This study was authorized by the Research Ethics Board of East Hospital, Tongji University (Shanghai, China). All recruited patients or their authorized family members were provided with a consent form. Peripheral blood (4ml) was taken from 7 eligible patients with pneumonia-induced sepsis, 7 pneumonia patients without sepsis and 7 healthy donors. The participants’ clinical parameters are shown in Supplementary Table 1 . The pneumonia patients were classified as sepsis according to the Surviving Sepsis Campaign definitions (17) from the emergency and/or general intensive care unit (ICU) of East Hospital. The pneumonia patients without sepsis who came from emergency internal medicine ward of East Hospital and healthy volunteers came to East hospital for routine physical examination. Pneumonia was defined by a new pulmonary infiltrate on chest radiograph accompanied with at least one of the following signs (18): (a) the presence of cough, sputum production, and dyspnea; (b) core body temperature > 38.0°C; (c) peripheral white blood cell counts > 10 × 10⁹/L or < 4 × 10⁹/L. Among the 21 samples, 3 sepsis samples, 3 pneumonia samples and 3 healthy people samples were used for RNA sequencing analysis, and the remaining samples were used for subsequent tests.

Peripheral blood mononuclear cells (PBMCs) were isolated from peripheral blood according to the protocol as previously reported (19) and CD14⁺ monocytes were sorted from PBMCs with a magnetic cell sorting system (Miltenyi Biotec, Germany). PBMCs were added with CD14 Microbeads(20µl/10⁷cells), and then the CD14⁺ monocytes were magnetically labeled with CD14 Microbeads. When PBMCs passed through a MACS column, the magnetically labeled CD14⁺ monocytes were retained within the column, and then CD14+monocytes were extracted as positively selected cell fraction.

Plasma from patients with pneumonia-induced sepsis, pneumonia patients without sepsis, and healthy individuals was isolated using TRIzol reagent (Invitrogen, USA) per the manufacturer’s instructions. NanoDrop ND-1000 was utilized to measure the RNA’s purity and concentration (NanoDrop Thermo). Through denaturing agarose gel electrophoresis, the RNA integrity of the samples was evaluated. The rRNA was removed using the Ribo-Zero rRNA Removal Kit (Illumina, San Diego, CA, USA). Cloud-Seq Biotech (Shanghai, China) performed the high-throughput whole transcriptome sequencing and subsequent bioinformatics analysis as previously reported (20). The sequencer Illumina HiSeq 6000 was used to obtain paired-end readings. The circular RNA was detected and identified using DCC software (v0.4.4) and the identified circular RNA was annotated using the circBase database and Circ2Tuits. Edger software (v3.16.5) was utilized to identify circRNAs with differential expression.

Total RNA (2µg) was extracted using TRIzol (Invitrogen, USA) followed by reverse transcription of mRNAs and circRNAs using PrimeScript II 1st Strand cDNA Synthesis Kit (Takara, Japan) per the standard manufacturer’s instructions. qRT-PCR assay was performed to measure mRNAs and circRNAs expression with SYBR^® Premix Ex Taq™ II (Takara, Japan) using the Roche 480 Real Time PCR System. GAPDH (encoding glyceraldehyde-3-phosphate dehydrogenase) was used as an internal control for circRNAs and mRNAs, and U6 was employed as an endogenous control for the miRNAs. Relative quantification (2^−ΔΔCT) was used for result analysis. All the primers used were included in Supplementary Table 2 .

The human monocytic leukemia cell line THP-1 was purchased from Chinese Academy of Sciences (Shanghai, China) and was grown in RPMI-1640 medium supplemented with 10% fetal bovine serum and 1% penicillin and streptomycin. THP-1 cells were differentiated into macrophages for 3 hr in the presence of 100 nM phorbol myristate (PMA) and replated. For stimulation, cells were primed with different or indicated concentrations (0.1, 0.5 and 1 μg/ml) LPS for 4 hrs in Opti-MEM, then stimulated with 10 μM nigericin for 2 hrs. GenePharma designed and produced the Circ_0075723 overexpression vector, miR-155-5p mimic, SHIP1 overexpression vector and circ_0075723 silence vector (sh-circ_0075723) (Shanghai, China). THP-1 cells were transfected with overexpression vector (2 μg), miRNA mimic (25 nM) or shRNA (25 nM) using Lipofectamine 3000 (Invitrogen) per the manufacturer’s instructions 24h before LPS/nigericin stimulation. After different stimulations, the supernatants of THP-1 cells were collected for IL-1β and IL-18 analyses or applied to culture human lung microvascular endothelial cells (HLMVEC) (Chinese Academy of Sciences, China) for 24 hours.

