Introduction

Sepsis is a persistent systemic inflammatory condition caused by an extreme immune response to infection, and it is always accompanied by multiple organ dysfunction [1]. To date, sepsis and septic shock have remained leading causes of death in critically ill patients, and the mortality rate of 30-day septic shock is as high as 34.7% [2]. In the initial stage of infection, the endothelial barrier and antimicrobial substances released by immune cells can cooperatively hinder pathogen dissemination [3, 4]. However, overwhelming amounts of inflammatory mediators damage endothelial barrier integrity, which contributes to microcirculatory disturbance and end-organ injury, especially lung injury [5, 6].

During sepsis, neutrophils are often the first immune cells to be recruited to infected sites [7]. Calprotectin (S100a8/a9), a heterodimeric Ca2+-binding protein mainly released by neutrophils, is thought to induce prolonged inflammation and endothelial injury through binding with its receptors, including toll-like receptor 4 (TLR4) and receptor for advanced glycation end products (RAGE) [8,9,10]. According to previous studies, inflammatory mediators, such as S100a8/a9 and neutrophil extracellular traps (NETs, web-like DNA structures adorned with bactericidal proteins), tend to induce mitochondrial metabolic disturbance and disrupt mitochondrial homeostasis [11, 12].

Mitochondrial dynamics (fission and fusion) and mitophagy are vital for maintaining mitochondrial quantity and quality. Specifically, fusion allows mitochondria to transfer gene products for optimal function [13], while fission induces the isolation of impaired mitochondria for degradation through mitophagy (a selective form of autophagy) [14]. Failure at any stage results in the accumulation of damaged mitochondria in the cytoplasm and ultimately cell death.

Previously, the different cell death patterns were considered independent. However, increasing evidence emphasizes the extensive crosstalk among different cell death patterns, and these pathways may be intertwined. Therefore, a novel form of cell death, PANoptosis, was proposed in 2019 [15]. It is a coordinated cell death pattern that involves apoptosis, pyroptosis and necroptosis [16]. Several studies have shown that Z-DNA binding protein 1 (ZBP1) can sense cytosolic DNA and consequently activate PANoptosis [17].

Here, our study provides evidence that S100A8/A9hi neutrophils are present specifically in lung tissues from septic mice. High expression levels of S100a8/a9 induce mitochondrial disorders in endothelial cells, including excessive fission and blocked mitophagy, mainly through Ndufa3 suppression in mitochondrial complex I. Finally, mtDNA released from damaged mitochondria activates ZBP1-mediated PANoptosis.

Results

Increased numbers of S100A8/A9hi neutrophils are found in the lung tissues of sepsis model mice, and these cells exhibit enhanced interactions with endothelial cells

Since the lung is considered as one of the most susceptible organs to sepsis [18], we selected publicly available scRNA-seq data from the lung tissues of sham and CLP mice for analysis (Fig. 1A). First, to explore the mechanisms underlying the excessive immune response to infection, all immune cells were clustered and identified by their marker genes (Fig. 1B, Supplementary Table 4). According to the relative percentages of each cell type, neutrophils accounted for the largest proportion of immune cells and were considerably more abundant in the CLP group (Fig. 1C). Therefore, neutrophils were extracted and reclustered into five subgroups. We found that a special subgroup, which showed greater expression of S100A8 and S100A9 than the other groups, was present only in the lung tissues of septic mice and represented a large proportion of the total neutrophil population (Fig. 1D–G, Supplementary Fig. 1A−D). Additionally, pseudotime analysis revealed that S100A8/A9hi neutrophils were specifically in the late stage of differentiation, indicating that neutrophils were induced to differentiate into S100A8/A9hi neutrophils during sepsis progression (Fig. 1H). Due to the presence of this subpopulation, neutrophils in the CLP group exhibited increased expressions of S100A8 and S100A9 and exhibited increased immune function and metabolism (Fig. 1I, Supplementary Fig. 1E–K).

