Introduction

Over the past few decades, significant advancements in perinatal and neonatal intensive care have been achieved1. However, bronchopulmonary dysplasia (BPD) continues to be a leading cause of morbidity in premature infants2, as well as a contributor to long-term respiratory complications3,4. Infants with moderate-severe BPD are at a higher risk of in-hospital mortality and postnatal growth failure, and have a poorer prognosis for both respiratory health and long-term neurodevelopmental outcomes compared to those with no or mild BPD5,6. The pathology of BPD is highly heterogeneous, encompassing regions of decreased alveolarization, cystic emphysema, fibrosis, and variable airway injury7. These characteristics are associated with persistent lung damage, which is mediated by pulmonary inflammatory mechanisms triggered through various factors such as oxygen therapy, intrauterine inflammation, and nosocomial infection8,9,10. Lung macrophage activation was identified as necessary and sufficient to disrupt normal lung developmental programs in experimental mouse models of BPD11,12. In addition, resident alveolar macrophages have been identified as master regulators of arrested alveolarization in experimental BPD13. In the hyperoxia-induced mouse model of BPD, lung macrophages polarized toward the M1 phenotype, thereby increasing the activity of pro-inflammatory factors such as interleukin-6 (IL-6) in the lungs and inhibiting the survival of type II alveolar epithelial cells8,14. These results from animal experiments strongly showed that lung macrophages (tissue-resident and alveolar macrophages) are key players in the inflammatory lung environment15,16. In clinical samples, it has been found that the inflammatory mediator genes of lung macrophages were more highly expressed in patients on the first day of life who eventually developed BPD than in other patients17. However, the key inflammatory cytokine that initiates persistent inflammation in the airway in premature infants with moderate-severe BPD, and IL-6 upstream anti-inflammatory therapeutic regimens targeting multiple inflammatory factors have not been discovered.

The aim of this study was to investigate the mechanisms underlying airway persistent inflammatory responses in premature neonates with BPD, offering new therapeutic avenues for the treatment of this disease.

Methods

Study design

Clinical samples were collected from May 2021 to December 2022 in the Neonatal Intensive Care Unit of the First Affiliated Hospital of the University of Science and Technology of China (USTC). The study population consisted of preterm infants born before 32 weeks of gestation and birth weight < 1500 g who required mechanical ventilation (the timeline of this study is presented in Figure S1). Preterm neonates with severe pneumonia, sepsis, pulmonary and cardiovascular developmental abnormalities, necrotizing enterocolitis, and Grade III/IV intracranial hemorrhage, as well as those who received antibiotics and/or corticosteroids in their first two weeks of life, were excluded. Patients requiring oxygen for 28 days were diagnosed with BPD, according to the 2001 National Institute of Child Health and Human Development (NICHD) criteria. Patients with mild BPD were able to breathe room air by 36 weeks of corrected age, while those with moderate BPD required a supplemental fraction of inspired oxygen (FiO₂) < 0.30 at 36 weeks of corrected age. Patients who required FiO₂ >0.30 or any positive pressure support at 36 weeks of corrected age were diagnosed with severe BPD18. Meanwhile, we supplemented analysis based on 2019 Jensen criteria, which define BPD as requiring any respiratory support at 36 weeks PMA (for infants < 32 weeks GA), with severity stratified exclusively by respiratory support modality, regardless of FiO₂19. Details are listed in Table S1.

To investigate the inflammatory response mechanisms and their relationship with disease severity, it was initially planned to collect tracheal aspirate (TA) samples from a cohort of premature infants in a single center. This cohort was subjected to two comparative analyses. Firstly, a baseline of clinical features between non-BPD and BPD groups was compared, followed by further analysis between mild and moderate-severe groups to identify clues related to the exacerbation of BPD. Then, multicolor flow cytometry was employed to analyze differences in immune cell composition and phenotype within the TA samples, while transcriptome sequencing was used to assess the cellular functional potential. Techniques such as cytokine arrays and enzyme-linked immunosorbent assay (ELISA) were utilized to confirm the presence of inflammatory factors. Finally, a hyperoxia-induced inflammatory model of neonatal rat lung tissue injury was established to simulate BPD, allowing for the evaluation of potential clinical value of inflammatory factor or pathway inhibitors in treating BPD.

Study approval

This clinical study was approved by the ethics committee of the First Affiliated Hospital of the USTC (approval numbers: 2021KY-080-A and 2022-BE(H)-134). Written informed consent was provided by the parents/guardians for the use of clinical data and samples of preterm infants in this study. The animal experiment was approved by the Experimental Animal Ethics Committee of the First Affiliated Hospital of the USTC (2022-N(A)-166).

