FPR1-dependent Pro-inflammatory Ccl4^high monocytes/macrophages drive and predict sepsis-induced acute lung injury

Abstract

The elusive pathogenesis of sepsis-induced acute lung injury (ALI) combined with the absence of reliable diagnostic biomarkers, significantly hinders the development of targeted therapies and precision medicine approaches for affected patients. In this study, we employed scRNA-seq to profile the transcriptional landscape of the lungs in well-established murine models of sepsis-induced ALI. These findings were further validated experimentally leading to the identification of tri-lineage candidate biomarkers for sepsis-associated ALI: SCGB3A2 (in epithelial cells), AKAP12 (in endothelial cells), and CCL4 (in monocytes/macrophages, Mo/Mφ). In pulmonary immune microenvironment of sepsis-induced ALI, Mo/Mφ were identified as the dominant contributors to pulmonary immune heterogeneity during the acute inflammatory phase of sepsis-induced ALI. Among these, the Ccl4^high Mo/Mφ subpopulation exhibited disease-specific lung infiltration and a distinct proinflammatory phenotype. Mechanistically, FPR1 was up-regulated in hyperinflammatory cluster of Ccl4^high Mo/Mφ. Pharmacological inhibition of FPR1 in vivo selectively reduced pulmonary infiltration of Ccl4^high Mo/Mφ and attenuated sepsis-induced ALI. Furthermore, we established a peripheral blood-based five-gene panel (CCL4, NFKBIA, IL1B, BATF, and XBP1) derived from signatures of Ccl4^high Mo/Mφ, which robustly predicted sepsis-induced ALI in clinical cohorts. Collectively, our work delineates the transcriptional alterations of lungs in sepsis-induced ALI, identifies candidate biomarkers for sepsis-induced ALI, and establishes a clinically applicable diagnostic model for early detection. These findings enhance our understanding of the underlying mechanisms of sepsis-induced ALI and offer new targets for its precision diagnosis and therapeutic intervention.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12931-025-03385-5.

Introduction

Sepsis is a life-threatening multiorgan dysfunction caused by a disordered host response to infection. It carries high morbidity and mortality rates. The lung is particularly vulnerable to sepsis-induced damage [1, 2]. Sepsis-induced acute lung injury (ALI) can rapidly progress to acute respiratory distress syndrome (ARDS) and respiratory failure, making it one of the leading causes of death in intensive care units (ICUs) [3, 4]. The mechanisms of sepsis-induced ALI pathogenesis are complex and remain unclear [5, 6]. Previous research on the pathogenesis of sepsis-induced ALI have primarily focused on single cellular components and lacked a systematic overview of alterations in the pulmonary microenvironment [7, 8].

Single-cell RNA sequencing (scRNA-seq) is a powerful tool that can be used to examine specific cellular components and states [9, 10]. Several studies have applied scRNA-seq to characterize the spectrum of immune cell states in the peripheral blood of patients with sepsis [11, 12]. However, previously identified peripheral blood signatures may lack tissue specificity for sepsis-induced ALI. To solve this problem, scRNA-seq of fresh lung tissues from sepsis-induced ALI models may provide valuable insights into the dynamic remodeling of the pulmonary microenvironment and helps identify highly tissue-specific biomarkers.

Epithelial and endothelial cells are primary injured targets in sepsis-induced ALI. Tracing their transcriptional reprogramming and crosstalk with immune cells contributes to elucidate mechanisms of sepsis-induced ALI pathogenesis [13–15]. Monocytes/macrophages (Mo/Mφ) play a central role in pathogenesis of sepsis-induced ALI, exhibiting remarkable plasticity and diverse functions (defense, inflammation, repair) [16–18]. Currently, a traditional binary classification scheme is widely accepted, categorizing macrophages into proinflammatory M1 and anti-inflammatory M2 types [19]. However, the traditional M1/M2 dichotomy oversimplifies their heterogeneity and targeting polarization has yielded limited clinical benefit [20]. scRNA-seq now enables precise identification of disease-specific Mo/Mφ subsets. This refined understanding of macrophage diversity promises to: (i) elucidate previously unrecognized pathogenic mechanisms, (ii) facilitate development of targeted immunomodulatory therapies, and ultimately improve clinical outcomes in sepsis-induced ALI [21].

C-C motif chemokine ligand 4 (CCL4), produced by stromal cells, epithelial cells, monocytes, lymphocytes, and neutrophils, serves as a potent attractant for immune cells expressing CCR1/CCR5 receptors [22, 23]. While earlier research primarily focused on CCL4 secretion and chemotaxis in neutrophils, emerging evidence underscores its critical role in inflammatory responses mediated by Mo/Mφ. Formyl peptide receptor 1 (FPR1) is a transmembrane G protein-coupled receptor [24] and is an important pattern recognition receptor on innate immune cells. The role of FPR1 in inflammatory injury is controversial. The overexpression of FPR1 on microglia exacerbates inflammatory injuries in the brain by regulating neutrophil homing and the migration of other immune cells [25, 26]. On the other hand, the insufficiency of FPR1 in mice compromises the elimination of pathogens, aggravating infection and inflammation and increasing mortality [27, 28]. The role of FPR1 in the pathogenesis of sepsis-induced ALI has not been studied in depth.

Here, we conducted scRNA-seq on lung tissues from well-established sepsis-induced ALI murine models and controls. We identified novel cell subpopulations significantly associated with pathogenesis of sepsis-induced ALI, including Scgb3a2^high alveolar epithelial cells, Akap12^high endothelial cells and Ccl4^high Mo/Mφ. Moreover, we propose a refined classification pattern for sepsis ALI-specific macrophage states and experimentally validated the role and mechanism of Ccl4^high Mo/Mφ proinflammatory subsets in mediating lung injury. These findings can help elucidate the pathogenesis of sepsis-induced ALI through insights into the cellular composition, state, and dynamics of the pulmonary microenvironment and can lead to the development of diagnostic and therapeutic targets for the clinical workflow of sepsis-induced ALI.

Methods

Sample collection and processing. Samples of patients

Fresh whole blood samples were collected from 16 patients diagnosed with sepsis within 24 h of admission to the ICU at Nankai hospital, Tianjin. Of the 16 patients, 10 exhibited oxygenation impairment (PaO₂/FiO₂ < 300 mmHg) and were diagnosed with sepsis-induced ALI. The remaining 6 sepsis patients without lung injury (PaO₂/FiO₂ ≥ 300 mmHg), were included as controls. The diagnosis of sepsis was based on the Sepsis 3.0 definition. Inclusion criteria included age ≥ 18 years, a confirmed diagnosis of sepsis, and the ability to provide informed consent. The infection sources of patients were confined to the gastrointestinal tract, pancreatitis, and the hepatobiliary system, based on the case distribution within our department. Exclusion criteria included a history of chronic respiratory disease, immunosuppressive therapy, and pulmonary edema caused by cardiac origin, or refusal to participate. Written informed consent was obtained from all participants or their legal representatives. The study was approved by the Institutional Review Board of Tianjin Nankai Hospital (Approval No. NKYY_YXKT_IRB_2024_072_01). All patients were received continuous monitoring of vital signs and managed similarly according to the standardized recommendations of our ICUs to ensure hemodynamic stability prior to sample collection. Blood samples were collected on the day of admission. Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood using density gradient centrifugation with Ficoll-Paque Plus (GE Healthcare) within 2 h of collection. The isolated PBMCs were cryopreserved in liquid nitrogen until further analysis.

Samples of animal

For the animal study, adult male C57BL/6J mice (aged 6–8 weeks, weighing 20–26 g) were provided by the Laboratory Animal Center of Tianjin Nankai Hospital. The animals were housed under specific pathogen-free conditions (22–24 °C, 12 h light-dark cycle) with access to water and chow ad libitum. Mice were randomly assigned into two groups of equal-sized litters of 6–8. This study was approved by the Animal Ethical and Welfare Committee of Tianjin Nankai Hospital, Tianjin Medical University (Approval No. NKYY_DWLL_2023_077).

Establishment of sepsis-induced ALI models

Murine models of sepsis-induced ALI were established using lipopolysaccharide (L4391, Sigma) or the cecal ligation and puncture (CLP) procedure. Following randomization, mice were either intraperitoneally injected with 15 mg/kg LPS or equivalent dose of PBS vehicle. Mice were monitored for clinical signs of distress and sacrificed at 18 h for tissue collection and analysis. In the CLP procedure, mice were subjected to CLP surgery under anesthesia. The cecum was ligated at approximately two-thirds of its length from the ileocecal valve. The cecum was perforated by single through-and-through puncture midway between the point of ligation and the tip of the cecum. The puncture was consistently made in a mesenteric-to-antimesenteric direction. 20-gauge needles were used for all procedures to ensure a uniform puncture size. The cecum was punctured twice through the same initial hole [29]. The abdomen was then closed with sutures. Sham-operated mice underwent the same procedure without cecal ligation or puncture. All CLP and corresponding control operations were carried out by one experienced technician to minimize technical variation. Mice were monitored for survival over a 48-hour period, and lung tissues were harvested at 18 h post-surgery for histological analysis and molecular studies.

Sepsis-induced ALI models validation

Twelve hours post the LPS intraperitoneal injection, the murine sepsis score (MSS) in the experimental group mice significantly increased. This score is a robust and refined tool used to assess the severity of sepsis in mouse models, with scoring items involving the general appearance of the mice (score 0–4), level of consciousness (score 0–4), activity level (score 0–4), respiratory rate (score 0–4) and quality (score 0–4), eye appearance (score 0–4), and response to handling/stimulus (score 0–4) [30]. The MSS scoring details for individuals were provided in Table 1. Mice were euthanized by i.p. injection of pentobarbital sodium (180 mg/kg). After carefully dissecting the thoracic cavity and exposing the lungs, the lung intact was removed. We weighed the freshly excised lungs immediately and recorded the weight as the wet weight (WW). Then, we placed a portion of the lungs in a drying oven at 65 °C for 24 h to ensure complete dehydration. After drying and cooling lungs to room temperature in the desiccator, we weighed the dried lungs and recorded the weight as the dry weight (DW). Subsequently, we calculated the ratio of wet lung to dry lung using the formula: W/D ratio = WW/DW to measure the permeability of alveolocapillary membrane and severity of lung edema. Additionally, a portion of the lungs were extracted and retained in 4% paraformaldehyde (P1110, Solarbio) fixation solution for 48 h at 4 °C for subsequent embedding in paraffin.

