Neutrophils with low production of reactive oxygen species are activated during immune priming and promote development of arthritis

Abstract

Rheumatoid arthritis (RA) is an inflammatory autoimmune disease mediated by immune cell dysfunction for which there is no universally effective prevention and treatment strategy. As primary effector cells, neutrophils are important in the inflammatory joint attack during the development of RA. Here, we used single-cell sequencing technology to thoroughly analyze the phenotypic characteristics of bone marrow-derived neutrophils in type II collagen (COL2)-induced arthritis (CIA) models, including mice primed and boosted with COL2. We identified a subpopulation of neutrophils with high expression of neutrophil cytoplasmic factor 1 (NCF1) in primed mice, accompanied by a characteristic reactive oxygen species (ROS) response, and a decrease in Ncf1 expression in boosted mice with the onset of arthritis. Furthermore, we found that after ROS reduction, arthritis occurred in primed mice but was attenuated in boosted mice. This bidirectional effect of ROS suggested a protective role of ROS during immune priming. Mechanistically, we combined functional assays and metabolomics identifying Ncf1-deficient neutrophils with enhanced migration, chemotactic receptor CXCR2 expression, inflammatory cytokine secretion, and Th1/Th17 differentiation. This alteration was mainly due to the metabolic reprogramming of Ncf1-deficient neutrophils from an energy supply pathway dominated by gluconeogenesis to an inflammatory immune pathway associated with the metabolism of histidine, glycine, serine, and threonine signaling, which in turn induced arthritis. In conclusion, we have systematically identified the functional and inflammatory phenotypic characteristics of neutrophils under ROS regulation, which provides a theoretical basis for understanding the pathogenesis of RA, to further improve prevention strategies and identify novel therapeutic targets.

1. Introduction

Rheumatoid arthritis (RA) is a chronic inflammatory autoimmune disease characterized by synovial hyperplasia, bone destruction, and joint deformity [[1], [2], [3]]. Early diagnosis and interventions are imperative strategies in the clinical management of RA. Exploring the specific mechanisms of early immune cell dysfunction in RA is of great scientific importance and clinical value to elucidate the inflammatory mechanisms and to screen targets for intervention.

Neutrophils have been recognized as critical players in the onset and progression of RA. They are the most abundant circulating leukocytes in humans with a robust innate immune response. The pathogenic role of neutrophils in RA lies in the alteration of several processes that perpetuate inflammation and lead to cartilage and bone destruction in affected joints, such as increased cell survival and migratory capacity, abnormal inflammatory activity, elevated reactive oxygen species (ROS) production and an exacerbated release of neutrophil extracellular traps (NETs) [[4], [5], [6], [7], [8]]. In RA patients, we found that neutrophils have a more pro-inflammatory phenotype than in healthy individuals, accompanied by significant ROS features [9].

The multi-component enzyme NADPH oxidases 2 (NOX2) is the major source of ROS production and is widely distributed on neutrophils. As one of the critical subunits of NOX2 complex, neutrophil cytoplasmic factor 1 (NCF1) exerts a central genetic control of arthritis by regulating the physiological levels of ROS [10,11], which regulates processes such as cell signaling, proliferation, differentiation, and redox homeostasis. NCF1 and the NOX2 complex are abundantly expressed in neutrophils. They function as the primary ROS producer during inflammatory responses [12]. Disease-causing Ncf1 mutation with low ROS production promotes the development of autoimmune diseases such as RA, systemic lupus erythematosus, and Sjögren's syndrome and their corresponding animal models [10,11,[13], [14], [15]], by mechanisms that include decreased clearance of apoptotic cells [16], downregulation of NETs formation [17], activation of autoreactive T cells [18], and enhanced type I IFN responses [15]. Transcriptomic analyses revealed that neutrophils were associated with the disease process of RA with increased NCF1 expression and enhanced intrinsic oxidative stress, compared with healthy subjects [9]. In collagen II (COL2) induced arthritis (CIA), mice with a mutation of the Ncf1 gene had more severe arthritis due to a break of tolerance and activation of self-reactive T cells [19]. This effect has been attributed to enhanced antigen presentation by macrophages and dendritic cells, but an additional role of neutrophils may contribute [20,21]. Therefore, we wanted to find out whether a possible bidirectional effect of ROS is related to the function of the neutrophils or the NOX2 complex in neutrophils, expecting to provide new potential targets for RA treatment.

Here, we used single-cell transcriptome sequencing (scRNA-seq) data to characterize neutrophil phenotypic changes in early CIA (primed state) compared to late CIA (boosted) and normal (naive) mice. Combining the analysis results, we characterized ROS regulation of neutrophil metabolic reprogramming, functional changes, and inflammatory phenotype shifts based on the presence or absence of NCF1-derived ROS. The present study's neutrophil phenotyping results identified specific features of neutrophil dysregulation during the initiation of arthritis and suggested new potential targets for immunophenotyping-based prevention strategies.

2. Materials and methods

2.1. Mouse strains

DBA1/J mice were purchased from Beijing Hua Fukang Biotechnology Co. Ltd. (Beijing, China). C57BL/6J WT and Ncf1 knockout (C57BL/6JCya-Ncf1^em1, here denoted as Ncf1^−/−) mice were purchased from Cyagen Biosciences Inc. (Jiangsu, China). C57BL/6J Ly6g-Cre-2A-tdTomato mice were purchased from Southern Biotechnology. All mice were bred, housed, and treated at a pathogen-free facility at Sichuan University. All experiments used age and sex-matched cohorts of females and males (8 weeks old) littermates. Animal experiments were approved by West China Hospital, Sichuan University Animal Ethics Committee (Approval number 20220128002).

2.2. Isolation of human and murine neutrophils

Human neutrophils were freshly prepared at room temperature from healthy donors after obtaining informed consent. Blood was loaded to the Histopaque-1077 at a ratio of 1:1 in volume in a tube, and was centrifuged at 500×g for 30 min with density gradient sedimentation [22]. Neutrophils were collected between the third and fourth layers. Erythrocyte lysis was carried out for 20 s in 36 mL of sterile distilled water, followed by adding 4 mL of 10 × PBS to neutralize and then centrifuged at 300×g for 5 min at 4 °C. After centrifugation of the cells with 1 × PBS and complete erythrocyte lysis again, follow the previous method [23]. Next, 1 × PBS was used to wash cells twice, and resuspend cells in RPMI-1640 media containing 10 % FBS. The purity of neutrophils was about 90 %, as determined by flow cytometry (FCM). Cells were counted, and 1 × 10⁶ cells were used for subsequent experiments.

Bone marrow (BM) cells were extracted from the hind limbs of mice, filtered with a 40 μm strainer, and washed with cold 1 × PBS at 300×g for 5 min at 4 °C. Then, cells were resuspended in 3 mL RPMI-1640 media containing 10 % FBS. In a 15 mL tube, 3 mL of Histopaque-1119 was added on the bottom, and then 3 mL of Histopaque-1077 was carefully layered on top. 3 mL cell suspensions were carefully added on top of Histopaque-1077, then centrifuged at 500×g for 30 min at room temperature without braking [24]. Neutrophils were collected at the interface between the Histopaque 1119 and 1077 layers and washed twice in cold 1 × PBS by centrifuging at 500×g at 4 °C for 5 min and resuspending in 1 mL RPMI-1640 media containing 10 % FBS. The purity of neutrophil isolations was routinely above 90 %. Cells were counted, and 1 × 10⁶ cells were used in each experiment.

2.3. Preparation of BM single-cell suspensions

BM cells were separately extracted from controls (naive), primed (immunized once with COL2 at D0), and boosted (immunized twice with COL2 at D0 and D42 (n = 3/group). Single-cell suspensions were obtained by pipetting and passing through a 40 μm cell strainer. After centrifugation at 300×g for 5 min, the cell pellets were lysed by incubation in 1 × RBC lysis buffer (eBioscience, San Diego, CA) for 5 min. After centrifugation, the cell pellets were resuspended in 1 % BSA in 1 × PBS. Cell viability was assessed with the trypan blue exclusion method, and only single-cell suspensions with >85 % viability were processed for further sequencing.