Fig. 1: Increased numbers of S100A8/A9hi neutrophils exist in lung tissues from CLP mice, and showed enhanced interactions with endothelial cells.
figure 1

A The main workflow of the scRNA-seq; B The UMAP plot was based on scRNA-seq data, and it showed five identified immune cell types; C The sankey diagram showed the percentages of five immune cells in sham and CLP groups; D Five clusters of neutrophils were identified in the UMAP plot; E The changes of five neutrophil subclusters percentage were shown on the sankey diagram; F Volcano map showed upregulated genes in S100A8/A9hi neutrophils compared with other subpopulations; G The expression levels of S100A8 and S100A9 in five neutrophil subclusters on dot plot; H The prediction of neutrophil differentiation trajectories using pseudotime analysis; I The expression levels of S100A8 and S100A9 of neutrophils in sham and CLP groups; J, K The number and strength of interactions among neutrophils, endothelial cells and epithelial cells analyzed by CellChat; L The expression levels of ligand-receptor pairs analyzed by CellChat (Since there is no ligand-receptor pair with high expression among epithelial cells in sham group, the dot plot only shows the expressions of ligand-receptor pairs among epithelial cells in CLP group.); M, N The distribution of S100A8/A9hi neutrophils and NET-related gene+ neutrophils on the UMAP plot. Wilcoxon rank sum test was used for the comparison between two groups. *p < 0.05, **p < 0.01 versus sham group.

Full size image

Previously, we demonstrated that neutrophils could disturb the metabolism of endothelial cells and alveolar epithelial cells to aggravate sepsis-induced acute lung injury (SI-ALI) [11, 19]. Consequently, we next explored the interactions among neutrophils, endothelial cells and epithelial cells by using the “CellChat” R package. The results indicated that the number of interactions among these three cell types increased (Fig. 1J). Notably, from an interaction strength perspective, neutrophils showed enhanced unidirectional effects on endothelial cells, and endothelial cells also affected epithelial cells unidirectionally (Fig. 1K). These findings suggested that endothelial cells might act as a “bridge” in neutrophil-induced epithelial cell damage. We then focused on the ligand‒receptor pairs with upregulated expression in neutrophils and endothelial cells. The results indicated that thrombospondin 1 (Thbs1)-CD47 and Thbs1-CD36 expressions were clearly upregulated (Fig. 1L). Previous studies have revealed that endothelial cells exhibit increased Thbs1 expression after treatment with S100a8/a9 [20]. The above results suggest that neutrophils induce endothelial damage mainly through the release of S100a8/a9.

In addition, our previous study revealed that neutrophil extracellular traps (NETs) released by neutrophils might also damage the endothelial barrier during sepsis [21]. Based on the 137 identified NET formation-related genes, GSEA suggested that neutrophils from the sepsis model mice were able to release more NETs [22] (Supplementary Fig. 1L, Supplementary Table 5). Moreover, the subpopulation with high expression of NET-related genes was mostly composed of S100A8/A9hi neutrophils (Fig. 1M, N, Supplementary Fig. 1M-N). These results indicate that S100A8/A9hi neutrophils, including those in the NET-related gene+ subgroup, play a critical role in endothelial injury during sepsis.

In conclusion, the results of the scRNA-seq data analysis suggest that the presence of S100A8/A9hi neutrophils, which exist especially in the lung tissues of septic model mice, might induce endothelial barrier damage to exacerbate lung injury during sepsis.

High expression levels of S100a8/a9 induce excessive inflammatory responses and acute lung injury, both of which are reversed by an S100a8/a9 inhibitor

Since S100a8/a9 usually exists as a dimer, we evaluated the number of S100A9+ neutrophils by flow cytometry. The results demonstrated that the percentage of S100A9+ neutrophils in the peripheral blood were significantly increased in septic mice compared to control mice (Fig. 2A, B). According to previous studies, the peak of lung vascular injury occurred at 24 h after CLP [57]. Exclusion criteria included: a history of cardiopulmonary arrest before being admitted to ICU; history of connective tissue diseases such as systemic lupus erythematosus (SLE), vascular embolism and pregnancy.

Cecal ligation and puncture (CLP) mouse model

Eight- to ten-week-old male C57BL/6 mice were used for experiment. Following random grou**, the cecal ligation and puncture (CLP) mice model was established using the procedures described in previous studies [58]. In brief, after intraperitoneal anesthesia with 1% pentobarbital sodium (1 mg/kg), the abdominal cavity was opened. The cecum was carefully separated, ligated using 5-0 suture, and punctured with a 20-gauge needle. Next, we extruded a small quantity of feces from cecum and repositioned it before closing abdominal cavity. Each animal received 0.5 ml/10 g of normal saline for rehydration. The sham group received the same surgery without CLP. When mice required treatments, the following drugs were injected intraperitoneally: Paquinimod (10 mg/kg, MCE, Shanghai, China), β-nicotinamide mononucleotide (NMN, 500 mg/kg, MCE, Shanghai, China), Selisistat (EX-527, 5 mg/kg, MCE, Shanghai, China) [46, 59].

Murine sepsis score (MSS)

Seven clinical variables in MSS, including appearance, level of consciousness, activity, response to stimulus, eyes, respiration rate and respiration quality were used to assess the severity of sepsis. Each indicator has a score of 0-4, with a full score of 28. Higher scores mean more severe injury [60, 61].