Table 1.

Murine sepsis score (MSS) to evaluate the severity of disease

Variable	Score
General appearance	0- Coat is smooth.
	1- Patches of hair piloerected.
	2- Majority of back is piloerected.
	3- Piloerection may or may not be present, mouse appears “puffy”.
	4- Piloerection may or may not be present, mouse appears emaciated.
Level of consciousness	0- Mouse is active.
	1- Mouse is active but avoids standing upright.
	2- Mouse activity is noticeably slowed. The mouse is still ambulant.
	3- Activity is impaired. Mouse only moves when provoked, movements have a tremor.
	4- Activity severely impaired. Mouse remains stationary when provoked, with possible tremor.
Activity level	0- Normal amount of activity. Mouse is any of: eating, drinking, climbing, running, fighting.
	1- Slightly suppressed activity. Mouse is moving around bottom of cage.
	2- Suppressed activity. Mouse is stationary with occasional investigative movements.
	3- No activity. Mouse is stationary
	4- No activity. Mouse experiencing tremors, particularly in the hind legs.
Respiratory rate	0- Normal, rapid mouse respiration.
	1- Slightly decreased respiration (rate not quantifiable by eye).
	2- Moderately reduced respiration (rate at the upper range of quantifying by eye).
	3- Severely reduced respiration (rate easily countable by eye, 0.5 s between breaths).
	4- Extremely reduced respiration (> 1 s between breaths).
Respiration quality	0- Normal.
	1- Brief periods of laboured breathing.
	2- Laboured, no gasping.
	3- Laboured with intermittent gasps.
	4- Gasping
Eye appearance	0- Open.
	1- Eyes not fully open, possibly with secretions.
	2- Eyes at least half closed, possibly with secretions.
	3- Eyes half closed or more, possibly with secretions.
	4- Eyes closed or milky.
Response to handling/stimulus	0- Mouse responds immediately to auditory stimulus or touch.
	1- Slow or no response to auditory stimulus; strong response to touch (moves to escape).
	2- No response to auditory stimulus; moderate response to touch (moves a few steps).
	3- No response to auditory stimulus; mild response to touch (no locomotion).
	4- No response to auditory stimulus. Little or no response to touch. Cannot right itself if pushed over.

Histopathology

Paraffin-embedded lung tissues were sliced and stained with hematoxylin and eosin (H&E) to assess the index of quantitative assessment (IQA) of lung injury for individual. We used a semiquantitative scoring system to evaluate lung injury based on the following indicators: alveolar edema, cellular infiltration, and hyaline membrane formation. Each index was classified into four levels: no abnormalities, less than 20% involved, 20–50% involved, and more than 50% involved, which were assigned scores of 0, 1, 2, and 3, respectively. The average values were evaluated by two blinded pathologists as a semiquantitative index of quantitative assessment (IQA) of lung injury. Tissue dehydration, paraffin embedding, sectioning, and staining were performed on histological and molecular experiment platform from Tianjin Nankai Hospital.

Single-cell isolation of lung tissues

Single-cell suspension was generated according to the published protocols [31]. After mice euthanasia and exposing their thoracic cavities, the lung tissues were removed and washed in pre-cooling RPM1640. The tissues were finely minced and dissociated using Tissue Dissociation Reagent A (K01301-30, SeekOne). With incubation at 37 °C until completely digestion, cell suspensions were filtered through 70 μm sterile cell filter nets and centrifuged at 300× g for 5 min at 4 °C. Resuspend the cell pellet in 1640 medium containing 2% fetal bovine serum (FBS) and remove red blood cells using Red Blood Cell Lysis Buffer (R1010, Solarbio). Then, we performed AO/PI staining and assessed cell quality using a fluorescence cell analyzer (Countstar^® Rigel S2). Finally, the cells were washed twice in RPMI1640 and resuspended in 1× PBS containing 0.04% BSA at a concentration of 1 × 10^6 cells/mL. All above procedures were performed on ice, and all cell suspensions were kept on ice to minimize the cell loss.

ScRNA-seq library construction and sequencing

Single-cell suspensions were ensured to meet for following criteria: viability of cells exceeding 90%, cell agglomeration rate < 10%, nucleation of cells > 70%, with a cell concentration 700 ~ 1200 cells/µL, and cell’s diameter varied among 5 ~ 40µm. Single suspensions were processed using the SeekOne^®Digital Droplet Single Cell 3’ library preparation kit (K00202, SeekGene). cDNA libraries were prepared according to the manufacturer’s protocol within 24 h. We utilized paired-end sequencing on the Illumina NovaSeq 6000 platform (PE150) to sequence the final libraries with an approximate depth of 20,000–25,000 reads/cell.

ScRNA-seq raw data processing and analysis

Raw sequencing reads were aligned with the mouse genome mm10 using the CellRanger v7.2.0 (10X Genomics) with default parameters. We utilized Seurat R package v5.0.1 to perform standard preprocessing and quality control for scRNA-seq data. Cells with less than 200 or more than 2000 detected gene, more than 15% mitochondrial genes and 10% ribosomal genes, or UMI < 500 were excluded. After filtering, a total of 23,215 cells with high quality were obtained from two groups, each consisting of three biological replicates of mice. Subsequently, we normalized the data to account for differences in sequencing depth across cells and identified highly variable features across the dataset for downstream analysis. After scaling the data, using Principal Component Analysis (PCA) technique to perfume dimensionality reduction. We constructed a K-nearest neighbor (KNN) graph based on the PCA dimensions. Harmony R package v1.2.0 was used to integrate the datasets and lessen the batch effects. The clustree R package v0.5.1 was performed to determine the optimal resolution for dimensionality reduction, set at 0.5. UMAP (Uniform Manifold Approximation and Projection) was utilized to visualize cells with approximate local neighborhoods in high-dimensional space into two dimensions. Further differential expression analysis was conducted in virtue of Seurat’s FindAllMarker and FindMarkers functions. For all comparisons between experimental conditions, the p-values derived from statistical tests were adjusted for multiple testing using the Benjamini-Hochberg method to control the False Discovery Rate (FDR). Genes with an adjusted p-value (FDR) < 0.05 were considered statistically significant. Cell types were manually annotated according to a compiled panel of canonical tissue compartment markers. The enrichment of 50 hallmark biological pathway from the Molecular Signatures Dataset (MSigDB, GSEA) in distinct clusters were assessed using Gene Set Variation Analysis (GSVA). We converted the GSVA scores into data frames and add them into Seurat object for subsequent visualization and interpretation.

Construction multi-gene signatures scoring for individual cell

The AddModuleScore() function in Seurat was performed to calculate the aggregated expression scores for self-defined sets of genes. This function evaluated the average expression of a predefined gene set and compared it to a random control gene set to determine if the module was specifically enriched in each cell, allowing us to assess the activity of specific pathways or biological processes in different cellular context.

ScRNA-seq pseudotime Analysis

We performed Monocle v2.26.0 pipeline for pseudo time inference in myeloid cells. As requested of the algorithm, we created Monocle CellDataSet (CDS) objects and preprocessed data using the default parameters. After dimensionality reduction, and cell clustering, order_cells() core function was conducted for pseudotime analysis. Sebsequently, Monocle’s built-in visualization tool plot_cells() was utilized to display the distribution of cells along the inferred pseudotime trajectory. Within the Ccl4^high macrophages subset, integration of cytoTRACE v0.3.3 and slingshot v2.7.0 dual pseudotime approach provided us a comprehensive understanding of cellular dynamics, of which cytoTRACE was used to estimate the differentiation potential of individual cells and determine the starting points of developemental trajectories, whereas slingshot inferred lineage trajectories and provided us insight into lineage relationships across cells.

Cell-cell communication inference

We employed the CellChat R package (v1.6.1) to investigate cell-cell communication networks among various pulmonary cell types. CellChat’s core principle is based on identifying intercellular interactions by quantifying the expression of ligands and their corresponding receptors across cell types or clusters, while also valuing the interaction weights to eliminate the effect of cell population size. The algorithm models these interactions by leveraging known ligand-receptor pairs and their associated signaling pathways, offering comprehensive insights into cell-cell communication mediated by secreted molecules. In each case, we also assessed whether the specific ligands and receptors were up-regulated in LPS-specific cell subsets.

Bulk RNA-seq analysis

Total RNA was extracted from PBMCs with TRIzol reagent (Invitrogen, Carlsbad, CA, USA). The RNA amount and purity of each sample was quantified using NanoDrop ND-1000 (NanoDrop, Wilmington, DE, USA). RNA library preparation was performed by using the VAHTS Universal V8 RNA-seq Library Prep Kit for Illumina (Vazyme, China). Pooled RNA-seq libraries were sequenced as paired-end reads at 2 × 150 bp sequencing (PE150) on an illumina Novaseq™ Xplus. Raw sequencing reads were processed using a standard bioinformatics pipeline. Reads were aligned to the reference genome (Homo sapiens GRCh38) using the STAR aligner. Expression levels of individual genes were quantified using featureCounts, and normalization was performed using FPKM to account for sequencing depth and gene length. Differential expression analysis was conducted using DESeq2. Genes with an adjusted p-value < 0.05 and a log2 fold change > 1 were considered differentially expressed.

Collection of publicly available transcriptomic data

We enrolled 4 datasets from Gene Expression Omnibus (GEO) for analyses. The transcriptional expression matrixs of PBMCs from 60 sepsis ALI patients and 79 sepsis controls without ALI were integrated using the sva R package from datasets GSE66890, GSE10474, and GSE32707, and were randomly divided into training and testing sets to develop the diagnostic model.

Construction of the diagnostic model and targeted drugs screen

In logistic regression analysis, GLM (Generalized Linear Model) is a supervised learning approach used to predict binary outcome variables. LASSO regression analysis was applied to optimize the initial predicition model and reduce overfitting. The predictive accuracy of the model was assessed by plotting the Receiver Operating Characteristic (ROC) curve and calculating the Area Under the Curve (AUC). Additionally, the performance and clinical utility of the predictive model were evaluated using a nomogram, decision curve analysis (DCA), and clinical impact curve (CIC). Focusing on the selected features in sepsis induced ALI patients, we utilized the Comparative Toxicogenomics Database to screen potential targeted drugs.