2.4. Collagen-induced arthritis (CIA)

DBA1/J (A^q, MHC-class II molecule) mice are susceptible to CIA, and are commonly used as RA models, showing pathological changes such as synovitis, cartilage and bone erosion, similar to the clinicopathological changes of human RA [25]. The CIA mouse model was prepared according to the standard protocols described previously [26]. 8-week-old male DBA1/J mice were immunized intradermally at the base of the tail with 100 μg chicken COL2 (Chondrex) emulsified in complete Freund's adjuvant (CFA, Sigma-Aldrich). On D21, a booster injection of 100 μg chicken COL2 (Chondrex) emulsified with incomplete Freund's adjuvant (IFA, Sigma-Aldrich) was administered. The stages of inflammation in the CIA were further divided into the primed stage (COL2 immunized once, early CIA) and the boosted stage (COL2 immunized twice, late CIA). For the late CIA intervention, mice were injected intraperitoneally (IP) with di-phenylene-iodonium (DPI, 100 ng/kg) or vehicle (10 % DMSO) daily until the experiment endpoint, starting on D21 when mice received their second immunization. For the early CIA intervention, mice were injected with different ROS inhibitors (i.p., 100 ng/kg/day DPI, 1 mg/kg/day GSK2795039 and 1 mg/kg/day MitoQ10), Anti-Ly6G (i.p., 100 μg/mouse, on day −2, 0, +2, and +4) or vehicle (i.p., 10 % DMSO/kg/day) until the experiment endpoint (D42), starting on D0 after when mice receive their first immunization, during which mice do not receive a second immunization with COL2. Arthritis was also elicited in C57BL/6J WT and Ncf1^−/− mice immunized with 100 μg of chicken COL2 emulsified 1:1 with CFA, 4 weeks after the first immunization (D28), mice received intraperitoneal injection of lipopolysaccharide (LPS) (25 μg/mouse) to induce arthritis.

The development of clinical arthritis was monitored every other day by visual scoring of the paws using a semi-quantitative scoring system (0–4 per paw; maximum score of 16) as previously described [27]. Cells from different organs such as BM, peripheral blood (PB), joint, and spleen were collected for FCM, and the hind limbs were collected for histopathological assessments as detailed below.

2.5. Bioluminescence imaging

Animals were anesthetized with isoflurane and maintained with 2 % isoflurane during imaging procedures. Each mouse was given an intraperitoneal injection with luminol (20 mg/100 g body weight, Sigma Aldrich) to detect mainly superoxide anion produced by human neutrophils [28]. Images were captured 10–15 min after the luminol injection. Bioluminescence intensity was quantified using the Living Image® advanced in vivo imaging software.

2.6. Micro-computed tomography (micro-CT) analysis

Hind paws were fixed in 10 % neutral-buffered formalin for 2 days at 4 °C and washed with 1 × PBS. Radiographs of the femora and tibiae were taken by soft x-ray (model CMB-2; SOFTEX Co. Ltd.). Three-dimensional images of posterior paws were obtained by micro-CT scanning (SMX-90CT; Shimadzu, Japan). Bone morphometric parameters and bone mineral density were analyzed using CTan software (SkyScan).

2.7. Histopathological and immunohistochemical analysis

Hind paws were fixed in 10 % neutral-buffered formalin, decalcified, and embedded in paraffin. Each section (4 μm) was stained with hematoxylin/eosin (HE) or Safranin O/Fast Green for histological scoring and assessment of cartilage morphology. Myeloperoxidase (MPO) was detected using the anti-MPO primary antibody and horseradish peroxidase conjugated goat anti-rabbit IgG secondary antibody for immunohistochemistry. Negative control included the absence of primary antibodies. The MPO-positive signals in histological and immunohistochemical images were quantified from randomly selected sections of at least five fields for each sample using Image J 1.52 software (NIH, Bethesda, MD).

2.8. Single-cell RNA sequencing and raw data preprocessing

The scRNA-seq libraries were constructed using the GEXSCOPE Single-cell RNA Library Kit according to the manufacturer's original protocol (Singleron Biotechnologies, Nanjing, China). Briefly, A cell suspension at a concentration of 1.5–5.0 × 10⁵ cells/mL was immediately loaded into scope-microfluidic chips to obtain the isolation of single cells, and the mRNA capture and labeling of single cells were completed by using Single Cell Barcode Beads. The mRNA captured by Barcode Beads was reverse transcribed into cDNA, and cDNA was amplified to construct cDNA libraries. Libraries were sequenced on the Illumina HiSeq X platform (Illumina, USA) with paired-end 150 bp reads.

Raw gene expression matrix data were generated by the CeleScope (https://github.com/singleron-RD/CeleScope) pipeline. Raw data for each sample were assessed for quality with FASTQC (version 0.11.7) and filtered with Cutadapt (version 1.17). Reads were aligned to the GRCm38/mm10 mouse genome reference using STAR (version 020201). Gene expression was subsequently estimated using FeatureCounts (version v1.6.2). Downstream bioinformatic analysis was performed in R using the package Seurat (v3.1.2).

2.9. Quality control, dimension-reduction, and clustering

Seurat v 3.1.2 was used for quality control, dimensionality reduction, and clustering. The expression matrix was filtered using the following criteria for each sample dataset: 1) Cells with less than 200 genes or with a gene count in the top 2 % were excluded; 2) Cells with a UMI count in the top 2 % were excluded; 3) Percentage of mitochondrial reads was determined with PercentageFeatureSet function with pattern = "^mt" -parameter. Cells with a mitochondrial content >20 % were excluded; 4) Genes expressed in less than 5 cells were excluded. After filtering, the gene expression matrix was normalized and scaled using the functions NormalizeData and ScaleData. The top 2000 variable genes were selected for the PCA analysis using FindVariableFeatures. Using the top 20 principal components and a resolution parameter of 1.2, the cells were separated into 10 clusters using FindClusters. Cell clusters were visualized with Seurat functions RunUMAP using Uniform Manifold Approximation and Projection (UMAP).

2.10. Differentially expressed genes (DEGs) analysis

To identify DEGs, we used the Seurat FindMarkers function based on the Wilcoxon rank-sum test with default parameters, selecting DEGs expressed in >10 % of cells in both compared cell groups and with an average log (Fold Change, FC) value > 0.25. The adjusted p-value was calculated using Bonferroni correction, with 0.05 as the statistical significance criterion.

2.11. Pathway enrichment analysis

To investigate the potential functions of G0-G4 clusters, GO analysis was used with the “clusterProfiler” R package v 3.16.1. Paths with a p_adj value of less than 0.05 were considered to be significantly enriched. Selected significant pathways were plotted as bar plots. Gene Ontology (GO) gene sets, including the categories of Molecular Function (MF), Biological Process (BP), and Cellular Component (CC), were used as a reference.

2.12. Cell-type recognition with Cell-ID

Cell-ID is a multivariate approach, extracting the genetic signature for each cell and performing cell identity recognition by hypergeometric testing (HGT) [29]. Dimensionality reduction, in which cells and genes are projected into the same low-dimensional space, was performed on the normalized gene expression matrix using multiple correspondence analysis. Next, the genes in each cell were ranked in order to identify the most distinctive genes in the cell. HGT was carried out on these sets of genes against the BM-specific reference from the SynEcoSys database, containing all genes expressed by the cell type in the specific tissues [30]. The identity of each cell was determined as the cell type with the lowest p-value of the HGT. To annotate clusters, the frequency of each cell type in each cluster was calculated and the cell type with the highest frequency was selected to identify the cluster.

Expression of canonical markers from the reference database SynEcoSysTM (Singleron Biotechnology) was used to identify the cell type of each cluster. SynEcoSysTM includes collections of canonical cell type markers for single-cell seq data from CellMakerDB, PanglaoDB and recently published literature [[31], [32], [33], [34], [35], [36]].

2.13. Subtyping of major cell types

To obtain a high-resolution map of Neutrophils, Mononuclear phagocytes (MPs), Granulocytes-monocyte progenitor cells (GMP), B cells, Erythroid progenitor cells (ProEryth), Hematopoietic stem cells (HSCs), T cells, Plasma Cells, Basophils, Erythroblasts from the specific cluster were extracted and re-clustered for detailed analysis following the same procedures described above and by setting the clustering resolution as 0.8.

Previous studies revealed various BM neutrophil subpopulations arising during differentiation and maturation [31,36]. We calculated the fraction of each scRNA-seq-defined cluster in the nine samples. Each cell in these samples was annotated using the above cell label transfer method. The cluster identity of each cell was inferred based on the transcriptomic similarity between this cell and the reference clusters (G0–G4) defined.

2.14. Pseudotime trajectory analysis: monocle2

Cell differentiation trajectory of monocyte subtypes was reconstructed using Monocle2 v 2.10.0 [37]. Seurat (v3.1.2) FindVariableFeatures selected the top 2000 highly variable genes, and DDRTree performed dimension reduction for trajectory construction. The trajectory was visualized by the plot_cell_trajectory function in Monocle2.