TA sample collection

Preterm neonates with GA < 32 weeks and BW < 1500 g, intubated for mechanical ventilation due to respiratory distress syndrome, were enrolled. TA samples were collected from these patients as part of routine respiratory care. The initial sample for each patient was obtained 96 h after intubation on their first day of life, with subsequent samples collected continuously until extubation as previously described20. Briefly, with the infant in the supine position and the head turned to the left, a 6-F catheter (KG Corporation, Shanghai, China) was carefully advanced through the endotracheal tube until resistance was met, and 1 mL/kg of sterile saline was instilled. The fluid was subsequently aspirated back into a sterile sputum collector connected to the catheter. Heart rate, respiratory rate, and oxygen saturation were monitored simultaneously. The lavage solutions were transferred to sterile Eppendorf tubes and centrifuged at 400 g for 10 min to separate the cells from the supernatant. The supernatant was stored at −80 °C for further experiments. The sediment was passed through a 200-mesh filter, and part of the sample (20 μL) was taken for absolute cell counting, while the remainder was stained and analyzed by flow cytometry.

Flow cytometry

Antibodies and reagents used for flow cytometry are listed in Table S2. The method has been previously described21,22. Cells collected from the TA of preterm infants were stained with fluorochrome-conjugated antibodies. Surface staining was performed for 20 min in the dark, followed by washing with phosphate-buffered saline (PBS) and centrifugation at 450 × g and 4 °C for 5 min. After extracellular staining, the cells were fixed and permeabilized for 30 min, followed by intracellular staining for 30 min in the dark at 4 °C. Cells were subsequently acquired on a flow cytometer (NovoCyte Flow Cytometer 3130; Agilent Technologies Inc., Hangzhou, China). Data were analyzed using NovoExpress software provided by Agilent Technologies Inc.

Magnetic-activated cell sorting (MACS) of CD14+ cells

We purified CD14+ cells from TA and blood samples collected from preterm infants by MACS, according to the instructions provided by the manufacturer (Miltenyi Biotec CD14 MicroBeads, Cat# 130-050-201, Cologne, Germany). Following magnetic bead separation to acquire CD14+ cells, a small sample was stained with anti-human-CD14 antibody and subjected to flow cytometry analysis, revealing a cell purity ≥ 90%.

RNA sequencing (RNA-seq)

Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, California) according to the instructions provided by the manufacturer. RNA purity and concentration were evaluated using the NanoDrop 2000 spectrophotometer (Thermo Scientific, Waltham, Massachusetts). RNA integrity was evaluated using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Next, the RNA libraries were constructed using the VAHTS Universal V6 RNA-seq Library Prep Kit according to the instructions provided by the manufacturer. The transcriptome sequencing was performed by OE Biotech Co., Ltd. (Shanghai, China). Principal Component Analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE) were performed.

Free DNA extraction and quantification

The frozen TA supernatants (at −80 °C) were thawed at room temperature and centrifuged at 800 × g for 5 min. Double-stranded DNA (dsDNA) was extracted from 500 μL of supernatant according to the instructions provided in the VAHTS Serum/Plasma Circulating DNA Kit N902 (Vazyme Biotech Co., Ltd., Nanjing, Jiang Su Province, China). The dsDNA concentration was quantified using the PicoGreen dsDNA Kit (Cat# P9740, Beijing Solarbio Science & Technology Co., Ltd., Beijing, China), according to the instructions provided by the manufacturer.

Cytokine arrays

Protein chip array detection from TA supernatants (500 μL) was performed according to the instructions provided in the RayBio® Human Cytokine Antibody Array 5 (G-Series) kit with the assistance of RayBiotech Co.in Guangzhou, China. The cytokine arrays include 80 cytokines as follows: ENA-78, TNF-b, PARC, MCP-4, IL-8, IL-6, MCP-1, MCP-3, GM-CSF, MIP-3 a, IGFBP-1, Eotaxin-2, IL-1b, GRO, NAP-2, Osteoprotegerin, FGF-7, IL-13, GCSF, TNF-a, EGF, Eotaxin, IL-1a, IL-7, GCP-2, TGF-b1, HGF, BLC, IL-2, MIP-1b, VEGF, IFN- g, TIMP-2, Eotaxin-3, PIGF, Leptin, IL-10, IL-15, IGFBP-4, TGF-b3, TARC, MCSF, PDGF-BB, GRO-a, Angiogenin, IL-12 p70, Flt-3 Ligand, Thrombopoietin, RANTES, NT-3, IL-5, GDNF, MIP-1d, IP-10, IGFBP-2, IL-16, MIF, Ckb 8-1, IGF-I, MDC, Oncostatin M, TIMP-1, BDNF, Osteopontin, SCF, FGF-6, LIGHT, Fractalkine, IL-3, TGF- b 2, MIG, IL-4, NT-4, MCP-2, LIF, FGF-9, IGFBP-3, SDF-1, FGF-4 and I-309.

ELISA

The concentrations of the following soluble cytokines in TA supernatants were examined by ELISA, according to the instructions provided by the manufacturer (as shown in Table S3).