Flow cytometry

Blood samples from mice were treated with red blood cell lysis buffer [46, 59] (Thermo Fisher). After centrifugation, the supernatant was discarded. The prepared cell suspensions were stained in PBS with the following antibodies: APC anti-mouse/human CD11b (BioLegend, 101211), FITC anti-mouse Ly-6G (BioLegend, 127605), PE anti-S100A9 (Cell Signaling Technology, #93941). Flow cytometry was performed using a BD FACS Aria III flow cytometer according to manufacturer protocol.

Histopathological analysis

Mice were sacrificed for several organs, such as lung, liver, kidney, spleen and intestine, and these tissues were fixed with 4% paraformaldehyde at room temperature for 24 h. Paraffin-embedded tissue sections were stained with hematoxylin and eosin (H&E). According to previous descriptions, the severity of acute lung injury was evaluated through a semiquantitative histology scoring method [62]. Specifically, the score of lung injury was based on these indicators: leukocyte infiltration, alveolar edema, haemorrhage and the thickness of alveolar septa. Two pathologists who blinded to the results graded each indicator from 0 to 3 (0 = normal; 1 = mild; 2 = moderate; 3 = severe), and finally calculated the total lung injury score. Furthermore, paraffin-embedded lung tissue sections were also stained with Masson dye to identify the degree of fibrosis within lung tissues.

TUNEL staining

Paraffin-embedded lung tissue sections were stained with TUNEL to detect cell apoptosis according to manufacturer protocol.

Lung wet-to-dry ratio

We harvested the left lung of mice and obtained its wet weight after drying the surface water. Subsequently, the tissue was dried at 70 °C for 48 h, and the dry weight was acquired. The wet/dry ratio was calculated by dividing the wet weight with dry weight. High wet-to-dry ratio means more severe lung edema.

Semi-quantification of inflammatory mediators

The levels of IL-1β, IL-6 and TNF-α in mouse serum were evaluated by Mouse IL-1β ELISA kit (mIC50300-1, mlbio, Shanghai, China), Mouse IL-6 ELISA kit (IC50325-1, mlbio, Shanghai, China), Mouse TNF-α ELISA kit (mIC50536-1, mlbio, Shanghai, China). Moreover, the levels of S100a8/a9 in mouse serum and human plasma were detected by Mouse S100a8/a9 ELISA kit (ml037985, mlbio, Shanghai, China) and Human S100a8/a9 ELISA kit (ml038517, mlbio, Shanghai, China).

Cell culture and treatments

We obtained Human Umbilical Vein Endothelial Cells (HUVECs) from American Type Culture Collection (ATCC; Manassas, USA) and cultured them in DMEM (Gibco) containing 10% fetal bovine serum (Gibco) and penicillin/streptomycin (Gibco) at 37 °C, 5% CO2 incubator. S100a8&a9 heterodimer protein was purchased from SinoBiological (Bei**g, China). Sirt1 activator SRT1720 was purchased from MCE (5 µM, Shanghai, China) [63]. HUVECs were transfected with Lentivirus-NC or Lentivirus-NRF1 (MOI = 10, Shanghai GeneChem, China) for 72 h to overexpress NRF1, and the sequence was listed in Supplementary Table 3.

Cell viability

A Cell-Counting Kit 8 (Do**do Corp., Kumamoto, Japan) was used to measure relative cell viability based on manufacturer’s instructions.

The Ad-mCherry-GFP-LC3B fluorescence microscopy assay

HUVECs were seeded in 48-well plates (5 × 104 cells/well) one day before transfection. Cells were transfected with adenovirus expressing mCherry-GFP-LC3B fusion protein (MOI = 20, Beyotime, Shanghai, China) for 24 h and then treated with or without S100a8/a9. The images were taken by Olympus microscope. In the absence of autophagy, mCherry-GFP-LC3B under fluorescence microscopy showed dispersed yellow fluorescence. However, in the presence of autophagy, mCherry-GFP-LC3B aggregated on the membrane of autophagosome showing yellow spots. After the fusion between autophagosomes and lysosomes, red dots could be observed, since GFP fluorescence was quenched partially.

Mitochondrial membrane potential assay kit with TMRE

HUVECs were seeded in 48-well plates (2 × 104 cells/well) with or without S100a8/a9 stimulation for 24 h. Then these cells were incubated with TMRE staining working solution (Beyotime, Shanghai, China) at 37 °C, 5% CO2 incubator for 30 min. After that, the supernatant was removed and the cells were washed with medium twice. The samples were then observed under immunofluorescence microscopy. Loss of mitochondria membrane potential was shown by diminished red fluorescence, which occurred in the early stage of cell apoptosis.