2.15. Measurement of ROS levels in neutrophils

DCF-DA is a versatile fluorescent probe for the detection of total of ROS (O₂^•-, H₂O₂ and ^•OH) generated in cells [38]. ROS production by neutrophils was detected with a Cellular ROS Detection Assay Kit (Sigma) following the manufacturer's manual. Neutrophils were isolated from PB or BM and diluted to the concentration of 1 × 10⁶ cells in RPMI-1640 medium supplemented with 10 % FBS. Pretreatment with vehicle (DMSO) or ROS inhibitors (DPI, GSK2795039, and MitoQ10) was carried out for the negative controls at 37 °C for 30 min. 100 μM tert-butyl hydroperoxide (TBHP) was used as the positive control. Neutrophils were incubated with 10 μM DCFH-DA for 30 min and then stimulated with 100 nM phorbol myristate acetate (PMA) (Sigma) or 100 nM N-formyl-methionyl-leucyl-phenylalanine (fMLF, Sigma) for 30 min and positive control (TBHP, Sigma) for 2 h. ROS level was determined based on the green fluorescence intensity of neutrophils gated on FSC/SSC using FCM.

2.16. Measurement of intracellular Ca²⁺ in neutrophils

Changes in intracellular Ca²⁺ were monitored using Fluo-3AM (Invitrogen) according to the manufacturer's directions. Briefly, neutrophils were suspended at a concentration of 1 × 10⁶ cells/mL and treated with or without ROS inhibitors for 30 min. Then, neutrophils were incubated with 4 mM Fluo-3AM for 20 min at 37 °C. Following a 30 s baseline recording, cells were challenged for other times with 100 nM fMLF or 100 nM PMA. The fluorescence intensities of the Fluo-3AM probes were measured via FCM.

2.17. Phagocytosis assay

Neutrophils at a concentration of 1 × 10⁶ cells/mL were incubated with E. coli expressing mCherry for 30 min at multiplicity of infection (MOI) 1:10 or 1:50. Cells were then washed with 1 × PBS for 3 times before being subjected to FCM (fluorescence-activated cell sorting) [39,40]. Phagocytosis ability was expressed as the percentage of neutrophils phagocytosing mCherry-E. coli.

2.18. Migration assays

Neutrophils were purified as previously described and resuspended in serum-free media at a concentration of 1 × 10⁶ cells/mL. Neutrophils were pre-incubated with or without ROS inhibitors for 30 min. Then, 200 μL of neutrophils were added to the top chamber of the transwell insert (3 μm pore size, Corning Costar, UK), and 600 μL of 20 ng/mL IL-8 or 100 nM fMLF was added to the bottom chamber. Neutrophils were incubated at 37 °C for 1 h. Neutrophils in the bottom chamber were collected and counted at least 3 times using an automatic cell counter. Data were expressed as mean ± SEM from at least 3 independent experiments.

2.19. Trans-endothelial migration assays

2 × 10⁴ human umbilical vein endothelial cells (HUVECs) were cultivated in the transwell chamber (Corning) and cultured in a complete medium until the cells reached 100 % confluence. HUVECs confluence was calculated using the IncuCyte ZOOM software. 1 × 10⁶ neutrophils were treated with or without ROS inhibitors for 30 min and then added to the upper chamber of transwells and allowed to migrate in the absence or presence of 20 ng/mL IL-8 or 100 nM fMLF in the lower chamber for 2 h. Cell counts were performed on neutrophils that had migrated into the lower chamber using an automatic cell counter at least 3 times. Data were expressed as mean ± SEM from at least 3 independent experiments.

2.20. Neutrophils adhesion assays

HUVECs were seeded in 96-well plates and allowed to proliferate to confluence, and then stimulated with 10 ng/mL TNF-α at 37 °C for 4 h. Neutrophils were pre-incubated with or without ROS inhibitors for 30 min and loaded to settle and adhere to HUVECs. After 60 min, the number of adhered neutrophils was counted by giemsa stain. Data were expressed as mean ± SEM from at least 3 independent experiments.

2.21. Neutrophil morphology

Neutrophils pre-incubated with or without ROS inhibitors for 30 min were stimulated with 100 nM PMA for 30 min, then loaded onto cytospin chambers with slides (Thermo Scientifics) and spun at 2000 rpm for 10 min. Giemsa stain of neutrophils was accomplished to visualize the neutrophils’ nucleus morphology under light microscopy.

2.22. Real-time confocal imaging

Neutrophils (Ly6G-tdTomato⁺, red) were pre-incubated with or without ROS inhibitors for 30 min, then stained with Hoechst 33342 (blue) for 10 min. Samples were added at 100 nM PMA concentration, and confocal images were obtained from 0 to 30 min using Andor Dragonfly 200 on a Nikon Ti-E microscope. Data were analyzed using Imaris Software (Oxford Instruments).

2.23. Neutrophils-mediated T-cell proliferation and differentiation assays

Neutrophils pre-incubated with or without ROS inhibitors for 30 min. After that, neutrophils were added to PB or lymph-node cells (1:1 ratio) in the presence of vehicle or 100 nM fMLF. All cells were co-cultured for 5 days at 37 °C in a cell incubator. Cells were harvested, and detected the proportions of CD3⁺ CD4⁺ T cells via FCM analysis. Next, CD3⁺ CD4⁺ T cells were gated to analyze populations of T cell subsets (Th1, Th2, Th17, and Treg cells).

2.24. Flow cytometry (FCM)

Cells were treated and cultivated as indicated in the corresponding section. 7-ADD was added to cells before analysis for live/dead cell discrimination. Then, cells were stained for fluorescently conjugated surface markers for macrophages (CD11b⁺ F4/80⁺), neutrophils (CD11b⁺ Ly6G⁺), and monocytes (CD11b⁺ Ly6C⁺). To analyze CD4⁺ T cells polarization (Th1, Th2, Th17, and Treg cells), surface staining was performed with fluorescence-conjugated MAbs to CD3, CD4, and CD25. Cells were then fixed, permeabilized using an intracellular staining kit (eBioscience), and incubated at room temperature with fluorescence-conjugated MAbs to Foxp3, IFN-γ, IL-4, and IL-17A. Samples were processed further for FCM analyses.

2.25. RNA extraction and qPCR

The neutrophils were collected, and total RNA was isolated by RNeasy® Mini Kit (Qiagen), where cells were directly sorted into 350 μL of RLT Plus buffer and RNA extracted per the manufacturer's protocol. Concentration was measured using the NanoDrop (Thermo Fisher Scientific, USA). 1 μg RNA per sample was used for cDNA synthesis using the iScript cDNA Synthesis Kit (Bio-Rad). qPCR was performed using iQ SYBR Green Supermix (Bio-Rad) in a CFX96 Real-Time PCR Detection System (Bio-Rad). Primer sequences were taken from the PrimerBank PCR primer public resource. The following forward/reverse primer pairs were used.

Ncf1	5′-GCTGACTACGAGAAGAGTTCGG-3′ 5′-CCTCGCTTTGTCTTCATCTGGC-3′
Cxcr1	5′-CCATTCCGTTCTGGTACAGTCTG-3′ 5′-GTAGCAGACCAGCATAGTGAGC-3′
Cxcr2	5′-CTCTATTCTGCCAGATGCTGTCC-3′ 5′-ACAAGGCTCAGCAGAGTCACCA-3′
Cxcr4	5′-GACTGGCATAGTCGGCAATGGA-3′ 5′-CAAAGAGGAGGTCAGCCACTGA-3′
Tnf-α	5′-GGTGCCTATGTCTCAGCCTCTT-3′ 5′-GCCATAGAACTGATGAGAGGGAG-3′
Ifn-γ	5′-CAGCAACAGCAAGGCGAAAAAGG-3′ 5′-TTTCCGCTTCCTGAGGCTGGAT-3′
Il-1β	5′-TGGACCTTCCAGGATGAGGACA-3′ 5′- GTTCATCTCGGAGCCTGTAGTG-3′
Il-6	5′-TACCACTTCACAAGTCGGAGGC-3′ 5′-CTGCAAGTGCATCATCGTTGTTC-3′
Il-17a	5′-CAGACTACCTCAACCGTTCCAC-3′ 5′-TCCAGCTTTCCCTCCGCATTGA-3′
Gapdh	5′-CATCACTGCCACCCAGAAGACTG-3′ 5′-ATGCCAGTGAGCTTCCCGTTCAG-3′
β-ACTIN	5′-CACCATTGGCAATGAGCGGTTC-3′ 5′-AGGTCTTTGCGGATGTCCACGT-3′
CXCR1	5′-TCCTTTTCCGCCAGGCTTACCA-3′ 5′-GGCACGATGAAGCCAAAGGTGT-3′
CXCR2	5′-TCCGTCACTGATGTCTACCTGC-3′ 5′-ACACACTTGGCGGTTCCTTCGA-3′
IFN-γ	5′-GAGTGTGGAGACCATCAAGGAAG-3′ 5′-TGCTTTGCGTTGGACATTCAAGTC-3′
CXCL1	5′-AGCTTGCCTCAATCCTGCATCC-3′ 5′-TCCTTCAGGAACAGCCACCAGT-3′
CXCL2	5′-GGCAGAAAGCTTGTCTCAACCC-3′ 5′-CTCCTTCAGGAACAGCCACCAA-3′
IL-1β	5′-AGCCATGGCAGAAGTACCTG-3′ 5′-TGAAGCCCTTGCTGTAGTGG-3′
TNF-α	5′-CACCACTTCGAAACCTGGGA-3′ 5′-AGGAAGGCCTAAGGTCCACT-3’.