A neonatal rat model of hyperoxia-induced BPD

Twelve healthy adult pregnant specific pathogen-free Sprague–Dawley rats (weight: 250–300 g) were raised in cages (one rat per cage). The room temperature and humidity were regulated in the range of 22 ± 2 °C and 55 ± 10%, respectively. Each pregnant rat could deliver 8–10 pups. Newborn Sprague–Dawley rats were randomly divided into four groups: air group (pups were arranged in cages full of ambient air from birth); hyperoxia group (pups were arranged in cages of 85% O2 and received PBS containing 5% trehalose from the fourth day after birth); granulocyte-macrophage colony-stimulating factor (GM-CSF) group (pups were arranged in cages of 85% O2 immediately after birth and received rat recombinant GM-CSF (rGM-CSF) dissolved in PBS containing 5% trehalose from the fourth day after birth); and baricitinib group (the pups were subjected to hyperoxia with rGM-CSF and treated with baricitinib). The newborn rats in the air group were fed in a normal air room (21% O2), and the newborn rats in other groups were exposed to 85% O2 immediately after birth to mimic BPD23. Animal experiments were repeated thrice. Rat body weights were recorded daily. The oxygen flow in the oxygen chamber was 1–3 L/min. The oxygen concentration curve in the chamber was recorded using a medical digital oxygen monitor over a 24-h period. On day 4 of the experiment, 1 μg of rat rGM-CSF was administered intranasally for 5 consecutive days to rats in the GM-CSF group, while rats in the baricitinib group received 5 μg/g oral baricitinib (diluted in PBS) for 7 consecutive days in addition to rGM-CSF. Dams were rotated between room air and hyperoxia every 24 h to avoid oxygen toxicity. On day 11, rats were euthanized and their lung tissues were extracted and placed in 4% formaldehyde. For the animal experiments, all procedures were performed in accordance with the Declaration of Helsinki for the use and care of animals.

Hematoxylin and eosin staining and immunohistochemistry

The left lungs of BPD model rats were fixed overnight in 4% paraformaldehyde at 4 °C and processed by successive dehydration with an alcohol gradient and xylene. The tissues were subsequently embedded in paraffin and cut into sections (thickness: 4 µm) for hematoxylin and eosin staining or immunohistochemistry. Immunohistochemistry was carried out to determine the infiltration of inflammatory cells in alveolar tissues. The tissue sections were analyzed by two experienced blinded pathologists. The F4/80-positive area and myeloperoxidase (MPO)-positive area were calculated by ImageJ software (National Institutes of Health, Bethesda, MD).

Statistical analysis

Continuous variables are presented as means ± standard error and analyzed using Student’s t test (normally distributed data, according to the Shapiro–Wilk test) or median (range) and analyzed using the Mann–Whitney test between two groups (non-normally distributed data). Categorical variables are presented as n (%) and tested using the chi-squared test or Fisher’s exact test. One-way analysis of variance was used for comparisons between multiple groups (normally distributed data). The Kruskal–Wallis test was used for comparisons among multiple groups (non-normally distributed data). Two-tailed P-values < 0.05 indicate statistically significant differences. Statistical analysis was performed using GraphPad Prism v.9.3 software (GraphPad Software Inc., San Diego, CA).

Results

CD14+CD16+ inflammatory monocytes were increased in TA obtained from infants with BPD

The baseline characteristics of preterm infants are summarized in Tables 1–4. Our primary focus was on the cellular composition and phenotype to investigate the immune characteristics of infants with BPD. We observed extremely low percentages of CD3−CD56+, CD3−CD19+, CD4+, and CD8+ subsets in the TA samples obtained from infants with BPD on days 4 and 5 of life (Figure S2A–D). This evidence suggests that T and B lymphocytes play a limited role in the pathology of early-onset BPD. Meanwhile, significantly increased frequencies of CD14−CD16+ neutrophils were observed in the BPD group compared to the non-BPD group on days 4–5 of life (Figure S2E). Yet we know that neutrophils are scarcely present in the alveoli under the physiological state, which suggests the potential local presence of neutrophil chemokines in the alveoli. Although the expression of CD14+CD68+ macrophages showed an upward trend in the non-BPD group, it did not reach statistical significance (Figure S2F).

Table 1 Clinical characteristics of Infants with BPD and non-BPD.
Full size table
Table 2 Clinical characteristics of Infants with BPD and non-BPD.
Full size table
Table 3 Clinical characteristics of Infants with mild BPD and moderate-severe BPD.
Full size table
Table 4 Clinical characteristics of Infants with mild BPD and moderate-severe BPD.
Full size table

We observed that the frequencies and counts of CD14+CD16+ inflammatory monocytes in the TA of patients were higher in the BPD group than the non-BPD group (Fig. 1a) indicating the potential involvement of these cells in the pathology of BPD. Similarly, the frequencies of CD14+CD163+ macrophages (Fig. 1b) were also elevated in the TA of patients in the BPD group (using 2001 NICHD criteria and 2019 Jensen criteria). Furthermore, comparisons of the cellular composition in TA samples obtained from infants who were later diagnosed with mild or moderate-severe BPD revealed elevated frequencies and counts of CD14+CD16+ cells in the moderate-severe group (Fig. 1c, d). These findings suggest that CD14+CD16+ inflammatory monocytes could be considered as key immune cells in moderate-severe BPD. When infants with BPD were stratified according to the 2019 Jensen criteria, the analysis results of immune cell composition in their TA were consistent with those obtained by stratification based on the classic 2001 NICHD criteria. This indicated that the immune cell composition characteristics identified in this study are not affected by the update of BPD stratification criteria, confirming the robustness of the results.

