-2’-deoxyuridine assay (Figure 2B) were performed and showed that circKrt4 silencing retarded hypoxia-induced proliferation of PAECs. Western blotting analysis showed that hypoxia increased the expression of PCNA (proliferating cell nuclear antigen), cyclin D, and cyclin E, which was reversed by knockdown of circKrt4 under the same conditions (Figure 2C). Additionally, the effects of circKrt4 on cell cycle progression were verified by flow cytometry, in which circKrt4 silencing reduced the percentage of cells in S+G₂/M phase under hypoxic conditions (Figure 2D).

To determine the role of circKrt4 in hypoxic PH in vivo, we used AAV5 to interfere with circKrt4 and constructed a hypoxia-PH model and a SuHx-PH model. We evaluated RV systolic pressure (Figure 2E; Figure S2A) and RV hypertrophy (Figure 2F; Figure S2B), and found that inhibition of circKrt4 inhibited the above PH index induced by hypoxia and SuHx. According to echocardiographic measurements, circKrt4 silencing attenuated the decrease of pulmonary artery velocity time integral and pulmonary artery acceleration time induced by hypoxia and SuHx which had no affect on LV ejection fraction (Figure 2G; Figure S2C). Furthermore, we found that knockdown of circKrt4 reversed hypoxia and SuHx-induced distal pulmonary vascular remodeling by measuring vascular wall thickening, vascular muscularization (Figure 2H; Figure S2D and S2E), and collagenation (Figure S2F). The interference efficiency of AAV5-circKrt4 is shown in Figure S2G, and the body weight of experimental animals was not affected (Figure S2H). The results suggested that circKrt4 plays a critical role in promoting PAEC proliferation and PH progression associated with hypoxia.

To explore the function of circKrt4 in PAEC dysfunction, we first sought to characterize the altered cellular phenotypes-EndoMT and the cell migration ability. Knockdown of circKrt4 significantly reversed hypoxia-induced decreased expression of E-cadherin (epithelial-cadherin) and increased expression of N-cadherin (neural-cadherin), Snail, and Vimentin (Figure 3A). Similarly, N-cadherin staining indicated that PAEC EndoMT was markedly decreased upon circKrt4 silencing (Figure 3B). Transwell and network formation assays revealed that circKrt4 deficiency inhibited the migration abilities of PAECs (Figure 3C). Increasing evidence demonstrates that mitochondrial injury and metabolic reprogramming are observed during epithelial-mesenchymal transition.^27,28 To evaluate the role of circKrt4 in mitochondrial function in PAECs, oxidative stress, and metabolic changes were determined. Hypoxia-induced oxidative damage, as evidenced by fluorescence staining and ELISA assays demonstrating increases in 8-hydroxy-2 deoxyguanosine, mitochondrial matrix O₂^- and cellular ROS, and decreases in SOD and GPx activities when compared with those of the normoxic group, whereas, ASO-circKrt4 administration partly counteracted the afore-mentioned variations (Figure 3D through 3F). Besides, hypoxia upregulated the expression of HK II (hexokinase II) and PKM2 (pyruvate kinase 2) and decreased the expression of PDH (pyruvate dehydrogenase), which were reversed by circKrt4 knockdown (Figure 3G). In addition, there was a clear decline in OCR (maximal respiration, spare respiration capacity, and ATP production) in mitochondrial function after hypoxia exposure, while circKrt4 silencing treatment alleviated oxidative respiratory dysfunction (Figure 3H). Meanwhile, glycolysis was monitored in real time using a Seahorse XFe24 Extracellular Flux Analyzer. As expected, circKrt4 knockdown decreased the glycolytic stress activation stimulated by hypoxia (Figure 3I). These results indicated that upregulated circKrt4 in hypoxia could result in excessive EndoMT and mitochondrial dysfunction in PAECs.