Cycle threshold (Ct) values were normalized per experiment and per gene. ΔΔCt was calculated using the housekeeping gene Gapdh (mice) or β-ACTIN (human). Each qPCR experiment was conducted in biological triplicate, and each individual run was repeated in technical triplicate. Statistical analysis was performed using a Student's t-test.

2.26. Metabolomics data generation and analysis

Neutrophils isolated from WT mice and Ncf1^−/− mice (n = 5, respectively), and re-suspended in RPMI 1640 media at a concentration of 5 × 10⁶ cells/mL. Neutrophils were either unstimulated or treated with fMLF (100 nM, 10 min) or PMA (100 nM, 30 min). Cells were centrifuged at 300×g at 4 °C for 5 min. The supernatant was aspirated, and cell pellets were re-suspended with ice-cold PBS, then centrifuged at 300×g at 4 °C for 5 min. The supernatant was discarded, while the pellets were stored at −80 °C before intracellular metabolite extraction. Samples were subjected to methanol extraction and then were split into aliquots for analysis by ultrahigh performance liquid chromatography/mass spectrometry (UHPLC/MS) in the positive, negative, or polar ion mode and by gas chromatography/mass spectrometry (GC/MS). Data acquisitions were performed using an LC-MS system, which is a Waters Acquity UPLC system coupled in tandem with a Waters photodiode array (PDA) detector and a SYNAPT G2-Si HDMS QTOF mass spectrometer (Waters, Manchester, UK). The multivariate data matrix containing information on sample identity, ion identity (retention time and m/z), and ion abundance was generated by centroiding, de-isotoping, filtering, peak recognition and integration. The intensity of each ion was calculated by normalizing the number of individual ions to the total number of ions in the entire chromatogram. The data matrix was further exported to SIMCA-P software (Umetrics, Kinnelon, NJ) and transformed using mean centering and Pareto scaling, a technique that enhances the importance of low-abundance ions without significantly increasing noise.

Metabolomics data were analyzed using R software (version 4.3.3). PCA and partial least squares discriminant analysis were used to analyze the quantitative metabolomic data imported into the R 2.00 package. Metabolite sets were significantly enriched at P < 0.05 and FC > 1.2. Metabolic pathway analysis was further performed using the KEGG enrich function in the R package clusterProfiler. Heatmaps for metabolic pathways were generated based on KEGG pathways.

2.27. Statistical analysis

Graph of FCM results and qPCR and statistical analysis were performed using Graphpad Prism 9.0 (Graphpad Software, CA, USA). Single comparisons were made using unpaired two-tailed t-tests. Multiple comparisons were analyzed using one-way or two-way ANOVAs with a Tukey's or Sidak post-hoc multiple comparisons test. Results were presented as the means ± SEM. P values < 0.05 were considered significant. All other statistical analyses were performed using R software (version 4.3.3). DEGs were selected by FC > 1.25 and P < 0.05, and the DEMs were identified by FC > 1.2 and P < 0.05, as parameters to define the statistical significance. GO terms with P < 0.05, and KEGG pathways with P < 0.05 were screened out.

3. Results

3.1. The ROS metabolic process in a specific neutrophil group is vital in the development of experimental arthritis

BM is the primary site of hematopoiesis and is responsible for producing all types of blood cells, including neutrophils, and is crucial for understanding the development and function of neutrophils in RA [31,[41], [42], [43]]. To characterize neutrophils during immune priming and arthritis onset, we made a scRNA-seq analysis of BM cells from 1) naive, 2) immune primed (day 0 (D0) and 3) immune boosted (a repeated COL2 immunization at D21) i.e. during arthritis onset. In total, 72388 cells were analyzed, identifying 10 different clusters based on the expression of canonical markers, including B cells, plasma cells, HSCs, ProEryth, erythroblasts, GMP, neutrophils, basophils, T cells, and MPs (Fig. 1A and B, S1A and B). Among these, the proportion of neutrophils varied significantly between naive and primed groups (Fig. 1C and D). However, no significant difference was observed between the naive and the boosted groups, probably because neutrophils of boosted mice were recruited more efficiently to the sites of inflammation, suggesting neutrophils migrate from BM to tissues without passing through the vascular compartment (Fig. 1C and D). We depleted neutrophils in DBA1/J mice with anti-Ly6G antibody and subsequently immunized them with COL2 and found that only neutrophil-deplete mice developed mild arthritis, suggesting that neutrophils could prevent primed mice from developing arthritis in the early stages of CIA (Fig. 1E and F).

Fig. 1.

scRNA-seq analysis of BM neutrophils in arthritis. (A) Schematic diagram of CIA model and overview of the study design for scRNA-seq. 3 groups of BM samples were processed by scRNA-seq at D42 (n = 9 samples) including naive, primed with immunized COL2 at D0, and boosted with immunized COL2 at D0 and D21, respectively. 5 mice were tested per group, and 3 mice were randomly selected for scRNA-seq. (B) UMAP plot shows that a total of 72388 single cells from BM identified 10 cell populations in each group. Each dot corresponds to a single cell. (C) Proportions of the 10 cells populations in 9 samples. (D) Percentage of neutrophils in 10 cells among 3 groups. Data represent means ± SEM (n = 3). (E) 100 μg of anti-Ly6G was injected i.p. into primed mice on day −2, 0, +2, and +4 until the experiment endpoint (D42). Clinical scoring was performed every other day based on the number of inflamed joints in each group until the experimental endpoint D42 (n = 5 animals per group). (F) Neutrophils were analyzed by FCM from BM of 3 groups. (G) UMAP representation of scRNA-seq data for the MPs and neutrophils from figure B in naive, primed, and boosted groups, colored by inferred cluster identity (G0-G4) (left) and absolute number of neutrophils among 3 groups (right). (H) GO analysis of the upregulation profiles of G4 cells in primed compared to naive and boosted. The ROS metabolic process was marked with a red font. (I) The dot-plot shows the percentage (pct. exp.) and average expression (avg. exp.) of Ncf1 and Cybb (NOX2 complex) in the G4 cluster. (J) UMAP plot shows that Ncf1 gene expression in the neutrophil G4 subpopulation. HSCs, hematopoietic stem cells; ProEryth, pro-erytrhrocytes; GMP, granulocytes-monocyte progenitor cells; MPs, mononuclear phagocytes. ∗P < 0.05. ns, nonsignificant.

To dissect neutrophil heterogeneity, we further examined neutrophil-related populations (MPs and neutrophils). Unsupervised clustering partitioned differentiating and mature neutrophils into five clusters (G0-4) (Fig. 1G and S1C). Likewise, the number of neutrophils in primed, boosted, and naive mice were ranked from highest to lowest (Fig. 1G). Single-cell trajectory analysis revealed a highly consistent time course of neutrophil maturation and differentiation, demonstrating a developmental trajectory of neutrophils from G0 to G4 (Fig. S1D). In addition, we observed a significant increase in the number of cells differentiating toward the G4 population in primed mice (Fig. S1E). In a GO analysis of DEGs, cell cycle-related genes were highly expressed in earlier phases of neutrophil maturation (G0, G1, and G2), and redox-related genes were upregulated at late phages (G3 and G4) (Figs. S1F and G). GO enrichment analysis of G4 cells revealed that they were closely related to the response to ROS metabolic process and regulation of inflammatory response (Fig. 1H). In G4 cells, NADPH oxidase, ROS production, chemotaxis, neutrophils activation, and maturation score were significantly upregulated in the early stages of arthritis (Fig. S1H). We also found that the G4 subpopulation highly expresses Ncf1, an essential subunit of the NOX2 complex responsible for ROS induction, further supporting that neutrophil ROS production was mainly mediated by phagocytic NADPH oxidase in the pre-arthritic phase (Fig. 1I and J).

3.2. ROS shows a dual role in the development of experimental arthritis

To explore the effect of ROS derived from the NCF1 component of the NOX2 complex at various stages of arthritis we treated the mice with the general NOX inhibitor DPI to block ROS production. The DPI dose (100 ng/kg) was selected based on a dose-response study (10, 100, 1000 mg/kg) and was found to be the lowest dose that significantly alleviated the arthritis (Fig. S2A).