Application of the FPR1 Inhibitor HCH6-1 in vivo

HCH6-1(Selleck, 1435265-06-7) is a selective competitive antagonist of FPR1. HCH6-1 was dissolved in DMSO to prepare a stock solution. For the in vivo application, a working solution was prepared by mixing the DMSO stock solution with corn oil, which contained 5% DMSO and 95% corn oil. HCH6-1 were injected intraperitoneally in mice either 6 h before LPS administration or CLP, at dose of 50 mg/kg. Mice were randomly assigned to HCH6-1 treatment group and vehicle-treated group using a computer-generated sequence to ensure comparable distribution. All outcome assessments, including clinical evaluation, sample processing, and data analysis, were performed by investigators blinded to the group allocation.

Cell culture and stimulation

Mouse lung epithelial cell line (MLE-12) and murine macrophages (RAW 264.7 cells) were obtained from the Wuhan Pricella Biotechnology Co., Ltd (Wuhan, China), while HPMEC from Shanghai Quicell Biotechnology Co., Ltd (Shanghai, China). All cell lines were authenticated by short tandem repeat (STR) profiling and tested free of mycoplasma. Cells cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, USA) supplemented with 10% fetal bovine serum (FBS, Gibco, USA), and maintained in a humidified atmosphere with 5% CO2 at 37 °C. To induce an inflammatory response, MLE-12 cells were stimulated with 10 µg/mL LPS (Sigma, USA) for 24 h while HPMEC with 5 µg/mL for 24 h and RAW 264.7 cells with 1 µg/mL LPS for 24 h. Control cells were cultured with 1x PBS treatment.

Western blotting

Protein was extracted from frozen lung samples or from cultured cell lines. The protein concentrations were quantified using a bicinchoninic acid assay (Sigma, USA). After boiling for 20 min in 5× SDS-PAGE sample buffer (E153-05, GenStar), protein samples were separated on 4–20% Tris-glycine BeyoGel™ Plus PAGE precasted gel (PO469S, Beyotime) and transferred to PVDF. Membranes were blocked with protease-free rapid blocking solution (PS108P, Epizyme) over 30 min at room temperature and incubated with anti-SCGB3A2 antibody (ER61918, 1:1000, HUABIO), anti-LCN2 antibody (PB9609, 1:1000, BOSTER), anti-AKAP12 antibody (25199-1-AP, 1:500, Proteintech), anti-CCL4 antibody (A27580, 1:10000, ABclonal), anti-FPR1 antibody (PA5-140980, 1:1000, ThermoFisher), anti-β-tubulin (TA-10,1:3000, ZSGB-BIO), anti-a-Tubulin (66031-1-IG,1:100000, Proteintech), anti-Gapdh (60004-1-IG, 1:50000, Proteintech), anti-β-actin (66009-1-IG, 1:50000, Proteintech), 4 °C, overnight. The next day, following three times washing of 1× Tris-buffered saline with 0.1% Tween (TBST), membranes were incubated in secondary antibodies (Goat Anti-Mouse/Rabbit IgG(H + L), HRP Conjugated, #LF101/LF102, Epizyme) diluted with universal antibody diluent (PS119, Epizyme). HRP signal was developed by StarSignal Chemiluminescent Assay Kit. Quantification of protein levels was conducted using ImageJ.

Quantitative RT-qPCR

Total RNA was purified using MolPure^® Cell/Tissue Total RNA Kit (19221ES50, YEASEN) and transcribed to cDNA using Hifair^® Ⅱ 1 st Strand cDNA Synthesis Kit (gDNA digester plus) (11121ES60, YEASEN). Quantitative RT-PCR was performed using the 7500 Fast Real-Time PCR system, with Hieff UNICON^® qPCR SYBR Green Master Mix (11199ES03, YEASEN). The pre-designed primer pairs were used:

• IL-1β-F: TGGACCTTCCAGGATGAGGACA, R: CTCTGCAGACTCAAACTCCAC

• TNF-α-F: ATGAGCACAGAAAGCATGATC, R:TACAGGCTTGTCACTCGAATT

• IL-6-F: TACCACTTCACAAGTCGGAGGC, R:CTGCAAGTGCATCATCGTTGTTC

• CCL3-F: ACTGCCTGCTGCTTCTCCTACA, R: ATGACACCTGGCTGGGAGCAAA

• Nos2-F: GAGACAGGGAAGTCTGAAGCAC, R: CCAGCAGTAGTTGCTCCTCTTC

Immunofluorescence

Lung tissues were fixed in paraformaldehyde, embedded in paraffin, and sectioned at 5 μm thickness. The sections were deparaffinized, rehydrated, and subjected to antigen retrieval using pH 9.0 EDTA. After three washes with TBST, the sections were treated with 3% H₂O₂ for 10 min and blocked with 10% FBS at 37 °C for 30 min. Primary antibodies, including anti-CCL4/MIP-beta antibody (ab45690, Abcam, 1:100) and anti-F4/80 antibody (sc-377009, 1:100, Santa Cruz Biotechnology), were applied overnight at 4 °C. Following three TBST washes, secondary antibody goat anti-Rabbit IgG (ab205718, 1:4000, Abcam) was incubated for 30 min at room temperature. The sections were counterstained with DAPI for 10 min, and images were captured using confocal laser scanning microscopy (Zeiss#LSM710).

Statistical analysis

For all statistical tests, samples from N ≥ 3 independent experiments (biological replicates) per condition were utilized. Data were evaluated using an unpaired two-tailed Student’s t-test (for comparisons between two datasets), applying either parametric or nonparametric tests depending on the normality of the data, as assessed by the Shapiro-Wilk normality test. For targeted comparisons between groups, a Bonferroni correction was applied where appropriate. P-values < 0.05 after correction were considered statistically significant. GraphPad Prism 9 (GraphPad Software, La Jolla, CA) was used for statistical analysis. Statistical significance was considered at P < 0.05 (*P < 0.05; **P < 0.01; ***P < 0.001).

Result

Mapping changes in the cellular composition of the murine lung in sepsis-induced ALI

As illustrated in the workflow diagram, we constructed sepsis-induced acute lung injury murine model by lipopolysaccharide (LPS) intravenous injection (Fig. 1A). We observed histopathological alterations in lungs of mice, including widening of the alveolar septa, hemorrhage, and exudation and swelling in the alveolar cavity, at 12 h post-LPS administrated. The index of quantitative assessment (IQA) for lung injury was considerably higher (Fig. 1B). Additionally, the wet-dry lung weight ratio (W/D) significantly increased in sepsis-induced ALI mice (Fig. 1C). Fresh lung tissues were collected from PBS- or LPS-treated mice for single-cell isolation and sequencing. Following stringent quality control and filtering, we obtained high-quality single-cell transcriptomic profiles for 23,215 cells (8,748 in the control and 14,467 in the LPS group) (Fig. S1A). To mitigate batch effects, data from different individuals were integrated using the Harmony algorithm [32]. Unsupervised clustering of the entire cells identified 19 distinct clusters, which were defined and categorized into four major cell types based on prototypical markers: immune (Ptprc), stromal (Acta2, Dcn, Col3a1, and Col1a2), epithelial (Epcam), and endothelial cells (ECs) (Vwf, Pecam1) [33] (Fig. 1D, Fig. S1B, C). The proportions of pulmonary epithelial and endothelial cells decreased substantially, whereas an increase in immune cell infiltration was recorded. In contrast, the proportion of stromal cells remained relatively stable between the two groups (Fig. 1E, Fig. S1D, Supplementary Table 1). To evaluate the potential risk of over-integration of Harmony algorithm [34], we compared differentially expressed genes between sepsis-induced ALI and control group before and after integration, which showed a substantial overlap across all major cell types (Fig. S1E-I). These demonstrate that the data integration process did not compromise the identification of key differential molecules in subsequent analyses.

Fig. 1.

Single-cell transcriptional atlas of of lungs in sepsis-induced ALI. A Schematic workflow of the study. The diagram outlines the step-by-step procedures. Each module represents a critical phase.B Histopathological evaluation of lung tissues and IQA scores for lung injury in control and LPS-induced ALI mice.C Wet-to-dry ratio of lungs in control and LPS-induced ALI mice.D UMAP plot showing coarse annotation of major cell types in two ACCEPTED MANUSCRIPT Accepted manuscript groups.E Stacked bar plot (The upper panel) showing the proportion of major cell types in each sample. The lower box plots showing cell type proportions in two groups of mice. Each dot within the box plot represents an individual sample.F UMAP plot of single-cell data from lung tissues in control and sepsis-induced ALI samples. Cells were colored according to their annotated cell types.G Stacked bar plot showing the proportion of samples within each cell type. Each bar represents a distinct cell type, and the segments within each bar are color-coded to indicate the relative contribution of individual samples to that cell type

At a more refined level of annotation, we further subclassified the cells into more detailed categories, specifically classifying the epithelial cells into AT1 (alveolar epithelial cells type I) (Hopx, Slc39a8, and Ager) and AT2 (alveolar epithelial cells type II) (Sftpa1, Sftpb, and Sftpc), where the AT2 cells were stratified into two distinct subclusters [35]. Concurrently, endothelial cells (ECs) were categorized into arterial (Efnb2, Sox17, Sema3g, and Hey1), capillary (Car4, Ednrb, Prx, and Kdr), and a unique subset of ECs was observed exclusively in the LPS group, designated Akap12^high ECs [35, 36]. Fibroblasts (Pdgfra, Dcn, Col1a1, Col3a1, and Col6a3) constituted the major component among the stromal cells (Fig. 1F, Fig. S1J). Regarding the subpopulations of immune cells, a significant reduction in lymphocytes (Cd19, Cd20, and Cd3) and pronounced infiltration of myeloid cells were found in the lungs of the LPS-challenged group, primarily comprising Mo/Mφ and neutrophils (Fig. 1G). Neutrophils were identified using an established gene marker set (Ngp, Ltf, Ly6g, and Camp). A dramatic increase in the proportion of neutrophils was found in the lungs following LPS stimulation. Mo/Mφ were annotated via an extensive marker set, encompassing Lyz2, Cd11b, Cd163, Cd14, Cd68, Cd163, and Cd115. The Mo/Mφ cell population could be distinctly stratified into four subgroups, designated Mo/Mφ 1–4. The fraction of the Mo/Mφ 1 subset significantly increased in the lungs of patients with sepsis-induced ALI, whereas the Mo/Mφ 4 subset decreased (Fig. 1G). Although Mo/Mφ 2 showed some transcriptional similarity to the Mo/Mφ 3 population and they were positioned close to each other in the two-dimensional subspace, Mo/Mφ 3 cells were only present in the lungs of the LPS-treated mice, whereas Mo/Mφ 2 was predominantly observed in the control lungs. The pronounced specificity of Mo/Mφ 3 in the context of sepsis ALI suggests that Mo/Mφ may serve as sepsis-associated Mo/Mφ (SAMs). We also found certain distinctive immune cell types, such as megakaryocytes/platelets (MK/PLT) (Cd41, Cd61, Pf4, and Pbpb) and erythrocytes (Hbb-bt, Hbb-bs, Hba-a1, and Hba-a2) [37]. The relative abundance of these cell types did not significantly differ between the two groups.

