The human umbilical vein endothelial cells (HUVECs) were available from the Institute of Immunology, Tsinghua University (Beijing, China) and were cultured in endothelial cell medium (ECM; Keycell, Wuhan, China) supplemented with 10% fetal bovine serum (FBS; TianHang, China) and 1% penicillin/streptomycin (Corning, China). Jikai Gene Technology Co. Ltd. (Beijing, China) constructed the lentivirus required for transfection. After transfection, the LPS-induced sepsis group was treated with 1000 ng/ml LPS for 12 h. LPS was purchased from Sigma-Aldrich Co. Ltd (Cat NO. L4516, St Louis, USA).

All procedures with animals were approved by the Institutional Animal Care and Use Committee (IACUC) of the Laboratory Animal Research Center, Tsinghua University. 8-week-old male C57BL/6J mice weighing 22 ± 2g, housed in SPF-grade facilities. All mice were administered adeno-associated virus (AAV) by tail-vein injection. AAVs (10×10¹¹ transducing units in 200 μl of phosphate buffer saline) were purchased from Jikai Gene Technology Co. Ltd. (Beijing, China). After 8 weeks, the efficiency of overexpression was confirmed via quantitative real-time polymerase chain reaction (RT-qPCR). A cecal ligation and puncture (CLP) surgical operation was performed to establish a sepsis model in mice.

Total RNA was extracted with Trizol Reagent (Cat NO. 15596-026, Ambion, Texas, USA) with or without RNase R (20U, 37°C, 30 min; Epicentre Technologies, Madison, WI). Specific convergent primers spanning the circRNA back-splice junction site were designed using the circPrimer (version 2.0). Other primers were designed similarly by Oligo 7 (version 7.37) software. The cDNA was synthesized with the PrimeScript™ RT reagent kit (TOLOBIO, Jiangsu, China). RT-qPCR was carried out in triplicate (biological & technical) with SYBR Green PCR Master Mix (ABI, NY, USA), analyzed by the 2^-ΔΔCt method normalized to GAPDH.

Total protein was extracted from HUVECs/tissues using radioimmunoprecipitation assay buffer (Wuhan Servicebio Technology Co. Ltd. Hubei, China). Proteins were separated by SDS-PAGE and transferred to activated PVDF membranes (Millipore, MA, USA). After blocking with 5% skimmed milk at room temperature for 2 h, membranes were incubated overnight at 4°C with primary antibodies: VE-cadherin (1:200, Santa Cruz, CA, USA), EIF4A3 (1:2000, Keycell, Wuhan, China), α-catenin (1:2000, Keycell, Wuhan, China) as loading control. Horseradish peroxidase (HRP)-labeled secondary antibodies (Beyotime Biotechnology, Shanghai, China; Keycell, Wuhan, China) were incubated for 1 h followed by chemiluminescence detection. A GAPDH antibody (1:1000, Goodhere Biotechnology, Hangzhou, China) was used as a protein loading control.

HUVECs were cultured as monolayers and fixed with 4% paraformaldehyde for 30 min. After blocking with 5% normal goat serum for 1 h, cells were incubated with appropriately diluted primary antibodies (according to manufacturer’s protocol) at 4°C overnight. Following cold PBS washing, Cy3-labeled secondary antibody was applied under dark conditions. Nuclei were counterstained with DAPI (Beyotime Biotechnology, Shanghai, China), and protein expression was analyzed using confocal microscopy (Lycra, Germany).

Collected the right lower lung tissue from mice, fixed it with 4% paraformaldehyde, removed the fixative after 48 h, washed it, and immersed the right lower lung tissue in alcohol for dehydration. Then, the lung tissue in a paraffin embedding machine for embedding, followed by sectioning and baking, dewaxing in xylene, alcohol cleaning, hematoxylin staining, fading in differentiation solution, PBS washing, eosin staining, and finally alcohol dehydration, xylene transparency, and gum sealing. Scanned the sections using a digital slide scanner and analyzed the results (Lycra, Germany).

To evaluate lung injury in septic mice, the right lower lung tissue was stained with HE. The slides were reviewed by two professionals in a double-blind manner, using the Smith scoring system. For each parameter, 20 high-power fields (400×) were randomly selected and scored. Additionally, 5 fields were graded based on the Smith scoring results.