The treatment with DPI (100 ng/kg/day) started before priming at day 0 (D0) (priming) or before onset of arthritis after the boost at D21 (onset) (Fig. 2A). Interestingly, the daily DPI treatment started before priming (D0) lead to the development of arthritis, with swelling, arthritis scores and severity of synovial hyperplasia at D42 (Fig. 2B–E). The later development of arthritis in the DPI-treated primed mice increased MPO activity, most likely reflecting an increased neutrophil activity (Figs. S2B–D). Moreover, DPI-treated primed mice developed arthritis with increased bone destruction (Figs. S2E and F). However, trabecular microstructural analysis showed no changes in bone volume/tissue volume ratio (BV/TV), trabecular number (Tb·N), trabecular thickness (Tb·Th), and trabecular separation (Tb.Sp) in primed mice with or without DPI treatment (Fig. S2G). We examined the neutrophil population and numbers in the PB at various time points. Prior to 21 days after immunization, ROS levels in PB neutrophils were elevated in primed mice. DPI could increase PB neutrophils in primed mice by inhibiting neutrophil ROS levels similar to naive mice, and accompanied by a maximum neutrophil level at D14 (Fig. 2F and G). We further examined neutrophils in BM, joints, and spleen of mice at D14 by FCM. Primed mice treated with DPI showed reduced BM neutrophil level, but no change in joints and spleen (Fig. S2H). These results suggested that DPI increased blood neutrophils in primed mice may be from BM during the priming phase which could have contributed to the earlier onset of arthritis.

Fig. 2.

ROS plays a bidirectional role in early protection and late pathogenesis of arthritis. (A) Schematic representation of the animal experimental protocol. Construct CIA models under different stages, primed and boosted, respectively (5–6 animals per group). Primed mice were treated or untreated with DPI (100 ng/kg/day) on D0 and boosted mice on D21. Mice were treated for 42 days, accompanied by recording clinical arthritis scores, and then sacrificed for subsequent experiments. In the primed stage, (B) representative images of swollen hind paws at D42, (C) clinical score of arthritis over time from D21 to D42, (D) synovial tissue via HE staining (40 × ), and (E) synovial proliferation score was shown on D42 for each group. (F) The proportion (left) and absolute number (right) of neutrophils (CD11b⁺ Ly6G⁺) was detected in PB by FCM for each group on days 7, 14, and 21. (G) The ROS level of neutrophils in PB was measured using the fluorescent probe DCFH-DA via FCM on days 7, 14, and 21. The MFI was used to indicate levels of ROS. In the boosted stage, (H) change in hind paw swelling on D42, (I) clinical arthritis score over time from D21 to D42, (J) knee joints with HE staining (40 × ), and (K) synovial proliferation score on D42 in each group. (L) The proportion and absolute number and (M) ROS level of neutrophils was detected in PB by FCM on D42 for each group. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001. ns, nonsignificant.

However, after the development of arthritis, the proportion and numbers of PB neutrophils were higher in DPI-treated mice (25–55 %, 400 counts) than in control-primed mice (10–20 %, 200 counts) at D42 (Fig. S2I). In this process, we also examined the percentage and number of monocytes and macrophages. There was a tendency for monocytes to increase under DPI intervention, but the multiplicity of increase was much less than that of neutrophils in primed mice (Fig. S2I). Further, in mice treated prior to arthritis onset, DPI was administered with once-daily interventions starting at D0 until D42. ROS levels in neutrophils were not suppressed by treatment with DPI at D42 (Figs. S2I and J). This phenomenon was also observed in macrophages and monocytes (Figs. S2I and J). The reason for the very high ROS level at this stage may be that ROS from the non-NOX2-dependent pathway was also involved in the most severe stages of arthritis. Nevertheless, treatment at day 21–42 (during arthritis onset) with DPI reduced the clinical scores for paw arthritis compared to non-treated mice (Fig. 2H and I), as well as synovial hyperplasia of the knee joint at D42 (Fig. 2J and K). Mechanistically, DPI was observed to reduce the level of neutrophils with inhibited ROS levels and not of monocytes and macrophages (Fig. 2L and M, S2K and L). Among these inflammatory cells, neutrophils are likely to play an important role.

Taken together, these results suggest that during immune priming, loss of ROS leads to the development of arthritis with increased neutrophil numbers in the joints. In contrast, during arthritis onset, the loss of ROS alleviates arthritis symptoms by decreasing neutrophil numbers, suggesting that ROS bidirectionally regulates arthritis.

3.3. ROS deficiency of neutrophils in the early stages of experimental arthritis aggravates joint inflammation

Induced activation of the NOX2 complex is the main source of ROS for neutrophils, another small part comes from mitochondria. To evaluate the role of different sources of ROS in early CIA, we employed GSK2795039, a NOX2-specific inhibitor, and the mitochondrial ROS inhibitor MitoQ10. We treated mice with GSK2795039 or MitoQ10 daily until the endpoint of the experiment immunized with COL2 at D0 (Fig. 3A). Mice immunized once (primed) developed arthritis, provided they were treated with inhibitors of NOX2 or mitochondrial ROS (Fig. 3B and C). In addition, it was observed that the levels of neutrophils in the joints were higher in primed mice with NOX2 or mitochondrial ROS inhibitors, with the largest numbers of neutrophils in the NOX2 inhibitor group (Fig. 3C). Thus, inhibition of NOX2 complex-derived ROS more potently promoted arthritis than inhibition of mitochondria-derived ROS (Fig. 2, Fig. 3B). These results suggested that primed mice were developed a higher susceptibility to arthritis after inhibition of NOX2-derived ROS, indicating that neutrophil-derived ROS is an important regulatory factor.

Fig. 3.

ROS deficiency triggers arthritis by recruiting neutrophils in the early stage. (A) Schematic drawing of the experimental protocol. Animals were randomly divided into 5 groups (5–6 animals per group), including [1] Naive group [2], Primed group [3], Primed + GSK2795039 (1 mg/kg/day) group [4], Primed + MitoQ10 (1 mg/kg/day) group. Mice were treated for 42 days, accompanied by recording clinical arthritis scores. (B) The representative pictures of the hind paws of primed mice with different treatments at D42. (C) Time course changes from D21 to D42 in arthritis score from primed mice treated with GSK2795039 (left), and MitoQ10 (center), respectively. The neutrophil percentage was determined by FCM in BM, PB, joint, and spleen for each group on D42 (right). (D) Schematic showing experimental design for COL2 treatment of Ncf1 knockout and control animals. Mice were divided into WT, Ncf1^−/−, Primed Ncf1^−/−, and Primed WT groups. Primed mice were injected with LPS (i.p., 25 μg/mouse) on D28 to induce neutrophil-driven acute inflammation (n = 5). (E) Representative photos (n = 3) and (F) clinical arthritis score showing hind-paw swelling of primed WT and primed Ncf1^−/− mice at D35 (n = 5). Red arrows indicated areas of swelling. (G) Representative FCM plots for PB neutrophils (CD11b⁺ Ly6G⁺) in each group at D14 and D35. Statistical chart of FCM result of the percentages (top) and absolute number (bottom) of neutrophils (H), FCM histograms and the MFI of ROS production by neutrophils from figure H (I), and the percentages (left) and absolute number (right) of CXCR2⁺ cells (J), at D14 and D35. (K) Statistical chart of FCM result of the percentages (left) and absolute number (right) of neutrophils in BM, PB, joint, and spleen at D35. (L) mRNA expression levels of chemokine receptors (Cxcr1, Cxcr2, and Cxcr4) and inflammatory cytokines (IL-1β, IL-6, IL-17, TNF-α, and IFN-γ) in the BM neutrophils of indicated groups at D35. ∗P < 0.05, ∗∗P < 0.01, and ∗∗∗∗P < 0.0001. ns, nonsignificant.

Next, we established a model of early arthritis employing a single immunization with COL2 boosted with an LPS injection, in C57BL/6J mice with Ncf1 gene knockout (Ncf1 ko, Ncf1^−/−) mice, with deficient NOX2 activity with or without a one-time COL2 priming immunization compared to WT mice (Fig. 3D–S3A and B). Importantly, the development of arthritis in this mouse, with the major histocompatibility complex type II allele A^b instead of A^q as in the DBA1/J mouse, is not dependent on a T cell response to COL2 but is entirely dependent on the development of arthritogenic anti-COL2 antibodies [44]. After the injection of LPS, joint swelling was only observed in Ncf1^−/− and not in WT mice (Fig. 3E and F, showing scoring at D35).