Fig. 1: Changes in immune cells in tracheal aspirations (TA) obtained from preterm infants.
Fig. 1: Changes in immune cells in tracheal aspirations (TA) obtained from preterm infants.The alternative text for this image may have been generated using AI.
Full size image

a, b Comparisons of the frequencies and counts of CD14+CD16+ inflammatory monocytes and CD14+CD163+ macrophages in TA between the BPD group (n = 40) and non-BPD group (n = 22) in the first week of life using 2001NICHD criteria and 2019 Jensen criteria. c, d Comparisons of frequencies and counts of CD14+CD16+ inflammatory monocytes and CD14+CD163+ macrophages in TA between patients who were eventually diagnosed with moderate-severe BPD (n = 21) and mild BPD (n = 19) or grade 1, 2, 3 using 2001 NICHD criteria and 2019 Jensen criteria separately. e The dynamic changes of frequencies of CD14+CD16+ inflammatory monocytes, CD14-CD16+ neutrophils, CD14+CD163+ macrophages and CD14+CD68+ macrophages in TA during the first month of life. Student’s t test or Mann–Whitney test was used for comparison between two groups. P < 0.05 indicates statistical significance.

Dynamic changes in these cell subpopulations were observed in the transitional TA samples. During the first month of life, the frequencies of CD14+CD16+ and CD14+CD163+ macrophages fluctuated between 5–15% and 15–25% in BPD group, respectively, with a notable decrease observed in the non-BPD group around 1 week after birth. The percentages of CD14−CD16+ neutrophils and CD14+CD68+ macrophages did not show opposite trends between groups in their first days (Fig. 1e).

Monocytes in TA were significantly involved in inflammatory responses of airways from infants with BPD

We then investigated the gene expression dynamics of CD14+ monocytes in TA samples obtained from preterm infants with BPD. Using t-distributed Stochastic Neighbor Embedding (t-SNE) analysis, we identified distinct clustering of CD14+ monocytes transcriptomes in non-, mild and moderate-severe cases of BPD, indicating differentially expressed genes between these groups (Fig. 2a). We observed that while PCA captured the overall variance, the t-SNE projection more effectively resolved the continuous, non-linear transcriptional trajectory associated with disease severity progression in our cohort (Figure S3). Pseudotime analysis revealed three subsets (i.e., C1, C2, C3); of those, C1 was associated with inflammation (Fig. 2b, c). Quality matrices for the bulk RNAseq were shown in Table S4.

Fig. 2: RNA-seq analysis revealed the inflammatory signature of CD14+ monocytes in TA purified by MACS.
Fig. 2: RNA-seq analysis revealed the inflammatory signature of CD14+ monocytes in TA purified by MACS.The alternative text for this image may have been generated using AI.
Full size image

a Dimensionality reduction analysis on the transcriptome information of CD14+ monocytes in TA by t-SNE. b Pseudotime analysis on the transcriptome information of a total of nine samples, which were classified as C1, C2, and C3 subsets. c Pathway enrichment analysis of gene clusters (C1–C3) from the pseudotime trajectory, showing C1 is enriched for inflammatory pathways. d Differentially expressed genes (DEGs) in CD14+ monocytes between the moderate-severe BPD and mild BPD groups are displayed as a heatmap. e GO term functional enrichment analysis on DEGs. f In terms of cytokine-cytokine receptor interaction, the moderate-severe BPD group showed significant upregulation of inflammatory factors (e.g., CSF2, CSF2RA, CXCL1), chemokine genes of the CCL family, and genes related to antigen presentation (e.g., HLA-DRA). The expression of SPP1 was also upregulated. g The expression of genes related to neutrophil activation and chemotaxis (e.g., CXCL8, CXCR1, CXCR2, S100A8, S100A9, and ELANE) were significantly upregulated in moderate-severe BPD. h In inflammatory signaling pathway, CSF2, MIF, PLAU, ICAM1, ICAM2, SPP1, GADD45G, TIMP1, and TNF were found to be significantly upregulated in moderate-severe BPD. CD14+ monocytes from TA samples were collected at 5–15 days after mechanical ventilation for RNA-seq.