To summarize, our initial analysis delineated the heterogeneous cellular compositions during sepsis-induced ALI at single-cell resolution. Several cell populations, such as Akap12^high endothelial cells and Mo/Mφ 3, were observed exclusively in the lungs of mice with sepsis-induced ALI, indicating their involvement in the pathogenesis of sepsis-induced ALI, which needs to be further investigated.

Sepsis-induced ALI elicits an inflammatory response primarily in Scgb3a2^high AT2 cells among epithelial cells

Alveolar epithelial cells constitute the primary interface with the external environment, rendering them critical and vulnerable targets in ALI. We found two predominant types of alveolar epithelial cells: AT1 and AT2. Two AT2 subclusters were named based on their specific expression of marker genes: Inmt^highAT2 and Scgb3a2^high AT2 cells (Fig. 2A). After administering LPS, the proportions of each type of AT cell decreased significantly, and Scgb3a2^high AT2 cells became the predominant epithelial component remaining in the lungs (Figs. 1G and 2B). The heatmap of gene set variation analysis (GSVA) revealed that AT1 cells were associated with a few biological processes from the 50-hallmark gene sets in the molecular signatures database (MSigDB), associated with the DNA repair and protein secretion signatures (Fig. 2C). Inmt^highAT2 cells were enriched in the mesenchyme-associated features of angiogenesis, coagulation, and epithelial-mesenchymal transition (EMT). EMT in alveolar epithelial cells is an important source of pulmonary fibroblasts and represents a crucial process influencing tissue injury repair and fibrotic remodeling in the lungs [38]. The sharp decrease in the proportion of Inmt^highAT2 cells after LPS stimulation suggested that the epithelial-mesenchymal imbalance caused by epithelial cell damage may be a mechanism underlying sepsis-associated ALI (Fig. 2C).

Fig. 2.

Single-cell landscape of the pulmonary epithelium in sepsis-induced acute lung injury.A UMAP plot highlighting epithelial cell populations. The plot revealed distinct three subpopulations corresponding different epithelial cell types. Feature plots showing marker genes for two AT2 cell subpopulations. The intensity of the color indicates the level of gene expression, with darker shades representing higher expression.B Pie charts representing the composition of epithelial cell types in control and sepsis-induced ALI groups. The plot displays the relative proportions and absolute number of epithelial cell subtypes.C Heatmap of GSVA pathway enrichment scores for each epithelial cell type. Inflammatory response-associated pathways are highlighted in orange.D Dendrogram of epithelial cell pseudotime trajectories. Each branch represents a distinct developmental state. Each dot represents cells in the current branching state, colored by annotated cell types. The right panel representing differential gene expression in the early stages of epithelial cell pseudotime.E KEGG pathway enrichment analysis of differentially expressed genes in Scgb3a2^high AT2 cells from sepsis-induced ALI compared to WT mice. The pathways are visualized with orange lollipops representing upregulated pathways and green representing downregulated pathways.F Volcano plot of differentially expressed genes in Scgb3a2^high AT2 cells between LPS and control groups. ACCEPTED MANUSCRIPT Accepted manuscriptG Western blot analysis showing the expression of SCGB3A2 and LCN2 protein in MLE-12 cells under different treatments. Data are presented as mean ± SEM. 95% CI of the difference: 0.08 to 0.39 (SCGB3A2), 0.20 to 0.30 (LCN2)

Scgb3a2^high AT2cells were enriched across a wide spectrum of biological pathways. These pathways included metabolism-related, immune regulation-related, and inflammation-related processes, such as ROS, IFNα, IL6/JAK3/STAT3, and the inflammatory response (Fig. 2C). In SCENIC prediction analysis, Ehf and Elf3 were predicted as specific transcription factors in Scgb3a2^high AT2 cells. Pseudotime analysis revealed that Scgb3a2^high AT2 cells function as progenitors of AT located upstream, whereas AT1 and Inmt^highAT2 cells are terminally differentiated AT cells (Fig. 2D). Through pseudotime analysis, Scgb3a2 and Lcn2 were identified as molecular markers for the early AT2 cell stage. These molecular characteristics align with the features of damage-activated AT2 cells [39, 40]. These findings collectively suggest that Scgb3a2^high AT2 cells exhibit progenitor-like properties, potentially serving as a reservoir for alveolar post-damage repair during sepsis-induced ALI through stemness maintenance and inflammatory adaptation.

We conducted a differentially expressed gene (DEG) analysis in Scgb3a2^high AT2 cells between the LPS and control groups. Scgb3a2^high AT2 cells in the LPS group presented increased enrichment of inflammatory response-related signaling pathways, including TNF signaling, chemokine activity, and bacterial infection, whereas processes associated with drug metabolism, biotransformation, and detoxification, such as cytochrome P450 and glutathione metabolism, were downregulated (Fig. 2E). Saa3, Lcn2, and Parp14 were the genes in the greatest differential upregulation of Scgb3a2^high AT2 during sepsis ALI (Fig. 2F). The recurrent identification of Lcn2 underscores its potential role in AT2 activation and orchestrating epithelial responses to inflammatory damage. In LPS-stimulated mouse lung epithelial cells (MLE-12), we observed a significant upregulation of SCGB3A2 along with the identified inflammatory marker LCN2, suggesting that SCGB3A2 may serve as a specific biomarker for inflammatory injury in pulmonary epithelium (Fig. 2G). Overall, Scgb3a2^high AT2 cells represent a crucial effector cell population involved in pulmonary inflammatory injury. The gene expression profiles of Scgb3a2^high AT2 cells in sepsis-induced ALI mice may help elucidate the dynamic regulatory mechanisms of these cells in response to inflammatory stimuli.

Akap12 as a potential biomarker for vascular injury and endothelial repair in sepsis-induced ALI

Endothelial cells are another primary injury target of sepsis-induced ALI. In this study, capillary ECs were identified in both groups, but their numbers were significantly lower in the LPS group (Fig. 3A and B, S2A, Supplementary Table 2). The expression of common inflammatory impairment markers of endothelial cells, including Vcam1, Icam1, and Nos2, significantly increased in capillary ECs treated with LPS. Additionally, the expression of the endothelial regeneration marker Cdh5 was significantly upregulated in the LPS group, which indicated that injured capillary ECs actively underwent self-repair and regeneration in the sepsis ALI group (Fig. 3C). DEG analysis revealed that the expression of interferon-stimulated genes (ISGs), including Gbp5, and several interferon-induced proteins with tetratricopeptide repeats (IFIs) was most highly upregulated in the capillary ECs of patients with sepsis-induced ALI, indicating a crucial role of interferon signaling in mediating capillary EC injuries in sepsis-induced ALI (Fig. 3D).

Fig. 3.

Single-cell transcriptional landscape of the pulmonary endothelial cells in sepsis-induced acute lung injury.A Pie charts representing the composition of endothelial cell types in control and sepsis-induced ALI groups. The plot displays the relative proportions and absolute number of endothelial cell subtypes.B Box plots comparing the average proportion of various endothelial cell types among all cells in two groups of mice.C Box plots comparing the expression of endothelial function-related genes in capillary endothelial cells between two groups of mice. Statistical analysis was performed using two-tailed unpaired Student t test.D Volcano plot of differentially expressed genes in capillary ECs between LPS and control groups.E Feature plot of Akap12 gene expression distribution and levels, split by group. The color intensity reflects the normalized expression level of the Akap12 gene.F Violin plots showing the expression of Akap12 across different cell types.G UMAP plot highlighting endothelial cell populations. H Heatmap of GSVA pathway enrichment scores for each endothelial cell type. The color scale indicates the strength of pathway enrichment, with red representing strong positive enrichment and blue representing negative enrichment

Arterial ECs were scarce in the LPS group. In contrast, a unique population of endothelial cells, Akap12^high ECs, was detected in the LPS group (Fig. 3A). Akap12^high ECs have the highest expression of the marker gene Akap12 and cannot be defined by existing EC-related gene sets (Fig. 3E and F). In the UMAP plot, Akap12^high ECs presented a similar distribution to that of arterial endothelial cells, indicating their similar transcriptomic profiles (Fig. 3G). However, GSVA of ECs revealed that several classical inflammation-related pathways, including the IFNα/γ, TNFα, IL6, and ROS pathways, were significantly enriched in Akap12^high ECs (Fig. 3H). Additionally, various proinflammatory molecules were prominently upregulated in Akap12^high ECs (Fig. 4A). By integrating and scoring common proinflammatory molecules expressed by ECs, we found that Akap12^high ECs presented the strongest inflammatory signature (Fig. 4B). The proliferation marker for ECs, Vegfa, was expressed exclusively in arterial ECs and decreased considerably in Akap12^high ECs, indicating that the proliferative capacity of Akap12^high ECs was markedly impaired. The expression of Cdh5 increased significantly in Akap12^high ECs, suggesting an increase in the regeneration of Akap12^high ECs (Fig. 4C and D). These findings indicated that Akap12^high ECs represent a unique EC subpopulation that responds to inflammatory stimuli and is involved in damage and regeneration during sepsis-induced ALI. In subsequent validation experiments, stimulation of human pulmonary microvascular endothelial cells (HPMEC) with LPS led to a significant increase in AKAP12 protein expression, confirming its response to inflammatory stimuli (Fig. 4E). However, through correlation analysis between the expression levels of Akap12 and functional markers in Akap12^high ECs, Akap12 was found to be positively correlated with the expression of the regeneration marker Cdh5. This positive correlation may reflect the protective role of AKAP12 in endothelial cell repair during sepsis-induced ALI (Fig. 4F). Hence, AKAP12 may serve as a valuable biomarker for vascular injury and endothelial repair in sepsis-induced ALI.

Fig. 4.