The location of circ_0075723 in THP-1 cells is determined by FISH. THP-1 cells are fixed with 4% paraformaldehyde and gradient dehydrated with ethanol. Fluorescent-labeled probe (1 µM) for circ_0075723 is applied during hybridization. We use DAPI (Beyotime, Shanghai, China) to stain the nucleus of macrophages.

The wild-type (WT) sequence and mutant-type (MUT) sequences (binding site mutation with miR-155-5p) of circ_0075723 and SHIP1 were amplified and cloned into PmirGLO reporter plasmid, respectively. The fusion plasmid was cotransfected with either miR-155-5p or miR-NC into HEK293T cells. 48 hours after transfection, the luciferase activity was measured using Picagene Dual SeaPansy luminescence kit (Toyo Inc., Japan) according to the manufacturer’s instructions as reported (21).

4 μg total RNA from THP-1 cells was either untreated (control) or treated with 20 units of RNase R (Epicenter; USA, RNR07250) in the presence of 1× reaction buffer and incubated for 30 min at 37°C. RNA was extracted using acid phenol-chloroform after digestion (5: 1). Then, reverse transcription and qRT-PCR were performed, as described in the RNA extraction and qRT-PCR section. THP-1 cells (1 ×10⁵) were placed in 24-well plates and treated with 250 ng/ml actinomycin D (Act D, Sigma) added to the cell culture medium. The levels of circ_0075723 and NUP153 were measured at 0, 8, 12, and 24 hrs.

Biotin-labeled circ_0075723 probe and oligo probe were obtained from Ribobio. THP-1 cells transfected with circ_0075723 probe or oligo probe were lysed and used for pull-down assay using the Pierce Magnetic RNA ProteinPull-down Kit (Thermo Fisher Scientific) in accordance with the instructions. qRT-PCR was used to detect the expression of specified miRNAs.

ELISAs were performed to measure the concentrations of IL-18 and IL-1β protein from supernatants according to the manufacturer’s instructions (R&D Systems).

Immunoblotting analysis was performed as described previously (19, 22). Densitometry analysis of immunoblot results was conducted by using ImageJ software. The results of three replicated experiments are expressed as mean ± standard deviation (SD) (primary antibodies are listed in Supplementary Table 3 ).

All experiments were done in triplicates and replicated at least three times and all experimental data are presented as the means ± SD. The two-tailed Student t-tests were used for comparisons between two groups, and one-way or two-way analysis of variance (ANOVA) were used for multifactorial comparisons. Statistical analyses were performed with SPSS 20.0 software (SPSS Inc., Chicago, IL, USA) or GraphPad Prism 8.0 (GraphPad Software, La Jolla, CA, United States). A value of P < 0.05 was considered to indicate a statistically significant difference.

To determine circRNAs expression profiles and to identify those that are differentially expressed in pneumonia-induced sepsis, we selected healthy people and pneumonia patients without sepsis as controls and performed RNA-seq analysis of circRNA in the plasma of these three groups. In total, 32,229 circRNAs were expressed in the plasma samples among the healthy people, pneumonia patients without sepsis and pneumonia-induced sepsis patients ( Supplementary Table 4 ). Using the cutoff values of fold change > 2.0 and P < 0.05, 382 circRNAs showed significantly differential expression between the pneumonia-induced sepsis patients and healthy people, including 233 circRNAs that were upregulated and 149 circRNAs that were downregulated ( Supplementary Tables 5, 6 ) ( Figures 1A, B , Supplementary Figure 1 ). Meanwhile, 172 differentially expressed circRNAs were detected between the pneumonia-induced sepsis patients and pneumonia patients without sepsis, in which 98 of them were upregulated and 74 were downregulated ( Supplementary Tables 7, 8 ) ( Figures 1A, B ). Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of differentially expressed circRNAs were showed in Supplementary Figure 2 . GO enrichment analysis showed that the differentially expressed circRNAs were involved in the biological processes, such as cell energy metabolism and histone modification. On the other hand, KEGG analysis revealed that RNA transport and MAPK signaling pathway, which is associated with inflammatory activation of macrophages, was related to the differential expression of circRNAs. Of all the differential circRNAs, we chose markedly downregulated circ_0075723 for next investigations because one of its most likely targeted gene, SHIP1(Src homology 2 domain–containing inositol-5-phosphatase 1), had been previously documented to be one of the negative regulators of TLR4 signaling which was involved in regulating NLRP3 inflammasome activation and pyroptosis (23, 24). Furthermore, we validated the expression of circ_0075723 in CD14⁺ monocytes by qRT-PCR among the three groups and found that circ_0075723 was significantly downregulated in pneumonia-induced sepsis patients comparing to the other two groups ( Figure 1C ). Taken together, we screened multiple differently expressed circRNAs in pneumonia-induced sepsis compared to healthy people and pneumonia without sepsis by RNA-seq and validated the significant downregulation of circ_0075723 in pneumonia-induced sepsis, the differential expression of circRNAs from sepsis suggest possible functions of circRNAs in pathogenesis of sepsis.