Heatmap analysis showed upregulation of pro-inflammatory genes (colony-stimulating factor 2 [CSF2], C-X-C motif chemokine ligand 10 [CXCL10], S100 calcium binding protein A8 [S100A8], S100A9, C-C motif chemokine ligand 3 [CCL3], CCL4, CCL5, and CXCL5) in moderate-severe BPD. CSF2 (GM-CSF) activates macrophages and neutrophils, CXCL10 recruits immune cells, and S100A8/S100A9 (calprotectin) regulates inflammation. Anti-inflammatory genes (interleukin 1 receptor type 2 [IL1R2], IL-10, and IL1R1) were downregulated, indicating a suppressed anti-inflammatory response (Fig. 2d). IL-10 limits immune responses, and IL1R2 acts as a decoy receptor to reduce inflammation. Gene Ontology (GO) term analysis showed enrichment of IL-8 (CXCL8) production, neutrophil chemotaxis, and chemokine signaling pathways in CD14+ monocytes of TA from moderate-severe BPD (Fig. 2e). Meanwhile, significant upregulation of inflammatory factors (CSF2, CSF2RA, CXCL1), chemokines (CCL family), and antigen presentation genes (major histocompatibility complex, class II, DR alpha [HLA-DRA], HLA-DRB1, HLA-DRB5) were displayed and the fibrosis-associated gene secreted phosphoprotein 1 (SPP1) (osteopontin), which is involved in tissue remodeling and fibrosis, was also upregulated in moderate-severe BPD samples (Fig. 2f). Neutrophil activation genes (CXCL8, C-X-C motif chemokine receptor 1 [CXCR1], CXCR2, S100A8, S100A9, elastase, neutrophil expressed [ELANE]) were upregulated, promoting chemotaxis and inflammation, especially peptidyl arginine deiminase 4 [PADI4]. This may lead to the formation of neutrophil extracellular traps (NETs) and subsequently induce toll-like receptor 9 (TLR9) activation in macrophages. The upregulation of CD40 and CD86 suggests activation of TLRs, with increases in Fc gamma receptor Ia (FCGR1A) and FCGR1B, which are essential for phagocytosis (Fig. 2g). In the inflammatory signaling pathway, the upregulation of genes (i.e., CSF2, macrophage migration inhibitory factor [MIF], plasminogen activator, urokinase [PLAU], intercellular adhesion molecule 1 [ICAM1], ICAM2, SPP1, growth arrest and DNA damage inducible gamma [GADD45G], tissue inhibitor of metalloproteinase 1 [TIMP1], tumor necrosis factor [TNF]) aligned with the pathology of severe BPD (Fig. 2h). ICAM1 and 2 are upregulated in leukocyte trafficking24. MIF acts as a pivotal chemokine-like factor governing leukocyte extravasation, sustains the pro-inflammatory phenotype of macrophages, and thereby prolongs and amplifies the innate immune response25,26. Matrix metallopeptidases (MMPs) are involved in tissue degradation with its natural inhibitor TIMP27. Nuclear factor kappa B subunit 1a (NFKB1a) is also a proinflammatory signaling molecule28. PLAU is involved in extracellular matrix degradation 29.

These findings indicate that monocytes-secreted inflammatory factors in the TA from patients with moderate-severe BPD probably converge into an integrated “adhesion–chemotaxis–matrix degradation–inflammatory amplification” axis that precisely steers leukocyte recruitment to inflammatory sites and perpetuates the subsequent immune response, highlighting the dominated role of inflammatory transcriptome profile of CD14+ monocytes in moderate-severe cases. This may provide new insights into the pathogenesis and potential treatment of moderate-severe BPD.

Monocytes upregulated TLR9 expression in the TA obtained from preterm neonates with BPD

To determine whether these CD14+ cells were inherent to local alveoli or recruited from peripheral blood, we detected the expression of CC chemokine receptor 2 (CCR2) in this subset of cells. We found that the proportion of CD14+CCR2+ was higher in the BPD group versus the non-BPD group. This result was in accordance with the RNA-seq profile which showed that CCL2 was upregulated especially in moderate-severe BPD, whereas the proportions of CD14+CCR5+ were not significantly different (Fig. 3a, b). It appears that some of these monocytes may originate from the circulation or bone marrow, rather than from resident populations. Next, we sought to identify the mechanism underlying CD14+ cell activation in BPD. We analyzed TLR expression in TA-derived CD14+ cells. Using flow cytometry, we did not observe significant differences in the expression of TLR2, TLR4, TLR5, and TLR7 in CD14+ cells from TA in the BPD group compared to the non-BPD group (Fig. 3c–f). We observed a significantly higher percentage of CD14+TLR9+ cells in the TA obtained from patients with BPD compared to those without BPD (Fig. 3g). Furthermore, the frequency of CD14+TLR9+ cells was negatively correlated with PaO2/FiO2, an indicator of oxygenation (Fig. 3h). In infants with moderate-severe BPD, TLR9 was more highly expressed on CD14+CD16+CD163+ cells than on the CD14+CD206+ cells (Fig. 3i). We also compared the DNA profiles in the TA obtained from the non-BPD and BPD groups, finding that higher dsDNA concentrations in TA obtained from infants with BPD were positively correlated with the frequency of TLR9+/CD14+ (Fig. 3j, k).