Akap12 as potential biomarkers for vascular injury and endothelial repair in sepsis-induced ALIA Box plots showing the expression levels of inflammatory molecules in endothelial cells across three types.B Box plots showing inflammation scores in three subpopulations of endothelial cells.C Box plots comparing the expression of endothelial proliferation markers across three distinct types of ECs.D Box plots comparing the expression of re-generation markers across three distinct types of ECs. ACCEPTED MANUSCRIPT Accepted manuscript E Western blot analysis showing the expression of AKAP12 protein in HPMEC cells under different conditions. Data are presented as mean ± SEM. 95% CI of the difference: 0.25 to 0.57. F Scatter plots showing the correlation between Akap12 expression and various endothelial functional phenotypes in Akap12high ECs

Ccl4^high Mo/Mφ that specifically infiltrate the lungs in sepsis-induced ALI exhibit a proinflammatory phenotype

Mo/Mφ constitute the predominant cellular components contribute to immune landscape heterogeneity in sepsis-induced ALI. Conventional M1/M2 classification based on established surface markers fails to capture the molecular complexity of Mo/Mφ in this pathological context. Through single-cell transcriptomic analysis, we identified four distinct Mo/Mφ subpopulations based on their signature genes: C1qc^high Mo/Mφ, Cxcr4^high Mo/Mφ, Ccl4^high Mo/Mφ, and Spp1^high Mo/Mφ (Fig. 5A and B).

Fig. 5.

Single-cell transcriptional atlas of the pulmonary Mo/Mφ in sepsis-induced acute lung injury.A UMAP plot highlighting Mo/Mφ cell populations, split by group.B Feature plots showing marker genes for four Mo/Mφ cell subpopulations, split by group. The intensity of the color indicates the level of gene expression, with darker shades representing higher expression.C UMAP plot showing the expression distribution of complement-related genes in lung cells. The distribution of expression is consistent with the localization of C1qc high Mo/Mφ.D UMAP plot showing the expression distribution of metabolism-related genes in lung cells. The distribution of expression is consistent with the localization of Spp1high Mo/Mφ.E This heatmap displays the GSVA enrichment scores for key inflammatory and metabolic pathways in various Mo/Mφ subpopulations.F Bubble plot showing the expression of M1-like and M2-like molecular features across different Mo/Mφ cell clusters. Bubble size represents the percentage of cells expressing each marker within a specific cell cluster and color indicates the expression level of each marker.G Predicted key TFs regulating distinct Mo/Mφ subsets, ranked by prediction score.H Violin plots showing the expression in markers of lung tissue-resident macrophages across different cell types of Mo/Mφ.I Pie charts representing the composition of Mo/Mφ cell types in control and sepsis-induced ALI groups. The plot displays the relative proportions and absolute number of Mo/Mφ cell subtypes. (n distribution of complementn distribution of complement) Box plots comparing the average proportion of various Mo/Mφ types among all cells in two groups of mice

C1qc^highMo/Mφ highly express several complement system-related molecules, such as C1qc, C1qb, C1qa, and C3ar1, which may indicate that this cluster is an important source of complement proteins in the lungs (Fig. 5C). C1qc^high Mo/Mφ exhibited a molecular signature of M1-like macrophages (Fig. 5F). SCENIC analysis suggested that the transcription factors Maf and Spic may regulate the signatures of C1qc^high Mo/Mφ (Fig. 5G). Secondly, Spp1^highMo/Mφ specifically overexpressed the M2 macrophage-specific receptors CD206 (MRC1) and CD68, which is consistent with the theory that SPP1 promotes the M2 polarization of macrophages (Fig. 5F). Spp1^high Mo/Mφ exhibited molecular features related to the regulation of cell metabolism, such as Fabp4 (fatty acid-binding protein 4), Lrp12 (LDL receptor protein 12), Cidec (death-inducing DFFA-like effector C), and Nrg4 (neuregulin 4) (Fig. 5D). GSVA of these cell populations revealed significant enrichment of genes related to cell metabolism (Fig. 5E). Tcf7l2 was predicted to be a regulatory TF that regulates metabolic gene signatures (Fig. 5G). Additionally, the Spp1^high Mo/Mφ subgroup possessed signatures of resident tissue macrophages. These cells highly expressed Marco and several scavenger receptor genes, Mrc1 and Olr1, which resemble alveolar macrophages (Fig. 5H). In the LPS group, the count and proportion of this cell type decreased significantly. The transition of these cell subpopulations with high metabolic activity may significantly influence the pulmonary immune microenvironment (Fig. 5I and J, S2A, Supplementary Table 2).

Additionally, Cxcr4^high Mo/Mφ and Ccl4^high Mo/Mφ subpopulations represented the most abundant Mo/Mφ subsets in the control and LPS groups, respectively (Fig. 5I). While their transcriptional profiles share some similarities, the cells from the two clusters are markedly distinct. CXCR4 has been used to distinguish monocyte-derived macrophages from tissue-resident macrophages, and the Cxcr4-CreER system has become a universal tool for studying bone marrow-derived cells [41], which suggests the bone marrow origins of these cells.GSVA revealed that Cxcr4^high Mo/Mφ subsets exhibited globally reduced activity across multiple biological pathways. However, we observed a pronounced increase in the expression of Hmgb2, a key regulator of effector T cells, in this cell population (Fig. 5H). CREM, which transcriptionally controls T-cell development, was predicted to be the key transcription factor for regulating the molecular features of this cell subgroup (Fig. 5G). These findings highlighted the significant adaptive immune regulatory potential of this cell population. In sepsis-induced ALI, the transformation of Cxcr4^high Mo/Mφ into Ccl4^high Mo/Mφ may impair the development and activation of T cells to some degree, leading to dysregulated immune responses to infections and thereby worsening sepsis-induced ALI [42].

The most remarkable feature within the Mo/Mφ landscape of sepsis-induced ALI was the Ccl4^high Mo/Mφ subpopulation, which exhibited exclusive localization in LPS-challenged lungs. This cell subset al.so presented features characteristic of the M1-like proinflammatory phenotype (Fig. 5F). GSVA revealed significant enrichment of several pathways associated with adaptive immunity, inflammation, and the stress response [43–45] in these cells (Fig. 5E). Subsequently, we integrated a set of classical proinflammatory molecules to assess the inflammatory properties of each cell. Our findings indicated that the inflammatory score (Infla score) was the highest in Ccl4^high Mo/Mφ (Fig. 6A). Importantly, Ccl4^high Mo/Mφ lacked the molecular characteristics typical of normal tissue-resident cells and exhibited a strong correlation with Cxcr4^high Mo/Mφ (Fig. 6B). Thus, Ccl4^high Mo/Mφ may represent a peripheral blood-derived, sepsis-induced ALI-specific pro-inflammatory subset. As a marker gene, the expression level of Ccl4 was significantly elevated in Ccl4^high Mo/Mφ, with further significant upregulation observed in the LPS group (Fig. 6C). Ccl4 mRNA expression levels were significantly positively correlated with the inflammatory score in each cell across Mo/Mφ (Fig. 6D). CCL4 expression was also significantly upregulated in LPS-stimulated inflammatory macrophages in vitro (Fig. 6E and F). Subsequently, we integrated three RNA-seq datasets from the GEO public database (GSE66890, GSE32707, and GSE10474), which included the transcriptomic data of PBMCs from patients with sepsis-induced ALI (n = 60) and those with sepsis without ALI (n = 79). Ccl4 expression was markedly elevated in the sepsis ALI cohort (Fig. 6G). To better characterize this unique subset of Mo/Mφ in sepsis-induced ALI, we developed another murine model of sepsis-induced ALI via cecal ligation and puncture (CLP) along with a corresponding sham-operated control group (sham). At 12 h post-surgery, we detected significant infiltration of CCL4⁺F4/80⁺ cells increased in the lungs of CLP mice (Fig. 6H). SCENIC analysis via scRNA-seq suggested that the unique molecular signature of Ccl4^high Mo/Mφ is predominantly governed by several pivotal transcription factors, including Batf, Xbp1, Rara, Interferon regulatory factors 5/7/9 (Irf5/7/9), and Nfkb2. These transcription factors are instrumental in driving inflammatory responses and fine-tuning the activities of immune cells (Fig. 5G).

Fig. 6.

Pro-inflammatory Ccl4^high Mo/Mφ specifically infiltrated in lungs of sepsis-induced ALI A Box plots showing inflammatory scores in four subpopulations of Mo/Mφ cells. ACCEPTED MANUSCRIPT Accepted manuscriptB Heatmap showing the correlation among the average transcriptional profiles of distinct Mo/Mφ cell populations.C Box plots showing the expression of Ccl4 across different cell types of Mo/Mφ and split by group (right panel).D Scatter plots showing the correlation between Ccl4 expression and inflammatory scores in each Mo/Mφ cell.E Western blot analysis showing the expression of CCL4 protein in Raw264.7 cells under different conditions.F Ccl4 mRNA levels in Raw264.7 cells under different conditions measured by RT-qPCR. 95% CI of the difference: 22.31 to 22.82.G Box plot showing CCL4 expression levels in PBMCs from sepsis patients with and without ALI. Data are presented as mean ± SEM. 95% CI of the difference: 5.83 to 6.25. Statistical analysis was performed using two-tailed unpaired Student t test. H Immunofluorescence staining of CCL4+F4/80+ cells in lung tissues of sepsis-induced ALI mice. The green color indicates F4/80, red in CCL4, and blue in DAPI staining. scale bars, 10 µm.I Dendrogram of Mo/Mφ cell pseudotime trajectories. Each branch represents a distinct developmental state. Each cells are colored by their pesudotime, developmental state, and cell type.J Expression profile of Ccl4 across developmental stages in pseudotime analysis. The x-axis denotes the pseudotime and the y-axis represents the normalized expression levels of Ccl4

Pseudotime trajectory analysis was performed among Mo/Mφ which showed distinct developmental stages among the four Mo/Mφ subpopulations. Cxcr4^high Mo/Mφ, which were predominantly in the early developmental state (State 1), may represent a population of low-differentiated, high-potency cells. C1qc^high and Spp1^high Mo/Mφ cells, which were largely in the mature state (State 2), may correspond to highly differentiated, highly functional Mo/Mφ. Moreover, Ccl4^high Mo/Mφ cells exhibited significant developmental heterogeneity, with a subset in the early state (State 1), another subset in the mature state (State 2), and a unique subset in another mature state (State 3). This trajectory from State 1 to State 3 may represent a unique developmental path of Mo/Mφ cells in response to sepsis-induced ALI (Fig. 6I). Notably, Ccl4 was identified as a DEGs in State 3 of Mo/Mφ cells along this trajectory, suggesting it may serve as markers for state transitions in Mo/Mφ cells during sepsis-induced ALI (Fig. 6J).