To investigate the molecular mechanism by which circKrt4 mediates hypoxic PH and PAEC dysfunction, we first predicted that circKrt4 may adsorb miRNAs based on bioinformatics (miRanda, pita, and RNAhybrid) and performed Gene Ontology and Kyoto Encyclopedia of Genes and Genomes analyses (Figure S3A and S3B). qRT-PCR was used to detect the expression of miRNAs with the highest interaction scores, including miR-107-3p, miR-326-3p, miR-6972-5p, and miR-7665-3p. The results showed that there were no detectable changes in these target miRNAs after silencing circKrt4 in hypoxia (Figure S3C). Recent studies have shown that circRNAs can function as protein scaffolds and interact with RBPs (RNA-binding proteins).²⁹ Therefore, we predicted target proteins interacting with circKrt4 by using the catRAPID program. Bioinformatics analysis predicted that the transcriptional activator protein Pura and mitochondrial-bound Glpk (glycerol kinase) could interact with circKrt4 (Figure 4A and 4B). We then used the HNADOCK Server to predict and visualize the 3-dimensional structural docking of circKrt4-Pura and circKrt4-Glpk (Figure 4C and 4D). To confirm such binding, RIP (RNA immunoprecipitation) experiments were performed and revealed that Pura and Glpk directly bound to circKrt4 in hypoxia, more obviously than the linear transcript Krt4 (Figure 4E and 4F). Then, according to the predicted binding site of circKrt4 and target genes (Figure S4A and S4B), we designed biotin-labeled specific probes to pull down the proteins directly bound to circKrt4, and the results showed that Pura and Glpk were indeed among the pulled protein complexes (Figure 4G). Consistent with these results, the FISH assay showed that circKrt4 colocalized with Pura and Glpk in the nucleus and cytoplasm, respectively (Figure 4H and 4I). Furthermore, as shown in Figure 4J and 4K, the expression of Pura and Glpk was increased in hypoxia; however, when circKrt4 was knocked down by ASO, the expression of Pura was markedly decreased, and the expression of Glpk was not affected.

Knowing that circKrt4 is abundant both in the nucleus and cytoplasm, we first sought to examine the function of circKrt4-associated protein Pura in the nucleus. Therefore, siRNAs for Pura were synthesized, and siRNA-3 showed the highest efficiency for Pura silencing and was used in the following assay (Figure 5A). Accordingly, silencing Pura led to decreased migratory capability of hypoxic PAECs (Figure 5B). To investigate how Pura modulates PAEC function, we sought to identify the transcription target genes regulated by Pura, specifically EndoMT-associated proteins. Immunoblotting assays demonstrated that the expression of EndoMT key regulators, including E-cadherin, Snail, and Vimentin, remained unaltered after Pura siRNA transfection (Figure 5C). However, Pura knockdown significantly inhibited the expression of N-cadherin in PAECs (Figure 5C and 5D), suggesting a functional connection between Pura and N-cadherin. To clarify whether Pura affects N-cadherin transcription, we divided the N-cadherin promoter into 5 segments (P1–P5; Figure 5E). The ChIP-qPCR results verified that Pura directly bound to promoter (P) 1 (P1) and P3 within the N-cadherin promoter in hypoxic PAECs, which was reduced by silencing circKrt4 (Figure 5F). Together, the data demonstrated that Pura, as a transcription factor, is required for hypoxia-induced N-cadherin and EndoMT activation.

Next, we studied the role of circKrt4 in the cytoplasm, especially in mitochondria associated with Glpk. Having determined that knockdown of circKrt4 did not affect the expression of the Glpk protein in cells (Figure 4K), the interaction of circKrt4 and Glpk prompted us to determine the role of circKrt4 in the mitochondrial translocation of Glpk. We then examined the expression of Glpk through cytoplasmic and mitochondrial protein isolation and found that the mitochondrial localization of Glpk decreased under hypoxia, and when circKrt4 was silenced, the translocation of Glpk into mitochondria was significantly increased (Figure 6A). Accordingly, an immunofluorescence assay showed that knockdown of circKrt4 promoted Glpk mitochondrial translocation in PAECs (Figure 6B), suggesting that the interaction of circKrt4 and Glpk blocked the mitochondrial shuttling of Glpk in hypoxia. Then, we identified a mitochondria targeting sequence in the Glpk protein using the PSORT II prediction tool³⁰ and found that there was a TRFLV (twin-arginine translocation–related flanking region with conserved Leu-Val residues) motif in the N-FGGY domain of Glpk (Figure 6C). To validate this prediction, we constructed a Glpk-overexpressing plasmid with an unaltered mitochondrial targeting motif (wild-type) and a mutated mitochondrial targeting motif (MUT) and confirmed the overexpression plasmid transfection efficiency (Figure S5A). When the TRFLV motif was mutated, Glpk translocation to mitochondria was diminished compared with wild-type–Glpk (Figure 6D). Then, we tested the influence of the translocation of Glpk from the cytoplasm to mitochondria on PAECs. Glpk mitochondrial translocation inhibition (MUT-Glpk) induced increased levels of mitochondrial ROS, total intracellular ROS, and oxidative DNA damage (8-hydroxy-2 deoxyguanosine; Figure 6E through 6G). In addition, there was a clear decline in maximal respiration capacity, spare respiration, and ATP production in mitochondrial function after transfection with the MUT-Glpk plasmid compared with wild-type–Glpk (Figure 6H). There was no significant difference in the extracellular acidification rate of the glycolysis index in the PAECs after mitochondrial targeting motif mutation (Figure 6I). As shown in Figure S5B and S5C, MUT-Glpk increased the proliferation of PAECs. These molecular data demonstrated that Glpk mitochondrial translocation inhibition resulted in PAEC dysfunction by inducing mitochondrial injury.