Mice received tail vein injection of 2% Evans blue (Shanghai hui Biotechnology, Shanghai, China). After 1 h anesthesia, systemic perfusion with 10 mL PBS was performed via left ventricle-right atrial route. Following 15-min perfusion, lung tissues were excised, weighed, and homogenized after PBS washing. Tissue homogenates (100 mg wet tissue/3 mL formamide) were incubated at 37°C overnight. Supernatants were collected by centrifugation and Evans blue content was quantified spectrophotometrically (Lycra, Germany) at 620 nm.

Taken the left lung of the mice, sucked up the surface liquid of the lung with absorbent paper, and accurately weighed and recorded the liquid as wet weight. Baked at 60°C for 72 h, weighed the dry weight, continued baking until the constant weight was reached (this was the final dry weight), and calculated the ratio of lung wet to dry weight.

RNA-Binding Protein Immunoprecipitation Kit (Cat NO. 17-700, Millipore, MA, USA) was available for RIP assay. HUVECs (1 × 10⁷) were lysed with RIP lysis buffer and then incubated with RIP buffer adding the magnetic beads and specific antibodies to EIF4A3 (Cat NO. 17504-1-AP, Keycell, Wuhan, China) and IgG (Cat NO. 3420, Cell Signaling Technology, MA, USA). RT-qPCR was carried out to detect co-precipitated RNAs. Three independent biological and technical duplications were conducted.

Fluorescein-labeled hsa_circ_0074158 probes (Robobio, Guangzhou, China) were used for FISH in fixed HUVECs (4% paraformaldehyde, Servicebio, Wuhan, China). After overnight probe hybridization, nuclei were counterstained with DAPI (Beyotime Biotechnology, Shanghai, China). Imaging was performed using an Olympus BX53 microscope and confocal microscopy (Olympus BX53 biological microscope), with signal detection via Fluorescent In Situ Hybridization Kit (Servicebio, Wuhan, China).

Biotin-labeled probes (Robobio, Guangzhou, China) and an RNA-Binding Protein Immunoprecipitation Kit (Millipore, MA, USA) were used for RNA pull-down in HUVECs (1 × 10⁷). Lysates were incubated with Streptavidin Magnetic Beads (Thermo Fisher Scientific, MA, USA) at room temperature (1 h). RNA and proteins were extracted, and RNA-binding proteins were analyzed via mass spectrometry (Wuhan Institute of Biotechnology, Wuhan, China).

HUVECs were treated with LPS (1000 ng/ml) for 12 h. And then culture the HUVECs in an endothelial cell medium containing 500 ng/mL actinomycin D (Cat NO. 50-76-0, Merck, Darmstadt, Germany) The HUVECs are collected for RT-qPCR at different times (0 h, 4 h, 8 h, 12 h, 24 h).

Data were analyzed using SPSS 26.0 and expressed as mean ± SD. Group comparisons were performed via one-way ANOVA (multiple groups) or Student’s t-test (two groups), categorical variables by chi-square/Fisher’s exact test, and survival outcomes by log-rank test/Cox regression. Statistical significance was defined as P < 0.05. Graphs were generated with GraphPad Prism 8.0.2.

The current understanding of sepsis pathophysiology highlights the crucial role of endothelial barrier dysfunction in the disease’s progression. Traditional approaches have mainly focused on targeting inflammatory mediators and activating immune cells, yet the underlying mechanisms contributing to endothelial barrier disruption are still not fully understood. Recent studies have implicated circRNAs as key regulators in various pathological processes, including sepsis. Among these, hsa_circ_0074158 has been identified as a potential modulator of endothelial cell function, especially in maintaining barrier integrity. Findings from this investigation demonstrate that the downregulation of hsa_circ_0074158 expression significantly mitigates endothelial barrier dysfunction in a murine model of sepsis. To identify circRNAs associated with CTNNA1, we searched the circBase database (http://www.circbase.org/), revealing 20 human circRNAs and 2 mouse homolog circRNAs associated with CTNNA1 ( Supplementary Table 1 ). Upon searching circBank (http://www.circbank.cn/), we discovered that hsa_circ_0074158 and mmu_circ_0000860 are homologous ( Figure 1A ), leading us to designate mmu_circ_0000860 as circ_Ctnna1 for further study. AAV was employed to downregulate circ_Ctnna1 in mice. We induced sepsis in mice via the CLP method to explore the impact of hsa_circ_0074158 on the endothelial barrier in sepsis. Measurements including lung injury score, Evans blue staining, HE staining, wet-dry weight ratio, and serum glycocalyx degradation products in septic mice indicated that downregulation of circ_Ctnna1 could reduce lung injury, alleviate endothelial hyperpermeability, decrease serum glycocalyx degradation ( Figures 1B-F ), and lower the 7-day mortality rate in mice ( Figure 1G ). These results underscore the significance of hsa_circ_0074158 in the pathogenesis of sepsis-induced endothelial injury and suggest that targeting hsa_circ_0074158 may offer a novel therapeutic strategy for preserving endothelial barrier function in sepsis.