Circ_0075723, located at chr6:17648038-17649531, which is derived from the human NUP153 gene and generated by back-splicing mechanism ( Figure 2A ). The sequence was located at the back-splice junction location of circ_0075723 according to Sanger sequencing ( Figure 2B ). Further, we treated THP-1 cells with RNase R exonuclease or actinomycin D to confirm circ_0075723 authenticity and found that the expression of circ_0075723 exhibited RNase R ( Figure 2C ) and actinomycin D resistance ( Figure 2D ), while that of NUP153 mRNA was significantly decreased. This indicated circ_0075723 was stable in THP-1 cells. We then investigated the sub-cellular location of circ_0075723. By RNA fluorescence in situ hybridization (FISH) assays, we found circ_0075723 was mainly localized in the cytoplasm ( Figure 2E ). These studies indicated that circ_0075723 as a circRNA, its biological stability may be advantageous to its function.

We then analyzed circ_0075723 function in pneumonia-induced sepsis. Given that pyroptosis, a typical inflammatory cell death, is a major features/characteristics of sepsis (7, 9), we wondered whether circ_0075723 was involved in the regulation of macrophage pyroptosis. Firstly, we found that CD14+ monocytes from pneumonia-induced sepsis patients had considerably greater levels of the proteins TLR4, NLRP3, ASC1, cleaved caspase-1, IL-1, and GSDMD in contrast to pneumonia patients without sepsis and healthy people ( Figure 3A , Supplementary Figure 3A ). In addition, pneumonia-induced sepsis patients showed a markedly enhanced expression of IL-1β and IL-18 from plasma in comparison to pneumonia patients without sepsis and healthy people ( Figure 3B ). These results indicated that pyroptosis is activated in pneumonia-induced sepsis. Due to the markedly downregulated expression of circ_0075723 in CD14⁺ monocytes from pneumonia-induced sepsis patients, we next examined the specific role of circ_0075723 in pyroptosis of pneumonia-induced sepsis. By usage of different doses of LPS together with nigericin to induce pyroptosis of THP1 in vitro, we found that LPS/nigericin treatment increased the production of proteins and cytokines associated with pyroptosis, including TLR4, NLRP3, ASC1, cleaved caspase-1, IL-1β and GSDMD, whereas downregulated the expression of circ_0075723 in a manner dependent on dose ( Figures 3C, D ). To further examine the direct impacts of circ_0075723 in macrophage pyroptosis, we transfected the circ_0075723-overexpressing vector (OE-circ_0075723) or circ_0075723 silence vector (sh-circ_0075723) to overexpress or knockdown circ_0075723 expression in LPS/nigericin-treated THP-1 cells ( Figure 3E ). Subsequently, we choose sh2-circ_0075723 for further experiments due to the relatively lower expression of circ_0075723 in transfected THP-1 cells than sh1-circ_0075723 and sh3-circ_0075723 ( Supplementary Figure 3B ). Overexpression of circ_0075723 in THP-1 cells showed a strong inhibition of pyroptosis-related proteins and cytokines expression, while silencing of circ_0075723 exhibited the opposite effect ( Figures 3F, G , Supplementary Figure 3C ). In general, these studies indicate that macrophages pyroptosis is activated in pneumonia-induced sepsis patients and circ_0075723 essentially prohibits macrophages pyroptosis in vitro.

Based on bioinformatic predictions from the miRanda and TargetScan databases, miR-155-5p was predicted to bind with circ_0075723, and circ_0075723–miR-155-5p network was depicted in Figure 4A . Additionally, we found the miR-155-5p level were considerably higher in CD14+ monocytes of pneumonia-induced sepsis patients than that of pneumonia patients without sepsis and healthy people ( Figure 4B ). Intriguingly, miR-155-5p expression was dramatically increased in LPS/nigericin-treated THP-1 cells in dose-dependent manner ( Figure 4C ). Further, miR-155-5p expression was modulated via circ_0075723 as miR-155-5p was downregulated by OE-circ_0075723 and upregulated by sh-circ_0075723 ( Figure 4D ). According to these results, we hypothesized that circ_0075723 may modulate macrophages pyroptosis by sponging miR-155-5p. Therefore, we performed RNA pull-down assay to verify the specific connection between circ_0075723 with miR-155-5p and found that miR-155-5p could directly interact with circ_0075723 ( Figure 4E ). Besides, dual-luciferase reporter assay revealed that miR-155-5p mimics dramatically decreased the activity of the circ_0075723 wild-type luciferase reporter but had no effect on the Mut luciferase reporter ( Figure 4F ). As we had observed that circ_0075723 could inhibit macrophages pyroptosis, we further adopted the rescue trials to find that miR-155-5p mimics could partly reduce protective effect of circ_0075723 on macrophages pyroptosis ( Figures 4G, H , Supplementary Figure 4 ). Collectively, our findings indicated that circ_0075723 acts as a “molecular sponge” for miR-155-5p.