Measurement of mitochondrial oxidation

The oxygen consumption rate (OCR) of HUVECs was measured by the Agilent Seahorse XF Cell Mito Stress Test on the Seahorse XFe and XF Extracellular Flux Analyzers. HUVECs (2 × 104 cells/well) were seeded in an XF96 plate and incubated in a medium containing glucose, pyruvate and glutamine. Oligomycin, FCCP and rotenone were used to evaluate OCR. Seahorse Wave software was used to assess all data.

Transmission electron microscopy (TEM)

HUVECs were fixed in 2.5% glutaraldehyde, and then were post-fixed with 1% osmic acid for 2 h. Next, gradient dehydration was performed by the usage of graded ethanol. The sample was embedded in 812 resin, which followed by thin section staining of 2% uranyl acetate. Finally, the ultrastructural images of mitochondria were acquired by the usage of transmission electron microscope (HT7700, Hitachi).

Cytosolic mtDNA isolation

After lysis, HUVECs were centrifuged at 700 × g for 10 min to remove nuclei. Then, we normalized the supernatant volume based on protein concentration. Cell lysate was further centrifuged at 10,000 × g for 30 min for cytosolic fraction isolation, which included mtDNA and nDNA [17]. mtDNA was assessed by RT-qPCR with gene sequences coding for human NADH dehydrogenase 1 as primers. Nuclear DNA was detected using sequences coding human b-globin as primers [23]. The primers for human NADH dehydrogenase 1 and human b-globin were listed in Supplementary Table 2.

NAD+/NADH measurement

NAD+/NADH Assay Kit with WST-8 (Beyotime, Shanghai, China) was used for measure mitochondrial complex I activity. HUVECs were seeded in a 6-well plate (1 × 106 cells/well) before they were lysed with 200 μl NAD+/NADH extraction solution. Then, 100 μl samples were added in centrifugal tubes and heated at 60 °C for 30 min to decompose NAD+. Supernatant was further mixed with working solution and the absorbance of samples was measured at 450 nm.

Immunofluorescence

HUVECs were seeded in 48-well plates (2 × 104 cells/well) with or without S100a8/a9 stimulation. 4% paraformaldehyde was used for fixation for 10 min. The cells were penetrated by using 0.1% Triton for 5 min and further blocked at room temperature for 30 min. Antibody against Ndufa3 (1:500, sc-365351, Santa Cruz Biotechnology), was used for incubation overnight at 4 °C, and Alexa Fluor® 594-conjugated goat anti-rabbit IgG (1:200, ab150080, Abcam) was used to incubation at room temperature for 1 h next day. Finally, the nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI). In order to visualize the expressions of target proteins in endothelial cells from mice lung tissues, paraffin-embedded tissue sections were deparaffinized, rehydrated for antigen retrieval. The primary antibodies used in this study included anti-S100a8 (1:500, GB11421-100, Servicebio), anti-S100a9 (1:750, GB111149-100, Servicebio), anti-CD31 (1:200, GB12063-100, Servicebio), anti-Ndufa3 (1:500, sc-365351, Santa Cruz Biotechnology), anti-LAMP1 (1:500, sc-20011, Santa Cruz Biotechnology); anti-ZBP1 (1:500, 13285-1-AP, Proteintech). And the secondary antibodies used here included iF488-Tyramide (1:500, G1231-50UL, Servicebio) and iF555-Tyramide (1:500, G1233-50UL, Servicebio).

Quantitative real-time PCR

The total RNA from cells and tissues was extracted by using TRIzol reagent (Thermo Fisher), and the quality and quantity of RNA were measured by NanoDropTM ND-1000. PrimeScript RT reagent kit (RR036A, Takara, Shinga, Japan) was used to reverse-transcribe RNA into cDNA. Then we used TB Green PCR kit (RR820A, Takara) and Bio-Rad system to perform RT-qPCR with two repetitions per well. The primer sequences were listed in Supplementary Table 2.