In PB of naive mice, not immunized with COL2, there was no difference in the percentage of the different immune cells between Ncf1^−/− and WT mice (Fig. 3G and H, S3C), as well as ROS levels (Fig. 3I and Fig. S3D). Primed Ncf1^−/− mice had increased levels of neutrophils and CXCR2-expressing cells with ROS deficiency on D14 without arthritis scores, after which the mice developed arthritis at the experimental endpoint of D35 (Fig. 3G–J) and the proportion of other cells (macrophages, monocytes, CD4⁺ T cells, and CD8⁺ T cells) with ROS deficiency was lower than that of WT mice (Figs. S3C and D). Incredibly, we found two CD4⁺ T cell populations, CD4⁺ T cells-1 and CD4⁺ T cells-2, which differed significantly in their ROS levels. In primed mice, CD4⁺ T cells-2 level was 2–3 times higher than CD4⁺ T cells-1 (Fig. S3E). When arthritis had developed, at D35, the number of neutrophils in BM was higher in WT mice than in Ncf1^−/− mice. The numbers of neutrophils were higher in Ncf1^−/−mice, especially outside BM, indicating an increased mobilization of NCF1 deficient neutrophils from the BM (Fig. 3K and Fig. S3F). Interestingly, with the onset of arthritis, ROS-deficient Ncf1^−/−mice had higher levels of neutrophil ROS than WT mice in PB (Fig. S3G), which was consistent with what was observed with the DPI intervention (Fig. S2I), suggesting that the source of late-stage ROS was complex. Furthermore, neutrophils isolated from the BM of Ncf1^−/− mice showed upregulated neutrophil chemotactic receptors CXCR1 and CXCR2, and the inflammatory cytokines IL-1β, IL-6, IL-17, TNF-α and IFN-γ, compared to primed WT mice at D14 and D35，and the results were even more significant at D35 (Fig. S3H and Fig. 3L). Notably, CXCR4 expression was reduced in Ncf1^−/− neutrophils, predicting a reduced BM homing capacity of Ncf1^−/− neutrophils compared to WT neutrophils from primed mice (Fig. S3H and Fig. 3L).

Collectively, these results suggest that ROS deficiency in the early stages of experimental arthritis promotes neutrophils to exhibit proinflammatory properties, contributing to the development of arthritis.

3.4. ROS deficiency affects the functions of neutrophils

Neutrophil viability was closely related to neutrophil oxidative burst activity, which is responsible for the production of ROS. We explored the time and concentration of ROS inhibitors to determine the optimal conditions for the experiments. With regard to the time factor, all of the ROS inhibitors seem to exert a significant inhibitory effect on neutrophil ROS levels, which appear to reach equilibrium at approximately 30 min (Fig. S4A). In the presence of ROS and ROS inhibitors for 30 min, the ROS production by PMA or fMLF-stimulated neutrophils was inhibited in a dose-dependent manner (Fig. 4A and S4B). The amount of ROS derived from NOX2 was approximately three times higher than that produced by mitochondria, and both sources of ROS (Fig. S4C). Calcium mobilization was required to regulate numerous neutrophil biological responses. Further exploring the correlation between ROS and calcium flux, we observed that ROS inhibitors downregulated the calcium flux of neutrophils triggered by PMA or fMLF in a dose-dependent manner (Fig. 4B and S4D). ROS-deficient Ncf1^−/− neutrophils also displayed a decreased neutrophilic calcium flux (Figs. S4E and F). This suggests that the latter was regulated by ROS-dependent signals involved in neutrophil biology.

Fig. 4.

Neutrophil function and inflammatory phenotype changes in the absence of ROS. Neutrophils were incubated with DMSO (solvent), DPI (1 μM, 10 μM, 100 μM), GSK2795039 (1 μM, 10 μM, 100 μM), MitoQ10 (0.5 μM, 1 μM, 5 μM), and TBHP (100 μM) for 30 min, and then treated with 100 nM PMA for 30 min. (A) The ROS level in neutrophils was evaluated via FCM. The bar graph shows the levels of intracellular ROS, as detected by DCF MFI (left). Representative flow cytometric histograms show a broad unimodal distribution of DCF fluorescence in different cell groups (right). (B) The Ca²⁺ flux in neutrophils was detected via FCM using Fluo-3AM. (C) Neutrophils (Ly6G-tdTomato⁺ cells, red) were incubated with 10 μM DMSO, 10 μM DPI, 10 μM GSK2795039, 1 μM MitoQ10 for 30 min, and then the nucleus was stained using Hoechst 33342 (blue). Morphological changes of chromatin (blue) and cell membrane (red) during neutrophils stimulated with or without 100 nM PMA for 30 min imaged by live-cell confocal laser scanning microscopy. Cells stimulated with 3 % H₂O₂ were used as positive control. (D) For adhesion and migration assays, neutrophils were pretreated with 10 μM DMSO, 10 μM DPI, 10 μM GSK2795039, 1 μM MitoQ10 for 30 min. HUVECs were cultured in medium alone or stimulated with 20 ng/mL IL-8, and then incubated with pre-treated neutrophils for 2 h. The number of neutrophils adhered to HUVECs was counted by giemsa stain. Each sample was counted at least 3 times, and the average was taken. (E) Pre-treated neutrophils were placed into transwell chambers in response to 20 ng/mL IL-8 in the bottom well. After 2 h, the number of neutrophils that migrated to the bottom layer was counted with an automatic cell counter, and images were taken, and neutrophils were counted three times, and an average was taken. (F) HUVECs were seeded into transwell filters one day before assay. Cells were grown to 80–90 % confluency and added pre-treated neutrophils to the upper chambers. Then, 20 ng/mL IL-8 was introduced in the lower chamber. After 2 h, the number of transmigrated neutrophils was counted using an automatic cell counter. (G) Transwell assays comparing the migration of neutrophils from WT and Ncf1^−/− mice in response to 100 nM fMLF or 20 ng/mL IL-8 for 2 h. (H) mCherry-E. coli was incubated with WT or Ncf1^−/− neutrophils at a neutrophil: E. coli ratio of 1:10. mCherry-E. coli uptake by neutrophils was analyzed by FCM after 30 min. (I) Neutrophils were isolated from mice BM, and were incubated with 10 μM DMSO, 10 μM DPI, 10 μM GSK2795039, 1 μM MitoQ10, and 100 μM TBHP for 30 min. Then, after stimulation with 50 nM fMLF for 12 h, FCM assessment of the development of Ly6G and CXCR2 expression. (J) Lymphocytes and neutrophils derived from BM were obtained from mice and incubated with 10 μM DMSO, 10 μM DPI, 10 μM GSK2795039, 1 μM MitoQ10, and 100 μM TBHP for 30 min. After added with 50 nM fMLF for 5 days, FCM analysis of the expression of CD3⁺ CD4⁺ T cells, and the proportions of Th subtypes (Th1, Th2, Th17, and Treg cells) in CD3⁺ CD4⁺ T cells in a co-culture system of neutrophils isolated BM and lymph node cells. Each above experiment was representative of three independent experiments. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001. ns, nonsignificant.

Next, we tested whether ROS directly caused the morphological changes in neutrophils. After PMA stimulation for 30 min, the round neutrophils flattened and adopted polarized morphology with depolymerized nuclei. DPI blocked these changes, but the effect of other inhibitors was not noticeable (Fig. 4C and S4G). A fter stimulation with H₂O₂, neutrophils underwent apoptosis and nuclear rupture, rather than the NETs induced by PMA stimulation for 2h (Fig. 4C and S4H). The above results showed that exogenous ROS (H₂O₂) induced neutrophil apoptosis, and endogenous ROS derived from NOX2 were involved in neutrophil activation. After inhibition of ROS, neutrophils showed reduced adhesion and increased migration and diapedesis (Fig. 4D–F). We also observed enhanced migration in Ncf1^−/− neutrophils (Fig. 4G). As for phagocytosis, functional knockout of NCF1, as well as pharmacological inhibition, decreased the phagocytosis of E. coli in ROS-deficient neutrophils (Fig. 4H and S4I).

Chemotactic receptors and inflammatory cytokines play an important role in neutrophil recruitment and propagation of the inflammatory response. Therefore, we evaluated the correlation of ROS with neutrophil chemoattractant receptors and inflammatory cytokines. In vitro, human neutrophils pre-treated with ROS inhibitors for 30 min were cultured with or without fMLF. qPCR analysis showed that the expression of chemokines (CXCL1 and CXCL2), chemoreceptors (CXCR1 and CXCR2), and inflammatory cytokines (IFN-γ, TFN-α, and IL-1β) were downregulated after 6 h of fMLF stimulation. In contrast, ROS inhibitors upregulated the expression of these genes, except for TNF-α, for which a ROS inhibitor targeting NOX2 was effective (Fig. S4J). Maturity analyses showed that neutrophils with reduced ROS had increased expression of CXCR2 and Ly6G after fMLF stimulation for 12 h (Fig. 4I). These ROS-deficient neutrophils likely exhibited the increase of the neutrophil maturation process, long-lived, with less apoptosis and less phagocytotic capacity. Moreover, they were likely to have less potent toxic granular. Therefore, these results suggested that ROS-deficient neutrophils did not have the classical proinflammatory phenotype.