Super enhancer (SE) is a large cluster of enhancers that drives gene transcription, including protein-coding genes and noncoding RNAs, by binding transcription factors and coactivators.^31,32 Analysis of SE active histone marker H3K27ac ChIP-seq data generated in mouse small-cell lung cancer identified a SE around Krt4 (Figure S6A), suggesting that Krt4 may be regulated by SE. To determine whether hypoxia-induced upregulation of circKrt4 expression is associated with SE in PAECs, we treated cells with JQ1 (BET bromodomain inhibitor), a SE inhibitor, and found that JQ1 specifically inhibited the expression of both Krt4 and circKrt4 in hypoxia (Figure 7A). ChIP-qPCR assays revealed that hypoxia promoted the binding of the H3K27ac to the individual constituent promoter (P1–P5) of Krt4 and H3K4me1 enriched in the promoter (P2, P3, and P5) of Krt4 in PAECs (Figure 7B). We next attempted to identify candidate transcription factors that contribute to the aberrant activation of circKrt4 in hypoxia. Knowing that Pura interacts with circKrt4 as a transcription factor (Figure 4), we first sought to examine whether the expression of circKrt4 is related to Pura. qRT-PCR showed that knockdown of Pura did not affect the expression of circKrt4 (Figure S6B). After, according to an epigenetic database (CHIP-Atlas) and promoter databases (JASPAR and AnimaiTFDB), we identified CEBPA, which could potentially combine with the promoter region of Krt4 (Figure 7C and 7D). Although the expression of CEBPA did not change in PAECs after hypoxia treatment (Figure 7E), ChIP-qPCR assays revealed that CEBPA could directly bind to the individual constituent promoter (P5) of Krt4 in PAECs under hypoxia (Figure 7F). The coimmunoprecipitation assay results verified the interaction of CEBPA with H3K4me1 and H3K27ac (Figure 7G). Electrophoretic Mobility Shift Assay results of nuclear proteins extracted from PAECs showed that the addition of a CEBPA-specific antibody in the binding reaction led to a super shifted band, indicating that CEBPA was indeed complexed with Krt4 probe (Figure 7H). Meanwhile, PAECs treatment with CEBPA siRNA (Figure S6C) counteracted hypoxia-mediated upregulation of Krt4 and circKrt4 (Figure 7I). We next cloned the promoter region of Krt4 into luciferase reporter vectors and measured their activities. The wild-type but not P5 MUT promoter region was activated by CEBPA in PAECs (Figure 7J). All of the above results indicated that circKrt4 expression is tightly regulated by the interaction between CEBPA and Krt4-acossiated SE (H3K27ac and H3K4me1).

To further explore the formation of circKrt4, we investigated the potential alternative splicing of the Krt4 gene by bioinformatic prediction (RBPmap and catRAPID) and highlighted candidate RBPs (RBM25 [RNA-binding-motif protein 25], KSHRP [KH-type splicing regulatory protein], RBM15B [RNA-binding motif protein 15B], UNK [unkempt family zinc finger], and MATR3 [matrin3]) possibly binding to the flanking sequences of exon 2 and exon 5 (intron 1 and intron 5) of Krt4, in which RBM25 (RNA-binding-motif protein 25) was reported to regulate a large number of alternating splicing exons and was confirmed to promote circRNA formation³³ (Figure 7K). To assess whether RBM25 was indeed able to regulate Krt4 splicing, we performed RIP assay and the results revealed significant binding between Krt4 pre-mRNA with RBM25 by anti-RBM25 compared with anti-IgG (Figure 7L). Subsequently, we knocked down RBM25 using siRNA (Figure S6D) and examined the expression of circKrt4. The results showed that circKrt4 but not Krt4 was significantly downregulated in which RBM25 was knocked down (Figure 7M). Therefore, we speculated that RBM25 was able to bind to flanking sequences of exon 2 and exon 5 to increase circKrt4 formation.

Figure S1–S6

Table S1–S2

Major Resources Table

Uncropped Blots