To summarize, we systematically delineated and characterized the heterogeneity of pulmonary monocytes and macrophages in sepsis-induced ALI. Notably, the disease-specific infiltration of Ccl4^high Mo/Mφ, which exhibit a pronounced proinflammatory phenotype, may drive tissue inflammatory damage, and CCL4 may serve as a risk factor in sepsis-induced ALI.

Identification of Fpr1as a marker gene in Ccl4^high Mo/Mφ proinflammatory subsets

Given the tight association between Ccl4^high Mo/Mφ and sepsis-induced ALI heterogeneity observed in Ccl4^high Mo/Mφ, coupled with the heterogeneity within this cellular population, we extracted and re-clustered these cells into nine distinct clusters (clusters 0–8) (Fig. 7A). GSVA revealed that the inflammatory response pathway was significantly increased in clusters 2 and 3, indicating that these clusters probably act as core proinflammatory subsets (Fig. 7B). Pseudotime trajectory analysis using the CytoTRACE and Slingshot algorithms revealed two distinct developmental trajectories in the Ccl4^high Mo/Mφ population (Fig. 7C). Both trajectories initiated and terminated with cell subpopulations enriched for high inflammatory response characteristics, whereas the intermediate states showed lower inflammatory features (Fig. 7C). Pseudotime differential gene analysis revealed stage-specific upregulation of FPR1 in both the earliest and latest developmental stages of Ccl4^high Mo/Mφ, while remaining relatively low during intermediate states. These findings suggested that FPR1 may serve as a marker for activating the inflammatory response in Ccl4^high Mo/Mφ cells during sepsis-induced ALI (Fig. 7D). Concurrently, Fpr1 expression was significantly upregulated in proinflammatory clusters (2 and 3) within Ccl4^high Mo/Mφ (Fig. 7E and F).

Fig. 7.

Identification of Fpr1as a marker gene in Ccl4^high Mo/Mφ pro-inflammatory subsets A UMAP plot showing the re-clustering ofCcl4^high Mo/Mφ. Cells are colored by clusters.B Feature plot showing GSVA enrichment scores of inflammatory response in Ccl4^high Mo/Mφ cells.C The UMAP plot (upper panel) shows the distribution of cells along two distinct trajectories in Ccl4^high Mo/Mφ cells, with arrows indicating the direction of developmental progression. The lower panel shows a landscape plot depicting the pseudotime stage of each cell subcluster in Ccl4^high Mo/Mφ.D Expression dynamics of Fpr1 along distinct developmental trajectories in pseudotime analysis.E Feature plots showing the expression of Fpr1 in Ccl4^high Mo/Mφ subpopulations. The distribution of Fpr1 expression is consistent ACCEPTED MANUSCRIPT Accepted manuscript with the localization of cells in high-inflammatory statesF Box plots showing the expression levels of Fpr1 across different cell subclusters inCcl4^high Mo/Mφ.G Western blot analysis showing the expression of FPR1 protein in Raw264.7 macrophages under different conditions. 95% CI of the difference: 0.04 to 0.38.H mRNA expression levels of pro-inflammatory markers in LPS-induced Raw264.7 macrophages. 95% CI of the difference:0.05 to 0.54 (IL-1β), 6.69 to 7.43 (TNF-α), 25.00 to 27.39 (IL-6), 10.90 to 13.48 (CCL3), 4.16 to 4.28 (Nos2)I Scatter plot showing FPR1 expression levels (FPKM) in PBMCs from sepsis patients with and without ALI. Data are presented as mean ± SEM. Statistical analysis was performed using two-tailed unpaired Student t test.J Chord diagram showing ligand-receptor interactions between epithelial cells and Mo/Mφ cells in control and sepsis-induced ALI groups

Given the potential role of FPR1 as an indicator of the inflammatory response in sepsis-induced ALI, we conducted experimental validation [46]. We found a marked increase in the protein level of FPR1 in macrophages with a proinflammatory phenotype (Fig. 7G and H). Moreover, FPR1 expression levels were significantly higher in the PBMCs of patients with sepsis-induced ALI, highlighting its potential as a biomarker and therapeutic target for sepsis-induced ALI (Fig. 7I).

Through subsequent analysis of cell-cell interactions, we found a significant increase in the frequency of Anxa1-Fpr1 interactions between alveolar epithelial cells and proinflammatory Ccl4^high Mo/Mφ in the context of sepsis-induced ALI (Fig. 7J, S2B). Under severe or continuous inflammatory stimuli, pulmonary nonimmune cells can promote substantial infiltration of inflammatory cells and thus exacerbate lung damage. The Anxa1/Fpr1 ligand-receptor pair is crucial for the recruitment and infiltration of immune cells [47, 48]. Therefore, the significant upregulation of the expression of Fpr1 on the surface of proinflammatory Mo/Mφ is highly correlated with the infiltration of specific proinflammatory cell subsets into lung tissues during sepsis-induced ALI, potentially contributing to or exacerbating tissue damage.

Inhibition of FPR1 could attenuate sepsis-induced ALI by reducing Ccl4^high Mo/Mφ infiltration

To further elucidate the role of FPR1 in sepsis-induced ALI, we intraperitoneally injected an FPR1 inhibitor (HCH6-1) or a PBS vehicle into mice in the LPS and control groups, respectively (Fig. 8A). The protein level of FPR1 was significantly elevated in the lungs of LPS-treated mice. Conversely, pretreatment with the FPR1 small-molecule antagonist led to a considerable reduction in the expression of the FPR1 protein (Fig. 8B). Our results revealed that HCH6-1 did not affect the survival of control mice but significantly prolonged the survival of LPS-induced sepsis ALI mice (Fig. 8C). Additionally, the murine sepsis score (MSS) was significantly lower in LPS mice treated with the FPR1 inhibitor compared to the MSS in those treated with PBS (Fig. 8D). Histopathological analysis of lung tissues revealed that the IQA score of lungs was decreased in LPS-treated HCH6-1-treated mice, indicating a marked improvement in lung injury imposed by HCH6-1 (Fig. 8E). Analysis of peripheral blood revealed that HCH6-1 significantly ameliorated the leukopenia caused by severe sepsis (Fig. 8F). These findings collectively suggested that FPR1 inhibition can alleviate LPS-induced sepsis ALI. In the LPS group treated with the FPR1 inhibitor, a significant increase in the absolute number and proportion of monocytes in the peripheral blood was recorded compared to those in the LPS + vehicle group, whereas the proportions of lymphocytes and granulocytes remained unchanged (Fig. 8F). These findings indicated that FPR1 levels primarily affect monocytes in the peripheral blood. A decrease in FPR1 levels may impair the recruitment of peripheral blood monocytes to lung tissues, thereby decreasing the infiltration of inflammatory monocytes and macrophages and alleviating lung injury.

Fig. 8.

Inhibition of FPR1 could attenuate sepsis-induced ALI by reducing pro-inflammatory Ccl4^high Mo/Mφ infiltration A Schematic workflow for the treatment and modeling analysis of FPR1 inhibitor in mice B Western blot analysis showing the expression of FPR1 protein in lungs from different experimental groups C Kaplan-Meier survival curves for mice in different experimental groups (D-E) The severity of sepsis-induced ALI in mice treated with and without FPR1 inhibitor assessed by murine sepsis score D, index of quantitative assessment for lung injury E F Peripheral blood profiles of four groups of mice G Kaplan-Meier survival curves for mice in different experimental groups H Histopathological evaluation of lung tissues and IQA scores for lung injury in four experimental groups of mice. Scale bar, 100 μm I Immunofluorescence staining and quantification of F4/80/CCL4 double-positive cells. Scale bars in left panel is 50 μm, right in 25 μm. Data are presented as mean ± SEM from three independent experiments. Statistical significance was assessed using one-way ANOVA

Given the limitations of the LPS model, we extended our study to the CLP model and corresponding sham-operated control groups, which were pretreated with HCH6-1 or PBS (Fig. 8A). FPR1 inhibition significantly prolonged survival and alleviated lung injury in CLP mice (Fig. 8G and H). In peripheral blood, we observed a similar trend as that in the LPS models: FPR1 inhibition partially alleviated leukopenia in CLP mice, with a significant increase in the number and proportion of monocytes compared to those in the CLP + PBS group, while the changes in lymphocytes and granulocytes were not statistically significant (Fig. S2C). To further investigate the infiltration of inflammatory cells in the lungs, we performed dual immunofluorescence staining and found that inhibition of FPR1 significantly decreased the infiltration of CCL4⁺F4/80⁺ inflammatory macrophages previously identified (Fig. 8I). These findings suggest that FPR1 inhibition alleviates ALI by reducing the infiltration of inflammatory Ccl4^high Mo/Mφ in the lungs. Our results revealed that FPR1 plays a crucial role in regulating the recruitment, migration, and differentiation of monocyte-macrophage from the peripheral blood to the lung tissue during sepsis-induced ALI. This regulation significantly affects pulmonary inflammatory responses and tissue infiltration, contributing to the pathogenesis of ALI during the acute inflammation phase. The development of therapeutic strategies targeting FPR1 may protect lung tissue from damage at the early stages of sepsis, thereby improving patient outcomes.