Fig. 3: TLR9 was upregulated in CD14+ monocytes in TA obtained from infants with BPD.
Fig. 3: TLR9 was upregulated in CD14+ monocytes in TA obtained from infants with BPD.The alternative text for this image may have been generated using AI.
Full size image

a The percentage of CD14+CCR2+ cells was higher in TA obtained from patients with BPD than those without BPD. b The percentage of CD14+CCR5+ cells was not significantly different between the two groups. c–f The percentages of TLR2-, TLR4-, TLR5-, and TLR7-expressing CD14+ cells in TA were not significantly different between the BPD and non-BPD groups. g Comparison of the percentage of CD14+TLR9+ cells (analyzed by flow cytometry) in TA obtained from patients with BPD and those without BPD. h The percentage of CD14+TLR9+ cells in TA was negatively correlated with PaO2/FiO2 in patients with BPD. i TLR9 was mainly expressed on CD14+CD16+CD163+cells derived from TA of patients with BPD. j Comparison of free dsDNA concentrations (detected using the Pico Green kit) in TA obtained from patients with BPD and those without BPD at 5–10 days after mechanical ventilation. k The percentage of CD14+TLR9+ cells in TA was positively correlated with the levels of dsDNA in patients with BPD. P < 0.05 indicates statistical significance.

GM-CSF stimulates persistent airway inflammation in infants with BPD

We sought to understand the reason for the persistence of proinflammatory CD14+ monocytes in the TA obtained from infants with BPD and the factors that contribute to their inflammatory signature. We used the protein chip array to detect inflammatory cytokines in TA supernatants and observed differential expression of protein among patients in the non-BPD, mild BPD, and moderate-severe BPD groups. Specifically, the data of the protein chip array showed higher levels of IL-8, GM-CSF, epithelial-derived neutrophil-activating peptide 78 (ENA-78), GRO, monocyte chemotactic protein 1 (MCP-1), TNF-α, and IL-6 in the TA obtained from patients with moderate- severe BPD (Fig. 4a). Subsequently, we used ELISA to further quantify the inflammatory cytokines in the TA. The levels of GM-CSF in TA were markedly elevated, especially in moderate-severe cases; they began to increase on about day 7 after birth, and remained persistently high for the next 7 days in the moderate-severe BPD group. In contrast, the cytokine levels showed a decreasing trend in the non-BPD group in the first 7 days (Fig. 4b, c). Of note, the levels of GM-CSF were negatively correlated with PaO2/FiO2 (Fig. 4d) and positively correlated with the percentage of CD14+TLR9+ cells (Fig. 4e). We again detected GM-CSF expression in CD14+ monocytes of TA by flow cytometry, finding its extremely high level in moderate-severe BPD group (Fig. 4f), illustrating that GM-CSF could be an indicator of moderate-severe BPD. Using ELISA, we found that IL-8, IL-6, MCP-1 and GRO (CXCL1, CXCL2, and CXCL3) concentrations were significantly different among the three groups with increasing trend in moderate-severe groups (Fig. 4g). In addition, we made correlation between these proinflammatory cytokines and oxygen supplement days (from birth to the day of TA collection) finding that GM-CSF, IL-8 and IL-6 levels were positively correlated with oxygen supplement length (Fig. 4h), indicating these cytokines were associated with the pulmonary inflammatory response. Next, we analyzed the expression of these cytokines in the TA obtained from patients with moderate-severe BPD by flow cytometry. GM-CSF was highly expressed in CD14+ cells from the TA and mainly expressed by CD14+CD16+CD163+CD206+ cells. IL-8 and IL-6 were mainly expressed by CD14+CD11c+CD68+CD66+ cells (Figure S4A–C).

Fig. 4: GM-CSF was the key inflammatory cytokine in TA obtained from infants who were later diagnosed with moderate-severe BPD.
Fig. 4: GM-CSF was the key inflammatory cytokine in TA obtained from infants who were later diagnosed with moderate-severe BPD.The alternative text for this image may have been generated using AI.
Full size image

a Protein chip array analysis of the TA supernatants obtained from infants who were later diagnosed with no BPD, mild BPD, or moderate-severe BPD. b Comparison of GM-CSF levels in TA examined by ELISA among patients who were diagnosed with no BPD, mild BPD, or severe BPD. c The levels of GM-CSF (examined by ELISA) in the TA supernatant obtained from patients with moderate-severe BPD increased at 7 days after birth and remained persistently high thereafter. d The levels of GM-CSF in the TA supernatant were negatively correlated with PaO2/FiO2. e The levels of GM-CSF in the TA supernatant were positively correlated with the percentage of CD14+TLR9+ cells in TA. f The proportion of GM-CSF in CD14+ cells in TA was tested by flow cytometry (TA samples were collected at 5–10 days after mechanical ventilation). g Levels of IL-8, IL-6, MCP-1 and GRO in TA were examined by ELISA. h Correlation between inflammatory cytokines and oxygen supplement days (including mechanical ventilation days).