To further study the downstream mRNA targets of circ_0075723-miR-155-5p ceRNA network, bioinformatic analysis of the TargetScan database revealed that miR-155-5p could target the 3′-untranslated region (UTR) of SHIP1 ( Figure 5A ). By dual-luciferase reporter assay, we showed that miR-155-5p mimics significantly inhibited the wild-type luciferase reporter activity of SHIP1 and validated the connection relationship between SHIP1 and miR-155-5p ( Figure 5B ). Therefore, SHIP1 might be the gene of interest for miR-155-5p. Additionally, SHIP1 expression was markedly diminished in CD14⁺ monocytes of pneumonia-induced sepsis patients compared with that of pneumonia patients without sepsis and healthy people ( Figure 5C ). Corroborating with the clinical findings, vitro tests also confirmed that LPS/nigericin treatment suppressed SHIP1 expression in a dose-dependent way ( Figure 5D ). As it had reported that SHIP1 was involved in the interplay of miR-155 and TLR4 activation by acting as a key negative regulator of TLR4 signaling (23, 25, 26), while TLR4 signaling might activate NLRP3 inflammasome and promote alveolar macrophage pyroptosis (24). We hypothesized that circ_0075723 might inhibit macrophages pyroptosis through promoting SHIP1 expression by sponging miR-155-5p. To investigate this assertion, we overexpressed miR-155-5p mimics in the LPS/nigericin activated THP-1 cells and found that SHIP1 expression was increased in the company of the circ_0075723 overexpression, meanwhile miR-155-5p mimics could markedly decrease SHIP1 upregulation induced by the circ_0075723 overexpression ( Figure 5E ). Besides, we further transfected SHIP1-overexpression lentivirus vector (OE-SHIP1) and (or) miR-155-5p mimics into the LPS/nigericin activated THP-1 cells, and found that the production of pyroptosis-related proteins and cytokines, such as TLR4, NLRP3, ASC1, cleaved caspase-1, GSDMD, IL-1β and IL-18, were decreased in OE-SHIP1 group, meanwhile miR-155-5p mimics could significantly reverse these effects of SHIP1-overexpression ( Supplementary Figures 5A, B ).

The pathological process of sepsis is complex, and the permeability change caused by vascular endothelial cell damage has a significant role in the pathophysiology of sepsis (27). We have shown that circ_0075723 in macrophages could inhibit pyroptosis-related pro-inflammatory cytokines IL-1β and IL-18 expression by circ_0075723/miR-155-5p/SHIP1 axis, and previous studies had documented that NLRP3 inflammasome associated cytokines IL-1β and IL-18 could increase endothelial cells permeability through inhibiting VE-cadherin expression in endothelial cells (28, 29). We further evaluated whether circ_0075723 in macrophages could modulate VE-cadherin expression in endothelial cells through inhibiting NLRP3 inflammasome associated cytokines IL-1β and IL-18 by circ_0075723/miR-155-5p/SHIP1 axis. To examine the proposition, we collected supernatant from aforementioned-LPS/nigericin activated THP-1 cells bearing altered expression of circ_0075723, miR-155-5p and SHIP1, and then used them to culture human lung microvascular endothelial cells (HLMVEC) in vitro. Firstly, Western blotting outcomes demonstrated that VE-cadherin expression in HLMVEC cells was downregulated culturing in supernatant harvested from LPS/nigericin-treated THP-1 cells in contrast to control and further dramatically downregulated in supernatant from sh-circ_0075723-transfected THP-1 cells, whereas transfected OE-circ_0075723 THP-1 cells supernatant exhibit the opposite ( Figure 6A ). Besides, the expression trend of VE-cadherin in HLMVEC cells was contrary to the expression of IL-1β and IL-18 in above supernatant ( Figure 3G ). Furthermore, rescue tests showed that the effect of circ_0075723 and SHIP1 on VE-cadherin expression in HLMVEC cells and IL-1β and IL-18 expression in the supernatant could be inhibited by miR-155-5p mimics ( Figures 4H ,  6B, C , Supplementary Figure 5B ). Collectively, these results may indicate that Overexpression circ_0075723 downregulates IL-1β and IL-18 expression, promotes VE-cadherin expression in endothelial cells and further protects endothelial cell integrity.