Construction of a diagnostic model for sepsis-induced ALI based on the molecular signature of Ccl4^high Mo/Mφ

To investigate whether single-cell findings from mouse models can be clinically translated through large-scale data validation, we identified several marker genes and specific transcription factors on Ccl4^high Mo/Mφ, which were also differentially highly expressed in State 3, a unique developmental state within the Ccl4^high Mo/Mφ. (Fig. 6I, Fig. S3A-B). As indicated in the work flowchart (Fig. 9A), we collected PBMCs from patients with sepsis-induced ALI and sepsis patients without ALI and further selected 10 candidate differentially expressed features (CCL4, SLPI, IL1B, NFKBIA, SRGN, CEBPB, IL1A, IL1RN, CCRL2, and CXCL3) and core transcription factors (XBP1 and BATF) in Ccl4^high Mo/Mφ, whose expression levels were significantly elevated in the PBMCs of the sepsis ALI group (Fig. 9B). After the three GEO datasets were integrated, the data were randomly divided into a training set (82%) and a validation set (18%). We used a generalized linear model (GLM) to develop a diagnostic model for sepsis-induced ALI, using the 12 identified genes as selected features. We performed LASSO regression to optimize the selection of genes and identified a more parsimonious subset of genes from the initial panel of 12 genes that could effectively predict sepsis-induced ALI while minimizing overfitting. We identified a five-gene panel comprising CCL4, NFKBIA, IL1B, BATF, and XBP1, which formed the basis of our optimized diagnostic model. In the internal test data, the five-gene diagnostic model performed well in terms of predictive performance, with a high AUC value of 0.857 (Fig. 9C). The clinical data were used for external validation, and the five selected genes were used to perform ROC analysis. The AUC in the external validation set was 0.867, suggesting excellent performance of the diagnostic model in sepsis-induced ALI (Fig. 9D). Decision curve analysis (DCA) revealed that, for threshold probabilities above 30%, the five-gene models provided greater clinical benefit than strategies of treating all patients or none. In the five-gene predictive model, the combined diagnostic prediction model exhibited a substantially greater net benefit than any single diagnostic gene (Fig. 9E). To further evaluate the predictive capacity of the diagnostic model, a nomogram for patients with sepsis-induced ALI was generated (Fig. 9F). The C-index of the nomogram was 0.70, and the calibration curves revealed that the incidence of ALI predicted by the nomogram based on the five-gene panel closely matched the observed incidence (Fig. 9G). We further plotted a clinical impact curve (CIC) to assess the clinical application value of the predictive model. When the risk threshold exceeded 0.4, the model’s predictions closely matched the actual values, indicating high clinical prediction efficiency (Fig. 9H). These findings highlighted the predictive value of the five-gene nomogram.

Fig. 9.

Construction of a clinical prediction model for sepsis-induced ALI based on Ccl4^high Mo/Mφ cell characteristics A The overall workflow of prediction model construction.B Heatmap showing the expression of Ccl4^high Mo/Mφ cell-specific genes in sepsis-induced ALI patients.C Receiver Operating Characteristic (ROC) curve of the prediction model constructed by logistic regression with LASSO analysis in the internal validation cohort.D ROC curve of the prediction model in our external validation dataset.E Decision Curve Analysis (DCA) curves of the prediction model.F Nomogram for predicting sepsis-induced ALI based on the constructed prediction model.G Calibration curves of the nomogram prediction.H Clinical Impact Curve (CIC) of the prediction model.I Venn diagram showing the intersection of targeted drugs for five candidate diagnostic genes in the prediction model

In the expression matrix of patients, the five genes were upregulated in samples from patients with sepsis-induced ALI. These four potential biomarkers are closely associated with inflammatory tissue damage. We subsequently used the comparative toxicogenomic database (CTD) to assess drugs that target these five genes further. According to the CTD data, antirheumatic agents, benzo(a)pyrene, and cyclosporine were identified as compounds associated with these genes (Fig. 9I). Only antirheumatic agents had a consistent influence on the five target genes, effectively downregulating their expression. This finding suggested that the application of antirheumatic agents may affect the inflammatory injury responses mediated by these target genes, conferring clinical benefits. However, the specific clinical contexts for their application, therapeutic efficacy, and underlying mechanisms need to be determined and confirmed. To summarize, our optimized five-gene diagnostic model provides a more clinically actionable tool for predicting sepsis-induced ALI, providing novel insights to improve clinical decisions and therapeutic strategies.

Discussion

In this study, we used a well-established murine model of LPS-induced sepsis ALI to obtain fresh lung tissues for single-cell RNA sequencing. The various special cell types and corresponding molecular features associated with sepsis ALI were delineated, including Scgb3a2^high AT2 (epithelial cells), Akap12^high ECs (ECs), and Ccl4^high Mo/Mφ (immune cells). We identified Scgb3a2 and Akap12 as potential biomarkers of inflammatory responses in alveolar epithelial and endothelial cells. The Mo/Mφ compartment, which constitutes the primary source of immune heterogeneity in sepsis-induced ALI, cannot be adequately characterized by the conventional M1/M2 classification [49]. Therefore, we have identified four distinct subsets based on their transcriptional profiles. Ccl4^high Mo/Mφ emerges as the predominant proinflammatory immune cells that specifically infiltrate the lungs in sepsis-induced ALI. We experimentally confirmed that proinflammatory Ccl4^high Mo/Mφ cells implicated in pathogenesis of sepsis-induced ALI in an FPR1-dependent manner. In the experimental design, we used multiple models, including LPS and CLP-induced sepsis ALI models, to overcome the limitations of a single model in accurately depicting the pathophysiological processes of the disease [50]. Subsequently, we developed a clinically applicable prediction model for sepsis-induced ALI using characteristic genes of the Ccl4^high Mo/Mφ subset, which demonstrated robust predictive performance when applied to peripheral blood samples. Our work delivers new diagnostic and therapeutic foundations for septic ALI management (Fig. 10).

Fig. 10.

Graphic abstract of our main findings in this work

Overall, the single-cell transcriptome atlas reflects pathological changes in sepsis-induced ALI to a certain extent. Our scRNA-seq profiling showed a significant reduction in epithelial and endothelial cells but greater immune cell infiltration. The reduction in alveolar epithelial cells and vascular endothelial cells highlights the vulnerability of these cell types to endotoxin-induced damage. In the ALI-induced cell niche, Scgb3a2^high AT2 cells are the primary epithelial cells that respond to inflammatory stimuli in sepsis-associated ALI. Secretoglobin (SCGB) 3A2 is a pleiotropic cytokine-like molecule known to exert anti-inflammatory and anti-fibrotic effects in various pulmonary disorders. In our study, it was identified as a key alveolar epithelial target in sepsis-induced acute lung injury, warranting further investigation into its precise mechanistic roles in septic ALI pathogenesis [51]. In endothelial cells, a unique sepsis-induced ALI-associated population comprising Akap12^high ECs exhibits activated inflammatory responses and higher regenerative potential. A-kinase anchoring protein 12 (AKAP12) plays a crucial role in maintaining vascular integrity and modulating angiogenesis and endothelial cell functions [52]. Benz et al. showed that AKAP12 was expressed at a low level in confluent endothelium but was upregulated in actively migrating endothelial cells. Endothelial cells migrate during angiogenesis and in damaged vessels to restore vascular integrity [53]. AKAP12 deficiency may compromise the migration of endothelial cells [54]. Therefore, the endothelial cells with high expression of AKAP12 in the sepsis-induced ALI group were likely to have migrated and were involved in vascular remodeling following inflammatory damage. The high expression of Akap12 suggests that it may serve as a valuable biomarker for vascular injury and endothelial repair in sepsis-induced ALI.

In the immune system, Mo/Mφ, the most abundant immune cells and the primary mediators of the innate immune system in the lungs, are closely associated with the pathogenesis of sepsis-induced ALI [55]. Monocyte/macrophage heterogeneity is popular in research on various diseases [21, 56]. Ma et al. reviewed macrophage diversity via single-cell omics and classified macrophages into seven major subtypes [55]. These subtypes reflect the functional plasticity and heterogeneity of macrophages across different tissues and disease contexts. In this study, we categorized pulmonary Mo/Mφ into four distinct subpopulations. The Ccl4^high Mo/Mφ and Cxcr4^high Mo/Mφ cell populations are key contributors to the heterogeneity of Mo/Mφ cells in the lungs of the sepsis-induced ALI and control groups, respectively. GSVA revealed that Cxcr4^high Mo/Mφ. Therefore, identifying the transcriptional differences between Ccl4^high Mo/Mφ and Cxcr4^high Mo/Mφ cell populations is crucial for understanding the molecular changes in monocytes/macrophages during sepsis-induced ALI. Ccl4^high Mo/Mφ constitute a unique proinflammatory cell population in sepsis-induced ALI. In contrast, Cxcr4^high Mo/Mφ cells, which exhibit a highly similar transcriptional profile to that of Ccl4^high Mo/Mφ cells, are predominantly found in the early stages of pseudotime development, whereas most Ccl4^high Mo/Mφ cells are positioned in the downstream stages of development. These findings suggest that these two cell populations may be highly homologous, with their distinct phenotypes and functions arising from differential responses to specific pathological stimuli. The specific marker of these unique proinflammatory Mo/Mφ cells, CCL4, can serve as a potential indicator of the transition into the inflammatory state. CCL4 (C-C chemokine ligand 4) is a key inflammatory chemokine that primarily promotes the migration and infiltration of inflammatory cells, thereby exacerbating tissue damage. Some studies have shown that serum circulatory CCL4 levels are closely related to sepsis and can serve as a proinflammatory factor for diagnosing sepsis [57, 58]. Our study revealed significant upregulation of Ccl4 mRNA levels in Ccl4^high Mo/Mφ, which were positively correlated with the proinflammatory scores of immune cells, highlighting the diagnostic potential of CCL4 in the hyperinflammatory phase of sepsis ALI. Furthermore, we extended our analysis to investigate the predictive potential of these identified biomarkers (CCL4, AKAP12, SCGB3A2) in human PBMCs. The evaluation of their individual diagnostic efficacy, particularly for CCL4 which showed a promising AUC value, suggests that the dysregulation of these markers is not confined to the lung tissue but is also reflected in the systemic circulation (Fig. S4A, B).

Through experimentation, we further revealed that the heterogeneity within Ccl4^high Mo/Mφ and identified a core proinflammatory subpopulation of Ccl4^high Mo/Mφ and their regulatory gene FPR1. FPR1 is an important pattern recognition receptor that is expressed primarily in phagocytic cells and regulates the migration of these cells. Our results showed that inhibition of FPR1 significantly reduces the infiltration of proinflammatory CCL4⁺ Mo/Mφ from the peripheral blood to the lungs in mice with sepsis-induced ALI, attenuating tissue damage and improving symptoms. When considering the therapeutic potential of FPR1 inhibition, it is informative to compare its efficacy with existing anti-inflammatory strategies. For instance, corticosteroids like dexamethasone are potent anti-inflammatory agents used in critical care. Literature reports indicate that dexamethasone treatment in similar septic models can improve survival [59]. In our model, HCH6-1 monotherapy also conferred a significant survival advantage. While a direct comparative study is needed to draw definitive conclusions on relative efficacy, the observed benefit of targeting the specific FPR1-CCL4 axis suggests a novel and mechanistically distinct approach to modulating sepsis pathophysiology, potentially complementing or offering an alternative to broad-spectrum immunosuppression. Remarkably, our findings were similar to those of Li et al., who demonstrated that FPR1-mediated inflammatory cell infiltration and tissue damage in brain injury are driven by microglial expression of FPR1 [26]. Similarly, Jun Li et al.. reported that FPR1 facilitates the migration of splenocytes to the brain, leading to higher proinflammatory cytokine release and worsening of brain injury. These findings highlighted the potential of FPR1 inhibition as a pan-protective strategy against tissue damage across different organ systems [25]. Our findings revealed that the infiltration of proinflammatory Ccl4^high Mo/Mφ cells in an FPR1-dependent manner contributes to or exacerbates tissue damage in sepsis ALI. Moreover, cell-cell interaction analysis revealed that the ANXA1-FPR1 signaling axis between epithelial cells and Ccl4^high Mo/Mφ is activated in sepsis-induced ALI. This pathway may represent a critical molecular mechanism by which injured epithelial cells recruit inflammatory macrophages, ultimately exacerbating lung tissue damage. However, but the downstream functional consequences of ANXA1-FPR1 interaction on epithelial cells and tissue remodeling are not thoroughly explored which merits further investigation. Overall, this finding not only elucidates the underlying mechanisms of sepsis ALI but also provides novel insights for the development of diagnostic and therapeutic targets in clinical settings.