Baricitinib alleviates hyperoxia-induced lung inflammation in the neonatal rat model of BPD

Since that CD14⁺ monocytes-derived GM-CSF in TA is pivotal to the pathogenesis of BPD, and we have concurrently documented an early accumulation of neutrophils in the TA from BPD infants, we next aimed to elucidate how CD14+ cells, GM-CSF and IL-8 (as revealed by our RNA-seq data) interact to drive the initiation and progression of BPD as GM-CSF and IL-8 function as potent neutrophil chemoattractants.

We used MACS to isolate CD14+ cells from the peripheral blood of healthy neonates and stimulate them with human recombinant GM-CSF (rGM-CSF). We observed that the percentage of CD14+IL-8+ cells increased from 14.23% to 37.25% (Fig. 5a, b). Janus kinase 2/signal transducer and activator of transcription 5 (JAK2/STAT5) signaling is the main downstream inflammatory signaling pathway of GM-CSF30. Therefore, we used the JAK2 inhibitor baricitinib to block the production of IL-8 in CD14+ cells finding that treatment with baricitinib (10 μg/mL) decreased IL-8 production in vitro (Fig. 5c). Next, we tested the effects of baricitinib in vivo using a previously established hyperoxia-induced neonatal rat model of lung inflammation mimicking BPD (Fig. 5d). We found that the daily weight change of newborn rats decreased significantly under hyperoxia with the GM-CSF treatment; however, the weight was restored to some extent after the administration of baricitinib (Fig. 5e). It clearly showed more severe lung inflammation of rats in the hyperoxia with GM-CSF group, as evidenced by thickened alveolar septum and infiltration of alveolar inflammatory cells, as well as an increase in MPO- and F4/80-positive areas, which became lower in the baricitinib treatment group (Fig. 5f–i).

Fig. 5: Baricitinib alleviated lung inflammation in the rat model of BPD.
Fig. 5: Baricitinib alleviated lung inflammation in the rat model of BPD.The alternative text for this image may have been generated using AI.
Full size image

a–b Recombinant human GM-CSF promoted IL-8 production by CD14+ cells purified from neonatal peripheral blood. c Baricitinib reduced IL-8 production by CD14+ cells (purified from neonatal peripheral blood). d Schematic outlining the experiments of the animal hyperoxia lung injury model. e Daily weight change among the different experimental groups. f Images of lung hematoxylin and eosin staining of different groups. g Immunohistochemistry of MPO and F4/80 in lung tissue of neonatal rats from different groups. h–i Comparisons of MPO-positive areas (h) and F4/80-positive areas (i) among different groups. The paired t-test was used for the analysis shown in (b). The Kruskal–Wallis test was used for the analysis shown in (c, h, i). Two-way ANOVA was used for the analysis shown in (e). P < 0.05 indicates statistical significance.

Discussion

In this study, we observed a significant aggregation of CD14+ monocytes in the airways of preterm infants during their early life, particularly in those who later developed moderate-severe BPD. These monocytes displayed a proinflammatory profile, with an even higher expression of inflammatory cytokine genes observed in moderate-severe BPD. GM-CSF, secreted by these airway monocytes, maintained high levels in the TA obtained from patients with BPD during the first two weeks of life. Hence, it emerged as a crucial factor in maintaining persistent airway inflammation. Additionally, treatment with baricitinib significantly reduced the aggregation of macrophages and neutrophils in the alveoli of neonatal rats with hyperoxia-induced lung injury mimicking BPD, thereby could be able to reduce alveolar inflammatory injury. The consistent findings of TA immune cell composition between stratification using the 2019 Jensen criteria and the 2001 NICHD criteria confirm that the identified local pulmonary immune cell composition characteristics are core biological features of infants with BPD. This reflects a high level of robustness and reproducibility of the core results in the present study.

Our data demonstrated that CD14+ monocytes from TA in the BPD group exhibited elevated TLR9 expression compared with the non-BPD group, and this elevation was positively corrrelated with the levels of cell-free dsDNA. The dsDNA likely originated from NETs, as TA samples from the BPD group were enriched with CD14-CD16+ neutrophils capable of releasing NETs containing dsDNA31, which play a pro-inflammatory role in the immune response32. Consistent with our findings that TLR9 expressions enhanced in BPD group, it was reported TLR9 was involved in impairing lung function and upregulate pro-inflammatory cytokines in conditions such as smoking-induced emphysema and chronic obstructive pulmonary disease, primarily through neutrophil accumulation and inflammation promotion33,34. Overall, these data highlight that the inflammatory milieu in the lungs of preterm infants during early life is detrimental to the development and maturation of alveolar cells8, thereby contributing to impaired alveolar development and the associated clinical manifestations observed in BPD.