Western blot

Cells and lung tissues were lysed with RIPA Buffer (Solarbio, Bei**g, China), which contains proteinase inhibitor cocktails. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) were used to separate proteins. And then proteins were transferred to polyvinylidene fluoride (PVDF) membranes. The membranes were immersed in blocking buffer and incubated with primary antibodies against S100a8/a9 (1:1000, ab288715, Abcam); phospho-MEK1/2 (1:1000, #9154 S, Cell Signaling Technology); phospho-Erk/2 (1:2000, #4370, Cell Signaling Technology); PGC-1α (1:1000, #AF5395, Affinity); Nrf1 (1:1000, 66832-1-Ig, Proteintech); Total OXPHOS Rodent WB antibody cocktail (6.0 µg/ml, ab110413, Abcam); Sirt1 (1:500, sc-74465, Santa Cruz Biotechnology); phospho-Drp (1:1000, #3455, Cell Signaling Technology); Fis (1:500, sc-376447, Santa Cruz Biotechnology); Mfn1 (1:500, sc-166644, Santa Cruz Biotechnology); Mfn2 (1:500, sc-100560, Santa Cruz Biotechnology); LC3B (1:1000, #2775, Cell Signaling Technology); LAMP1 (1:500, sc-20011, Santa Cruz Biotechnology); ZBP1 (1:500; sc-271483, Santa Cruz Biotechnology); caspase 3 (1:500, sc-56053, Santa Cruz Biotechnology); cleaved caspase 3 (1:500, ab2302, Abcam); GSDMD (1:1000, ab219800, Abcam); N-terminal GSDMD (1:1000, ab215203, Abcam); MLKL (1:5000, 66675-1-Ig, Proteintech); phospho-MLKL (1:1000, ab196436, Abcam); phospho-MLKL (1:1000; ab187091, Abcam); GAPDH (1:1000, GB15004-100, Servicebio); β-Actin (1:1000, GB15003-100, Servicebio).

Bulk RNA-seq and scRNA-seq data

Bulk RNA-seq data from pneumonia-induced sepsis patients (GSE65682) and sepsis patients in ICU (GSE185263) were used to reanalyzed. scRNA-seq data of lung tissues from sham and CLP mice (GSE 207651) were selected for further analysis.

ScRNA-seq data pre-processing

We transferred merged matrix into the R statistical environment for further analysis through Seurat package (v. 4.0.4). Cells expressing <200 or >2500 genes, >5% mitochondrial reads were removed. “NormalizeData” function was performed to normalize the gene expression matrix, and 2000 highly variable genes (HVGs) were identified using “FindVariableFeatures” function. Then, the data were integrated among different samples based on identified anchor points using “FindIntegrationAnchors” function. Finally, we used “FindNeighbors” and FindCluster” functions to cluster and identify cells. And cell clusters were visualized by “RunTSNE” and “RunUMAP” functions.

Cluster marker identification and cell annotation

We identified the differentially expressed genes (DEGs) of each cluster using “FindAllMarkers” function, and the clusters were annotated based on classic marker genes [64].

Pseudotime analysis

We constructed differentiation trajectory using “Monocle 2” with DDRTree and the default parameter.

Pathway and functional enrichment analysis

Gene Ontology (GO) analysis was performed by “clusterProfiler” R package. Gene set enrichment analysis (GSEA) was performed by GSEA software. And gene set variation analysis (GSVA) scores were calculated through “gsva” function. We showed gene lists in supplementary tables.

Cell-cell communication analysis

“CellChat” package was used to evaluate the interactions between cells.

Correlation analysis

The correlation of genes in every endothelial subcluster was analyzed by “corrplot” R package.

The analysis of immune cells proportion

22 immune cells proportions were calculated by the CIBERSORT algorithm.

Survival analysis

“Survival” and “Survminer” R packages were used for Survival analysis. Sepsis patients were classified as “S100A8high” and “S100A8low” groups using “surv_cutpoint” function.

Logistic regression analysis

We took an intersection of marker genes of S100A8/A9hi neutrophils and gene lists from peripheral blood leukocytes of sepsis patients, and then conducted univariate logistic regression analysis. 24 genes (p < 0.001) were selected for multivariate logistic regression analysis to seek independent risk factors of high SOFA scores (>median of SOFA scores).

Cox regression analysis

We selected significantly upregulated marker genes in S100A8/A9hi neutrophils according to scRNA-seq data. Then, we took an intersection of marker genes of S100A8/A9hi neutrophils and gene lists from peripheral blood leukocytes of sepsis patients. The top 40 genes ordered by log2FC were chosen for univariate and multivariate cox regression analyses.

Statistical analysis

We carried out all statistical analyses using v.4.0.0 R and GraphPad Prism 8 software. Experimental data were expressed as means ± standard error of the means (SEM). Unpaired t-test and Wilcoxon rank sum test were used for the comparison between two groups, and one-way ANOVA was used for three or more groups. p < 0.05 was considered statistically significant (*/#/▲p < 0.05, **/##/▲▲p < 0.01).