To further explore the mechanism of the effect of hsa_circ_0074158 on endothelial barrier function in sepsis, cellular experiments were conducted. We found that hsa_circ_0074158 was distributed in both the nucleus and the cytoplasm, suggesting that hsa_circ_0074158 may have a biological function in both the nucleus and the cytoplasm ( Figures 2A, B ). To investigate the pathogenesis of hsa_circ_0074158 in sepsis, we performed RNA pull-down experiments and mass spectrometry. By analyzing the mass spectrometry, the qualitative or quantitative information of the proteins was selected, and the experimental group and the control group proteins were identified with 1086 hsa_circ_0074158 specific binding proteins ( Figure 2C ). The analytical protein results showed that hsa_circ_0074158 bound to RBP more highly than the other proteins ( Figure 2D ). RBP is a class of proteins that can associate with RNA molecules and accompany RNA to regulate metabolic processes. The interaction between circRNA and RBP can affect the mRNA stability of host genes (11, 14, 22, 23), thus affecting tumor progression. However, the effect of hsa_circ_0074158 combined with RBP on the endothelial barrier function in sepsis state was not reported. Meanwhile, we predicted RBP combined with hsa_circ_0074158 by CircInteractome, RPIseq, and RPIseq SVM databases and drew a Wayn diagram, which showed a high probability of hsa_circ_0074158 combined with AGO 1, AGO 2, AGO 3, EIF4A3 and CAPRIN1 ( Figure 2E ). Therefore, we speculated that hsa_circ_0074158 might bind to RBP, where the above five RBP are more likely. We have found that hsa_circ_0074158 and EIF4A3 colocalized ( Figure 2F ). The Circular RNA Interactome database (https://circinteractome.nia.nih.gov/) shows EIF4A3 binding to the gene sequence and flanking region of hsa_circ_0074158, suggesting that EIF4A3 may not only interact with hsa_circ_0074158 but also affect the generation of hsa_circ_0074158 ( Figure 2G ; Supplementary Figure 1 ). More importantly, we have found that hsa_circ_0074158 forms a complex with EIF4A3 ( Figure 2H ). Thus, we speculated that EIF4A3 may be associated with sepsis endothelial barrier function and participate in the development of sepsis.

The results of hsa_circ_0074158 mass spectrometry showed that hsa_circ_0074158 is combined with EIF4A3 ( Figure 3A ). EIF4A3 is an adenosine triphosphate-dependent RNA helicase, which is a core component of the exon-junction complex and plays an important role in RNA metabolism. After binding with circRNA, EIF4A3 can prevent the nuclear export of host gene mRNA and thus inhibit tumor metastasis, promote tumor progression by enhancing mRNA stability, and protect atherosclerotic plaque stability by inhibiting autophagy. In conclusion, EIF4A3 has important significance in the development and progression of diseases, but the correlation of EIF4A3 with endothelial barrier function in sepsis remains unclear. To screen proteins bound to hsa_circ_0074158, we performed RNA pull-down experiments and mass spectrometry on hsa_circ_0074158 and analyzed proteomic results. Gene Ontology (GO) showed that proteins interacting with hsa_circ_0074158 were distributed in the nucleus, cytoplasm, and cell junction, and can bind to DNA and RNA, and these proteins participate in biological processes such as DNA transcription and molecular function regulation ( Figures 3B-D ). Kyoto Encyclopedia of Genes and Genomes (KEGG) showed that the proteins interacting with hsa_circ_0074158 were mainly enriched in the pathways associated with infection ( Figure 4E ). Therefore, we hypothesized that EIF4A3 may be associated with the sepsis endothelial barrier function and may be involved in the regulation of the endothelial barrier function in sepsis.