Furthermore, we identified the special molecular features of Ccl4^high Mo/Mφ cells, ultimately constructing a novel five-gene optimal diagnostic prediction model for sepsis-induced ALI using publicly available datasets. Previous efforts to construct diagnostic models for sepsis-induced ALI have focused primarily on identifying DEGs in the PBMCs of patients [60, 61]. In contrast, our study provided lung tissue-specific diagnostic markers that exhibit similar expression patterns in the peripheral blood. This offers a readily accessible and tissue-specific approach for detection, bridging the gap between tissue-specific pathophysiology and clinically feasible diagnostics for sepsis-induced ALI. The model, comprising CCL4, NFKBIA, IL1B, BATF, and XBP1, showed excellent predictive performance. These four genes are significantly upregulated in the PBMCs of patients with sepsis-induced ALI. Unlike CCL4, which has proinflammatory effects, NFKBIA encodes the NF-κB inhibitor α, a key endogenous inhibitor of NF-κB signaling [62]. This protein acts by binding to NF-κB, thus inhibiting its nuclear translocation and effectively curbing the overproduction of inflammatory mediators [58]. Elevated expression of NFKBIA plays a crucial role in mitigating inflammation, highlighting its potential as a protective target in managing sepsis-induced ALI [63]. IL1B, encoding the proinflammatory cytokine IL-1β, is a key mediator of sepsis-induced inflammation and tissue injury and a proinflammatory marker [64]. BATF is a key transcription factor predicted to regulate the proinflammatory phenotype in Ccl4^high Mo/Mφ cells. As a member of the AP-1 transcription factor family, BATF can form dimeric complexes with key components of AP-1, such as the c-Jun and Fos proteins, to jointly regulate the transcription of various immune-related genes. Some studies have shown that BATF plays an essential role in the differentiation and functional regulation of dendritic cells, T cells, and B cells [65, 66]. Dysregulation of BATF expression is closely associated with immune dysregulation in multiple diseases. In 2024, Daly et al. reported that the mRNA expression level of BATF significantly increased in mouse bone marrow-derived macrophages (BMDMs) following stimulation with Lipid A, a component of bacterial endotoxins. This increase was further enhanced with higher doses of Lipid A, suggesting that BATF is a potential key factor in macrophage responses to endotoxin stimulation [67]. Additionally, a study reported that the expression level of BATF in human bone BMDMs plays a role in macrophage polarization and the regulation of inflammatory responses [68]. In our study, the expression of Batf was significantly upregulated in proinflammatory Mo/Mφ cells, suggesting its involvement in the phenotypic regulation and proinflammatory processes of Mo/Mφ cells during sepsis-associated ALI. These findings indicated that BATF is closely related to immune-inflammatory regulation and the pathogenesis of sepsis-induced ALI. XBP1 is a key transcription factor in the endoplasmic reticulum stress response (ERS), regulating cellular protein homeostasis and inflammatory responses through the IRE1α-XBP1 signaling pathway [69]. The genes regulated by XBP1 are involved in various cellular processes, including the endoplasmic reticulum stress response, secretory function, lipid metabolism, glucose homeostasis, and inflammatory responses. Some studies have shown that XBP1 enhances the release of proinflammatory cytokines in macrophages and is closely associated with proinflammatory immune responses [70]. The role of high levels of XBP1 in the proinflammatory effects of Ccl4^high Mo/Mφ on sepsis-associated ALI necessitates further investigation. Overall, the five-gene diagnostic model provides a readily detectable diagnostic approach for sepsis-associated ALI. Previous multi-omics studies investigating biomarkers and pathogenic molecules in sepsis patients have been limited by the challenges of obtaining clinical tissue samples, relying primarily on peripheral blood mononuclear cells. In contrast, our study integrates clinical peripheral blood samples with lung tissues from well-established murine models of sepsis-induced lung injury. This approach has enabled us to identify clinically detectable biomarkers with lung tissue-specific pathological relevance, which could significantly improve future patient diagnosis. However, we acknowledge that the sample size of the initial training cohort poses a potential risk of overfitting. Future studies with larger training samples will be valuable to further refine and solidify the model. In addition, the applicability of our predictive model across different populations or etiologies of sepsis remains to be fully elucidated, also necessitating further validation with an expanded sample size and stratified analyses as well.

Additionally, the CTD database revealed that antirheumatic drugs can target al.l four of these molecules and reduce their expression levels. Corticosteroids and nonsteroidal anti-inflammatory drugs (NSAIDs), which are the main agents used in the treatment of rheumatic diseases, are widely used in the management of sepsis [71, 72]. However, a recent cohort study showed that biological disease-modifying antirheumatic drugs (bDMARDs) reduce the risk of sepsis progression and mortality after severe infections in patients with rheumatoid arthritis [73]. The use of DMARDs in sepsis-induced ALI remains uninvestigated. Our findings suggest that investigating the role of DMARDs in ALI may provide new therapeutic insights, necessitating further experimental validation.

In this study, we identified several novel specific biomarkers for sepsis-induced ALI and primary validation were performed in murine lung tissues and cell lines. However, it does not include mechanistic loss-of-function or gain-of-function experiments to establish direct causality. Each of these promising candidates warrants independent, in-depth functional investigation. In fact, based on these findings, we have already commenced follow-up studies using molecular and genetic tools to elucidate their precise roles in the pathogenesis of sepsis-induced ALI, which will be the focus of a future publication. In addition, although we provided supporting evidence from human PBMCs, direct validation in lung tissues from sepsis-induced ALI patients was difficult due to the significant ethical and practical challenges associated with obtaining samples. This study has additional limitation which the initial translational verification using our local clinical cohort was limited by a small sample size (n = 16). While this cohort served its primary purpose of bridging our murine findings to human disease and the key signals were independently validated in a large database, future prospective studies with larger, multi-center patient populations are warranted to further solidify the clinical relevance and generalizability of our proposed mechanism across diverse clinical scenarios. Additionally, we proposed a novel therapeutic method based on sequencing analysis but did not assess the safety, efficacy, or overall benefits of these drugs in animal models or cell lines. Further studies are needed to investigate.

Conclusions

Overall, through systematic transcriptomic analysis, we delineated the altered molecular landscape in septic ALI lungs, identifying SCGB3A2 and AKAP12 as key biomarkers of inflammatory responses in alveolar epithelial and endothelial cells. Additionally, we identified a novel population of disease-specific Ccl4^high Mo/Mφ exhibiting proinflammatory transcriptional signatures, closely associated with pathogenesis of sepsis-induced ALI. During sepsis-induced ALI, the core proinflammatory subcluster within Ccl4^high Mo/Mφ mediates lung tissue damage through FPR1-dependent mechanisms. Pharmacological inhibition of FPR1 could reduce pulmonary infiltration of Ccl4^high Mo/Mφ and ameliorate sepsis-induced ALI. Furthermore, leveraging Ccl4^high Mo/Mφ-derived molecular signatures, we constructed a five-gene diagnostic prediction model (CCL4, NFKBIA, IL1B, BATF, and XBP1). This model demonstrates high detectability and robust diagnostic performance for predicting sepsis-induced ALI from peripheral blood samples (Fig. 10).

Abbreviations

ALI

Acute lung injury

ARDS

Acute respiratory distress syndrome

ICU

Intensive care units

scRNA-seq

Single-cell RNA sequencing

Mo/Mφ

Monocytes/macrophages

LPS

Lipopolysaccharide

IQA

Index of quantitative assessment

W/D ratio

Wet-to-dry lung weight ratio

ECs

Endothelial cells

SAMs

Sepsis-associated Mo/Mφ

MK/PLT

Megakaryocytes/platelets

GSVA

Gene set variation analysis

DEGs

Differential genes expression

ISGs

Interferon-stimulated genes

TAMs

Tumor-associated macrophages

CLP

Cecal ligation and puncture

Sham

Sham-operated

TFs

Transcriptional factors

DCA

Decision curve analysis

CIC

Clinical impact curve

CTD

Comparative toxicogenomic database

Authors’ contributions

Q.Z.,Y.Z., J.Y., Y.L., and J.S. conceived and designed the study; Y.L., J.S., T.Y., and M.P. carried out the studies; J.T., Y.M., X.L., and H.L. acquired and analyzed the data; L.G., S.D. contributed to methodology; Y.L., J.S., T.Y., and M.P. drafted the manuscript; Q.Z.,Y.Z., and J.Y. edited the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by National Key Research and Development Program of China (No. 2023YFC3605200) and Major Research Plan of National Natural Science Foundation of China (No.92163213) for Qiang Zhang, and National Natural Science Foundation of China (No. 82372154) for Jianbo Yu. This project was also funded by Tianjin Key Medical Discipline Construction Project (No. TJYXZDXK-3-013B).

Data availability

The raw data of single-cell RNA-seq generated in this study have been deposited in China National Center for Bioinformation with project number PRJCA037564.

Declarations

Ethics approval and consent to participate

The animal study was approved by the Animal Ethical and Welfare Committee of Tianjin Nankai Hospital, Tianjin Medical University (Approval No. NKYY_DWLL_2023_077). Clinical trial number: not applicable. The collection of PBMC samples from patients was approved by the Institutional Review Board of Tianjin Nankai Hospital (Approval No. NKYY_YXKT_IRB_2024_072_01).

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.