The mutual activation of neutrophils and T cells could contribute to the perpetuation of the local inflammatory process, and eventually to the destructive process in RA [45]. As early effector cells of arthritis, the ROS-driven neutrophil's role was further explored in the regulation of T cells-mediated adaptive immune responses. To this end, we isolated PB cells, lymph node cells, and BM neutrophils. PB cells, lymphocytes, lymphocytes in co-culture, and neutrophils were pretreated with various ROS inhibitors for 30 min in vitro, followed by the addition of fMLF. Cells were then cultured for 5 days. FCM analyses revealed that ROS inhibition increased the proportion of CD4⁺ T cells, Th1 and Th17 cells. Proliferation and differentiation of T cells were not seen in lymph node cells (Figs. S4K and L). However, in lymphocyte and neutrophil co-culture, we observed an increase in the proportion of CD4⁺ T cells, Th1, and Th17 cells after ROS inhibition and a decrease of Th2 cells (Fig. 4J). Inhibition of ROS resulted in a higher proportion of Th1 and Th17 cells in the lymphocytes/neutrophils co-culture, and showed higher proportions than PB cells (Fig. 4J and S4M). These results suggested that intervening with NOX2-derived ROS enhanced the ability of neutrophils to promote the differentiation of CD4⁺ T cells into Th1 and Th17 cells, which could contribute to arthritis development.

In summary, these data together demonstrated that neutrophils with ROS deficiency were self-sustaining by reducing oxidative respiratory bursts, calcium flux, and morphological non-expansion. This altered level of ROS leads to a shift in neutrophil function, primarily in the form of increased migration, transvascular endothelial capacity, and inflammatory immune response, coupling with decreased phagocytosis and adhesion, for rapid migration to the inflammatory fraction to exert their effects.

3.5. Neutrophils with ROS deficiency perform a more robust inflammatory metabolism

Activation of neutrophils is associated with a large metabolic demand. To characterize the metabolic changes associated with ROS-regulated neutrophil activation, we investigated the response of neutrophils from WT and Ncf1^−/− mice stimulated in vitro in the presence and absence of PMA and fMLF (Fig. 5A). Box plots show metabolite distribution, median and extremes for these samples (Fig. 5B). PCA revealed no difference in metabolites between Ncf1^−/− and WT neutrophils in the absence of stimulation, but metabolites changed rapidly in response to fMLF or PMA (Fig. 5C). After PMA stimulation, we identified 92 and 61 differentially expressed metabolites (DEMs) in Ncf1^−/− and WT neutrophils, respectively (Fig. 5D, left and 5E, top). Moreover, 31 DEMs were identified in Ncf1^−/− neutrophils and 36 DEMs in WT neutrophils in response to fMLF (Fig. 5D, right and 5E, bottom). This shows that Ncf1^−/− neutrophils have a unique metabolome.

Fig. 5.

Inflammation metabolic reprogramming in ROS deficient neutrophils.

Neutrophils isolated from WT and Ncf1^−/− mice treated with or without 100 nM fMLF for 10 min and 100 nM PMA for 30 min separately (n = 5 for each group). The quantitative metabolomics analysis of neutrophils intracellular metabolites was performed by LC-MS using an untargeted metabolomics platform. (A) Strategy map for the metabolic analysis. (B) A box plot was used to show the relative levels of metabolites in each group. (C) PCA data showed that the metabolite profiles differed among neutrophil samples. (D) Volcano plot showed the DEMs in neutrophils (Ncf1^−/− and WT neutrophils) with PMA stimulation (left) or fMLF (right), compared to the control group (Ncf1^−/− and WT neutrophils). (E) Venn diagram showed several significant DEMs of neutrophils (Ncf1^−/− and WT neutrophils) with or without stimulation (PMA, top; fMLF bottom). 92 unique DMEs in Ncf1^−/− PMA vs Ncf1^−/− Control, and 61 unique DMEs in WT PMA vs WT Control; 31 unique DMEs in Ncf1^−/− fMLF vs Ncf1^−/− Control, and 36 unique DMEs in WT fMLF vs WT Control. These DEMs were further used for KEGG pathway enrichment analysis, with heat maps showing the top 30 pathways. (F) Metabolic pathways that WT and Ncf1^−/− neutrophils were enriched to in response to PMA stimulation. The red front indicated a significant difference. (G) Metabolic pathways that WT and Ncf1^−/− neutrophils were enriched to in response to fMLF stimulation. The red front indicated significant differences, as well as the PMA stimulation; the green color indicated unique significant differences with fMLF stimulation. (H) Schematic of the hypothesized mechanism by which defective NCF1-derived ROS mediate pro-inflammatory phenotypic shifts in neutrophils that mediate arthritis.

After stimulation with PMA in vitro, Ncf1^−/− neutrophil metabolic pathways differed from that of WT neutrophils. In the former, the enhancement was focused on inflammation- and immune-related pathways, like metabolism of histidine, glycine and serine and threonine signaling (3 FC, P < 0.005) (Fig. 5F, left). In contrast, neutrophil pathways were focused on energy metabolism and biodegradation without inflammation, including transaldolase, ribose-5-phosphate isomerase, glucose-6-phosphate dehydrogenase, pentose phosphate pathway (PPP), fructosuria, fructose and mannose degradation, and fructose intolerance hereditary (3 FC, P < 0.005) (Fig. 5F, right). In addition to inflammatory and immune-related pathways, Ncf1^−/− neutrophils stimulated by fMLF were found to upregulate metabolic pathways involved in the action of drugs like fluvastatin and alendronate (2 FC, P < 0.005) (Fig. 5G). These results suggest that ROS deficiency leads to abnormal neutrophil metabolism, closely related to aberrant inflammation and immune activation, in addition to imbalances in energy regulation and drug metabolism.

In an immune priming environment, ROS-deficient neutrophils undergo a phenotypic shift from pathways associated with energy metabolism and biodegradation towards inflammation and immunometabolism, thereby promoting inflammation.

4. Discussion

NCF1-derived ROS plays a central regulatory role in the initiation of arthritis. Neutrophils with deficient NCF1, leading to low ROS-inducing capacity during the immune priming, is associated with a higher susceptibility to arthritis. NCF1 deficiency leads to activation of neutrophils, including ROS metabolic pathways and inflammatory immune response. However, with a reduction of NCF1-derived ROS during late arthritis, the mice were relieved of their symptoms, accompanied by a decrease of neutrophils, suggesting that NCF1-derived ROS may modulate the involvement of neutrophils in the arthritis process. In an in vitro induced inflammation model, NCF1-deficient neutrophils exhibited the increase of the neutrophil maturation process, migration, chemotactic receptor and inflammatory cytokine expression, and adaptive immune response. Furthermore, the metabolism of NCF1-deficient neutrophils undergoes reprogramming from a predominantly energy metabolism to an inflammatory immune metabolism to maintain the inflammatory profile of neutrophils, which in turn is involved in early inflammation in arthritis.

Neutrophils are believed to be critical players during the gradual progression of RA from an initial loss of immunoinflammatory tolerance to persistent synovial inflammation [4,6,8]. We used scRNA-seq to analyze BM from CIA mice at different stages, from immune priming to arthritis onset, and found that neutrophils change most dramatically in the early immune priming stage. Our findings may contrast with current dogma as neutrophils have been shown to be pathogenic in the development of arthritis. For instance, the knockdown of granulocyte colony-stimulating factor inhibits the ability of BM neutrophils to generate and form NETs in vitro, which in turn prevents CIA [46]. Knockdown of CXCR2 leads to decreased recruitment of neutrophils to the site of joint inflammation, which in turn attenuates K/BxN serum-transfer arthritis [47]. Gfi-1-deficient mice with abnormal neutrophil differentiation are resistant to K/BxN serum-transfer arthritis with some resistance [48]. Together, these findings reveal a vital pathogenic role for neutrophils in arthritis, whereas our finding provides the first strong evidence that neutrophils play a protective role during immune priming leading to arthritis. However, these findings are not contrasting as NCF1 expressing neutrophils play a protective role during immune priming to prohibit the initiation of autoimmune inflammation, whereas they are associated with, and maybe also driving, ongoing inflammation.

By scRNA-seq analysis, we further found that the highly mature G4 neutrophil subpopulation was abundantly expressed in the BM of early CIA mice. The specific G4 cells were predominantly enriched in pathways related to ROS production and metabolic pathways. In RA, neutrophil-generated ROS is manifested by the oxidative modification of IgG to produce new epitopes that induce RA autoantibodies and by cytotoxic effects that lead to the degradation of cartilage and collagen [5,49]. ROS production primarily depends on NOX2-mediated oxidative burst, with a minor contribution from mitochondrial ROS. Our results show that different ROS inhibitors induced different degrees of arthritis at the onset of inflammation. Among them, the NOX2 inhibitors DPI and GSK2795039 induced the most severe inflammation, while mitochondria-targeting MitoQ10 was slightly less effective. Moreover, MPO is stored in azurophilic granules within the neutrophil. The NOX2 generates superoxide (O₂^•-) that is converted into H₂O₂. MPO uses H₂O₂ to catalyze the production of potent antimicrobial molecules, such as hypochlorous acid (HOCL, OCl⁻) [50]. The interaction of hypochlorite with H₂O₂ leads to the generation of singlet oxygen (ROS) [51]. Thus, ROS inhibitors, by reducing upstream NOX- or mitochondria-driven ROS, would further reduce downstream MPO-driven ROS. Therefore, we infer from the above results that ROS inhibitors affect the activity of MPO during inflammation mainly through a series of related inflammatory features that are mediated on the basis of altered ROS levels, such as a pro-inflammatory phenotypic shift. We also observed that arthritis was induced in NCF1-deficient mice in the early stages of CIA. Furthermore, it has been shown to cause severe arthritis in both Ncf1 mutant mice and rats [10,19]. These results confirm that Ncf1-derived ROS play an important protective role in the development of arthritis and that NCF1-expressing neutrophils contribute to this effect.