VE-cadherin and the integrity of the glycocalyx system, a crucial molecular component of endothelial barrier function, demonstrate significant impairment during the pathological process of sepsis, a notable alteration validated by various methodologies, including RT-qPCR, WB, and electron microscopy techniques ( Figures 4A-F ). Importantly, the RIP assay in preliminary experiments not only confirmed the molecular interactions between hsa_circ_0074158 and EIF4A3 but also revealed, for the first time, that EIF4A3 formed a specific binding with the mRNA of the host gene CTNNA1 ( Figure 4G ). Based on the findings that (i) hsa_circ_0074158 is derived from reverse splicing of the CTNNA1 gene; (ii) hsa_circ_0074158 significantly affects endothelial barrier function in a sepsis model; and (iii) the CTNNA1 gene encodes α-catenin, a key regulatory protein of the endothelial barrier, we propose the scientific hypothesis that hsa_circ_0074158 may regulate endothelial barrier function through EIF4A3. hsa_circ_0074158 could be involved in the pathological process of sepsis by modulating the expression of the host gene CTNNA1. Experimental data indicated that both mRNA and protein expression levels of CTNNA1 exhibited significant down-regulation in the sepsis group (P<0.01), and hsa_circ_0074158 was negatively correlated with CTNNA1 expression ( Figures 4H-L ). These findings suggest that hsa_circ_0074158 may inhibit the transcription of the host gene CTNNA1 through an epigenetic regulatory mechanism, exacerbating sepsis-associated endothelial barrier dysfunction.

Our previous investigation revealed distinct expression patterns of hsa_circ_0074158 and its associated molecules in sepsis pathogenesis. Specifically, hsa_circ_0074158 showed significant upregulation in sepsis (8), while the transcriptional activity of its host gene CTNNA1 demonstrated concurrent suppression ( Figures 4H, I ). Complementary analyses using RT-qPCR and immunoblotting further identified a notable increase in EIF4A3 expression in the sepsis cohort, with a strong positive correlation observed between circ_0074158 and EIF4A3 levels ( Figures 5A-D ). RIP experiments confirmed a direct molecular interaction between EIF4A3 and hsa_circ_0074158 ( Figure 4G ), suggesting functional interplay within this regulatory axis. To delineate the mechanistic significance of the hsa_circ_0074158/EIF4A3/CTNNA1 axis, we employed siRNA-mediated EIF4A3 knockdown followed by comprehensive phenotypic assessments. Subsequent analyses revealed a significant recovery of CTNNA1 expression at both transcriptional and translational levels after intervention ( Figures 5E-G ). Notably, EIF4A3 suppression reduced sepsis-induced glycocalyx degradation ( Figure 5H ). Collectively, these findings establish a pathogenic cascade wherein: (i) sepsis-induced hsa_circ_0074158 overexpression exacerbates endothelial barrier dysfunction; (ii) concomitant EIF4A3 upregulation synergistically worsens vascular compromise; and (iii) bidirectional molecular interactions between these effectors create a self-amplifying circuit driving disease progression.

To delineate the regulatory mechanism of the hsa_circ_0074158/EIF4A3 axis on its host gene CTNNA1, we established the following genetic perturbation models in HUVECs: si-circ_0074158, OE-circ_0074158, and si-EIF4A3. Multidimensional validation through RT-qPCR and WB demonstrated that hsa_circ_0074158 interacts with EIF4A3 to modulate sepsis-induced endothelial barrier dysfunction by regulating CTNNA1. Rescue experiment design: (i) Control group; (ii) LPS group; (iii) LPS + OE-circ_0074158 + si-EIF4A3; (iv) LPS + OE-circ_0074158; (v) LPS + si-circ_0074158; (vi) LPS + si-EIF4A3. RT-qPCR results indicated that in sepsis, compared to the LPS group, the expression of CTNNA1 was up-regulated in the LPS + OE-circ_0074158 + si-EIF4A3 group, and down-regulated in the LPS + OE-circ_0074158 group. The expression of CTNNA1 was also up-regulated in the LPS + si-circ_0074158 group, and in the LPS + si-EIF4A3 group, with the differences being statistically significant ( Figure 6A ). The expression was further confirmed by WB, IF and electron microscopy ( Figures 6B-D ). To assess the post-transcriptional regulation of CTNNA1, actinomycin D was added to HUVECs subjected to circ_0068631 knockdown (si-circ_0068631), hsa_circ_0074158 overexpression (OE-circ_0074158), and EIF4A3 knockdown (si-EIF4A3). Notably, CTNNA1 mRNA stability increased due to hsa_circ_0074158 depletion or EIF4A3 depletion and decreased due to hsa_circ_0074158 overexpression ( Figures 6E-G ). The mechanism of the hsa_circ_0074158/EIF4A3/CTNNA1 axis in sepsis is depicted in Figure 7 .