In addition, our findings indicate that these inflammatory monocytes in the TA obtained from patients with BPD likely originate from CCR2+ monocytes. These monocytes reside in the peripheral blood, constantly patrolling the body for signs of infection or injury. Upon encountering such signals, they migrate to the site of inflammation and differentiate into local macrophages35,36. Consistent with this mechanism, we observed elevated levels of the pro-inflammatory cytokine MCP-1 (also termed CCL2) in the TA supernatant obtained from patients with BPD. This cytokine serves as a high-affinity ligand for CCR2 and has been implicated in pulmonary hypo alveolarization37, a hallmark feature of BPD. Furthermore, in an animal model of BPD, treatment with CCL2 blocking antibody alleviated the lung pathology38. Based on the data, we inferred that CCR2+CD14+ monocytes are recruited to the lung and subsequently differentiate into macrophages, thereby contributing to the inflammatory environment characteristic of BPD.

Activated macrophages drive the inflammatory response by producing a wide range of pro-inflammatory cytokines39. In this study, it inferred that the cytokine GM-CSF produced by proinflammatory monocytes was the core factor of local persistent inflammation of preterm infants with BPD. However, the role of GM-CSF in the pathology of BPD remains unclear due to conflicting findings. The levels of GM-CSF increased suddenly in the bronchoalveolar lavage fluid of patients with BPD by the third postnatal day40. Higher levels of serum GM-CSF were found in BPD infants versus non-BPD infants within 48 h after birth41. But in extremely low birth weight infants, the levels of GM-CSF detected in blood spots collected at different time points were reported not to correlate with the subsequent development of BPD42. In the current study, we were particularly intrigued to find that the concentration of GM-CSF was significantly elevated in the TA obtained from infants who later developed moderate-severe BPD, especially after one week of mechanical ventilation, and maintained at high levels for several days in the first month of life. This evidence suggests that GM-CSF levels are closely associated with the severity of local airway inflammation. GM-CSF is a cytokine that promotes myeloid cell development, maturation, survival, activation, differentiation, and polarization signals to various immune cell subsets43. This cytokine, virtually undetectable in the systemic circulation under homeostatic conditions44, is currently recognized as a multi-origin and pleiotropic cytokine, encompassing both innate and adaptive immunity, and plays a pivotal role in the pathogenesis of numerous autoimmune and inflammatory diseases45,46. Hence, we hypothesized that the sustained presence of elevated levels of GM-CSF is crucial for maintaining the activation and functionality of neutrophils and macrophages in patients who later progressed to moderate-severe BPD, and likely perpetuates a persistent inflammatory microenvironment within the airways of these patients. Consequently, this ongoing inflammation impedes the ability of the neonate to exchange gases in the airways and contributes to the inability to wean off supplemental oxygen.

In addition, our findings align with the “CSF network” hypothesis, which originally proposed the interdependent regulation between GM-CSF and pro-inflammatory cytokines47,48. Within this network, IL-6 functions as a component of an autocrine/paracrine signaling loop49,50 and can enhance the activity of immune cells induced by GM-CSF, promoting inflammatory responses51. Accordingly, we propose that in moderate-severe BPD, a GM-CSF-mediated inflammatory positive-feedback loop could initiate in the airway as we found airway monocytes accumulated and released abundant pro-inflammatory mediators, prominently GM-CSF, IL-6 and IL-8; GM-CSF, in turn, feeds back to sustain pro-inflammatory monocytes activation and survival, thus driving disease progression.

Importantly, the JAK/STAT pathway serves as a central hub for downstream signaling of GM-CSF and IL-6 receptors52,53. The JAK transphosphorylation and the downstream cascade of the STAT5 play a crucial role in mediating GM-CSF receptor signaling54. Using hyperoxia-induced rat model of BPD, we demonstrated that the JAK inhibitor baricitinib effectively reduces myeloid cell infiltration and pulmonary inflammation. These findings are consistent with reports showing that oral administration of baricitinib significantly diminishes the aggregation of inflammatory airway cells in the lungs of rhesus monkeys infected with the novel coronavirus55. Thus, we inferred that baricitinib attenuates hyperoxia-induced lung injury, which was associated with inhibiting the downstream signaling of GM-CSF, which is upstream of the IL-6/STAT3 pathway. And its potential clinical value in BPD requires confirmation through future high-quality clinical trials.

However, the present study has limitations. Further research is needed to determine whether neutrophils produce GM-CSF and to elucidate their interactions with pulmonary stromal cells in conjunction with macrophages. Additionally, the long-term effects of baricitinib on pulmonary fibrosis should be explored and paid more attention in further research.

Conclusions

Our study elucidated the pathogenic mechanism driving persistent pulmonary inflammation in severe BPD, wherein inflammatory monocytes serve as the primary cellular source of proinflammatory cytokines, particularly GM-CSF. Our findings suggest a novel therapeutic option for severe BPD, proposing targeted inhibition of the JAK/STAT signaling cascade downstream of inflammatory cytokine receptors as an anti-inflammatory intervention.