At the cellular level, targeted blockade of NOX2-derived ROS production was linked to increased neutrophil migration, maturation process, chemotactic receptor CXCR2 and pro-inflammatory cytokine expression, and Th1/17 differentiation, whereas targeted blockade of mitochondrial ROS was not. Furthermore, in Ncf1^−/− mice after COL2 immunization, neutrophils were observed to switch to an inflammatory phenotype mediating arthritis, including increased chemotactic receptors CXCR1 and CXCR2 and inflammatory cytokines, as well as attenuated CXCR4-mediated neutrophils homing to BM, indicating that Ncf1^−/− neutrophils had a prolonged retention time outside the BM involved in the joint inflammatory response. Lack of ROS also increased IL-1β production by activating inflammasome-independent protease activity, and increased IFN-γ responses via the STING and JAK1/STAT1 pathways [52,53], which may be a major pathway mediating neutrophil inflammation. In addition, IL-17A expression was highest on neutrophils within the inflamed synovial tissue, linking T cell activation to neutrophil recruitment and activation [54,55]. In Ncf1-deficient mice, it has been reported to promote the expansion of IL-17A-producing T cells (Th17 cells) specific for immunodominant COL2 peptides, interferon signature, T cells autoreactivity, and high titers of anti-COL2 Abs, leading to severe arthritis [18,56]. We observed that neutrophils with ROS deficiency highly expressed IL-17A and IL-1β, and enhanced the differentiation of Th17 cells, suggesting that ROS deficiency promoted neutrophil-mediated T cells differentiation into Th17 cells, leading to severe arthritis. However, in primed mice, we observed that CD4⁺ T cells-2 had high ROS level compared to CD4⁺ T cells-1 in both WT and Ncf1^−/− mice. This in turn resulted in ROS level in CD4⁺ T cells similar to those in macrophages. More work is needed to explain this finding, which is at odds with the low expression of NCF1 in these cells. Moreover, Ncf1-deficient mice were also associated with increased germinal center formation, a stronger autoreactive B cells (C1-specific B cells) response and intramolecular epitope spreading (a spreading of antibody specificity to non-C1 epitopes present on the COL2 triple helical molecule), a promoting effect during the late phase of CIA [57,58]. Thus, NCF1-derived ROS have an important role in priming and later disease time-point.

Interestingly, ROS inhibitors failed to reduce neutrophil ROS levels in advanced arthritis, likely due to high ROS production from non-NOX2 sources that could not be blocked by DPI. NOX2 catalyzes the production of O₂^•- from oxygen; O₂^•- further generates H₂O₂ and ^•OH by the action of SOD enzyme. But in this process, in addition to NOX2, mitochondria, peroxisome and ER all can produce O₂^•- and H₂O₂. Another possible reason is that the balance between generating and scavenging late ROS is complex. In RA, oxidative stress is increased, suggesting that there is a dysregulation of the antioxidant system. There are various types of antioxidants, including the glutathione system, enzymes, thioredoxin system, NADPH supply system, vitamins, thiols, etc., which work to cope with ROS- and free radical-associated stress [59]. However, ROS is a double-edged sword. At low concentrations, ROS maintains cellular homeostasis and regulates signal transduction [60,61], whereas at high concentrations, ROS induces genomic mutations and irreversible oxidative modifications of proteins, lipids, and polysaccharides, promoting disease [[62], [63], [64], [65]]. A dual role of ROS is observed also in the initiation and metastasis of cancer. NCF1-expressing neutrophils have been shown to downregulate the anti-cancer inflammatory response [66], whereas high levels of ROS in cancer cells are detrimental to their growth [67]. New anticancer therapeutic approaches developed for ROS duality based on the modulation of ROS generation or antioxidant mechanisms have made good progress [68,69]. In the present study, the dual role of ROS in arthritis has been suggested to also involve neutrophils. The bidirectional action of ROS not only casts a new perspective on the mechanism in early RA development, but may also help to develop novel detection tools for real-time monitoring and early intervention in at-risk populations, offering the potential target for RA prevention and treatment.

Under pathogenic microbial infection, neutrophils deficient in mitochondrial respiration primarily rely on glycolysis and PPP for energy supply and regulation of functional effects [70,71]. During the oxidative burst, neutrophils switch from a glycolysis-dominated metabolism to a unique metabolic pattern known as the "pentose cycle" [72]. This remodeling maximizes NADPH production and allows them to establish a first line of defense against pathogens rapidly. On the contrary, the PMA-induced oxidative respiratory burst is dominated by glycolysis, and the PPP, NCF1-deficient neutrophils switch from energy metabolic pathways to inflammatory- and immune-related metabolism pathways associated with histidine, glycine, serine, and threonine signaling, etc. [[73], [74], [75], [76], [77], [78], [79]]. This leads to functional abnormalities and pro-inflammatory neutrophil phenotypes. Overall, these data demonstrate the remarkable metabolic flexibility of ROS-driven neutrophils. Neutrophils with intense oxidative bursts tend to form NETs to capture, immobilize, and destroy pathogens and inflammatory mediators [80]. They operate via glycolysis and the PPP pathway, whereas NCF1-deficient neutrophils enhance inflammatory responses by modulating the metabolism of critical proteins via inflammatory signaling pathways. Thus, the energy pathways by which ROS drives neutrophil conversion to a pro-inflammatory phenotype are essential for elucidating neutrophil involvement in RA (Fig. 5H).

This study supplements conventional research and, for the first time, investigates the role of neutrophil ROS at different inflammation stages. The development of novel strategies to detect and treat neutrophils based on ROS effector molecules will provide effective and safe guidance for early intervention and treatment of RA, and other autoimmune diseases.

CRediT authorship contribution statement

Tao Chen: Writing – review & editing, Writing – original draft, Project administration, Methodology, Funding acquisition, Formal analysis, Data curation, Conceptualization. Zhen Zhou: Methodology, Data curation. Yi Liu: Methodology, Conceptualization. Jiayi Xu: Methodology, Data curation. Chenxi Zhu: Methodology, Formal analysis. Rui Sun: Methodology, Data curation. Huifang Hu: Methodology, Data curation. Yan Liu: Data curation. Lunzhi Dai: Data curation. Rikard Holmdahl: Writing – review & editing, Funding acquisition. Martin Herrmann: Writing – review & editing. Lulu Zhang: Writing – review & editing. Luis E. Muñoz: Writing – review & editing, Funding acquisition. Liesu Meng: Writing – review & editing, Project administration, Funding acquisition, Conceptualization. Yi Zhao: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.

Data availability

All scRNAseq data generated in this study have been deposited at NCBI's Gene Expression Omnibus (GEO) repository and are accessible through GEO Series accession number GSE268409. Metabolomic data used in this publication has been deposited to the EMBL-EBl MetaboLights database with the identifier MTBLS10179. The complete data set can be accessed at https://www.ebi.ac.uk/metabolights/MTBLS10179. Source data are provided with this paper.

Funding

This work was supported by the Sichuan International Science and Technology Cooperation Project (2024YFHZ0231, 2024JDHJ0044), Science Popularization Base Project of Chengdu Science and Technology Bureau (2022-GH03-00003-HZ), West China Hospital 1.3.5 Project for Disciplines of Excellence (21HXFH002), West China Hospital Hospital-Enterprise Cooperation Clinical Research Innovation Project (21HXCX004), Post-doctor Research Fund of West China Hospital, Sichuan University (2024HXBH038), the National Natural Science Foundation of China (82402111, W2431021, 82171784), Vetenskapsrådet (2019-01209), Cancer foundation (222350Pj01H) and Leo foundation (LF–OC–22-001023), Deutsche Forschungsgemeinchaft DFG MU 4240/2-1 (470134687) and the International Collaborative Project of Science & Technology Department of Sichuan Province (2022YFH0023). We also acknowledge financial support by Deutsche Forschungsgemeinschaft and Friedrich-Alexander-Universität Erlangen-Nürnberg within the funding programme “Open Access Publication Funding”.

Declaration of competing interest

We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work. There is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of the manuscript entitled.

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

Data will be made available on request.