Airway microbial dysbiosis and oxidative mitochondrial DNA damage in the development of bronchopulmonary dysplasia

Abstract

Background

This study investigated the association between airway microbiome composition, oxidative mitochondrial DNA (mtDNA) damage and the development of bronchopulmonary dysplasia (BPD) in preterm infants.

Methods

A prospective cohort study enrolled 82 very low birth weight preterm infants (<32 weeks’ gestation). Tracheal aspirates (TA) were collected at birth and on postnatal day 28. Airway microbial diversity and composition were assessed using 16S rRNA sequencing. Oxidative mtDNA damage was measured using 8-hydroxy-2′-deoxyguanosine (8-OHdG) levels in TA samples. We used PICRUSt2-based metagenome predictions from 16S rRNA gene sequencing of TA samples, with functional pathway annotations based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database.

Results

Infants who developed BPD (n=25) had lower gestational age, birth weight and prolonged ventilatory support (p<0.05). Oxidative mtDNA damage was significantly higher in infants with BPD, particularly in moderate-to-severe cases (p<0.05). BPD was associated with reduced microbial alpha diversity and distinct beta diversity clustering. Infants with BPD exhibited higher relative abundance of Proteobacteria and lower relative abundance of Firmicutes, with enrichment of Stenotrophomonas, Acinetobacter and Serratia (p<0.05). By day 28, KEGG-based functional predictions revealed enrichment in microbial pathways related to bacterial motility proteins, circadian rhythm signalling pathway, MAPK signalling pathway and α-linolenic acid metabolism. Proteobacteria abundance correlated positively with oxidative mtDNA damage (r=0.49, p<0.01).

Conclusions

Airway microbial dysbiosis and oxidative mtDNA damage are strongly associated with BPD severity. Targeting oxidative stress and microbiome modulation may offer potential strategies for BPD prevention and management.

Shareable abstract

A significant association between airway microbial dysbiosis, characterised by increased Proteobacteria and reduced diversity, and oxidative mitochondrial DNA damage in very low birth weight infants who develop bronchopulmonary dysplasia https://bit.ly/46qtkal

Introduction

Bronchopulmonary dysplasia (BPD) is a major cause of respiratory morbidity and mortality in very low birth weight (VLBW) preterm infants. Its complex pathogenesis involves mechanical ventilation, oxygen toxicity, inflammation and arrested lung development [1–3]. Oxidative stress (OS) plays a crucial role in BPD development, arising early in fetal life and persisting postnatally [4]. Preterm infants are particularly vulnerable due to an imbalance between excessive reactive oxygen species (ROS) production and immature antioxidant defenses [5, 6]. This imbalance amplifies inflammation, promotes lung injury and disrupts alveolar development [2, 4, 7]. While the mechanisms linking OS, inflammation and lung damage remain incompletely understood, clarifying these pathways is essential for developing targeted preventive therapies.

The airway microbiome, a diverse community of microorganisms in the respiratory tract, plays a critical role in shaping immune and inflammatory responses in the developing lung [8]. In healthy neonates, microbial colonisation follows a dynamic, sequential pattern, with early exposure to commensal bacteria supporting immune maturation and lung development. In contrast, preterm infants often experience disrupted microbial colonisation due to factors such as antibiotic use, mechanical ventilation and prolonged hospitalisation, resulting in dysbiosis [9]. High-throughput 16S rRNA sequencing studies have identified distinct microbial signatures associated with respiratory diseases such as cystic fibrosis [10] and asthma [11]. In preterm infants, initial airway colonisation is typically dominated by organisms such as Staphylococcus and Ureaplasma [12–14]. In particular, two studies reported significant differences in alpha and beta diversity between preterm infants with and without BPD. The development of BPD has been associated with dynamic shifts in the airway microbiome, including increased abundance of Stenotrophomonas [15] and Gammaproteobacteria [16], which have been implicated in adverse respiratory outcomes and BPD pathogenesis [17].

OS-induced disruptions in the respiratory microbiome can trigger immune dysregulation and impaired lung development, contributing to progressive lung injury [12, 18, 19]. Mitochondria, central to energy metabolism and immune signalling, are particularly vulnerable to oxidative damage in preterm infants receiving prolonged ventilation and oxygen therapy [4, 20]. Our previous research has demonstrated that mitochondrial DNA (mtDNA) damage, reflected by elevated 8-hydroxy-2′-deoxyguanosine (8-OHdG) levels, is strongly associated with sustained inflammation and the progression of BPD [6, 7].

Despite increasing recognition of these mechanisms, precise interactions between microbial composition, metabolic activity and oxidative mtDNA damage remain inadequately understood. To our knowledge, no longitudinal studies have examined airway microbial dynamics in relation to OS-induced mtDNA damage in preterm infants, limiting insights into causal mechanisms in BPD pathogenesis [13]. Additionally, the functional impact of microbial metabolism on host mitochondrial integrity remains largely unexplored [21]. This study aims to investigate the association between airway microbiome composition, metagenomic pathways and oxidative mtDNA damage in VLBW preterm infants who develop BPD. By integrating microbial and OS profiling, it aims to clarify how respiratory dysbiosis contributes to disease progression. These findings may inform novel biomarkers and therapeutic strategies for early intervention in BPD.

Methods

Study design and population

This prospective observational cohort study enrolled 82 VLBW preterm infants (gestational age <32 weeks, birth weight <1500 g) who developed respiratory failure requiring mechanical ventilation at the neonatal intensive care unit (NICU) of Changhua Christian Children's Hospital. Clinical characteristics and tracheal aspirates (TA) from preterm infants were collected at two time points: within the first 3 days of life (DOL 0–3, baseline) and on postnatal day 28 (DOL 28, follow-up), to assess microbial composition and oxidative profiles over time. BPD was diagnosed according to the 2019 Jensen criteria, based on the type of respiratory support at 36 weeks postmenstrual age (PMA) [22]. Mild BPD was defined as the use of low-flow nasal cannula (≤2 L·min⁻¹); moderate BPD as the need for high-flow nasal cannula (>2 L·min⁻¹), continuous positive airway pressure or noninvasive positive pressure ventilation; and severe BPD as requiring invasive mechanical ventilation at 36 weeks PMA. Group labels such as “BPD (baseline)” and “BPD (follow-up)” indicate the infant's BPD diagnosis at 36 weeks PMA and the corresponding time point of sample collection. Infants with congenital anomalies, maternal chorioamnionitis, congenital heart disease, hereditary metabolic disorders or who died within the first 7 days of life were excluded. The attending physicians reached consensus on all major clinical decisions, while management protocols, clinical practices, equipment and key personnel in the NICU remained consistent throughout the study period. All infants were managed by the same group of attending neonatologists in accordance with previously described institutional protocols and standardised care practices [6, 7, 23].

Of the 90 eligible infants, 8 did not participate, yielding a final cohort of 82 infants. An additional 14 were excluded due to congenital anomalies (n=1), severe pneumonia or sepsis (n=2), loss to follow-up (n=4) or poor DNA quality (n=7). The remaining 68 infants were included in the final analysis and classified into BPD (n=25) and non-BPD (n=43) groups. The BPD group was further stratified into mild (n=14) and moderate-to-severe (n=11) subgroups (supplementary figure S1). Clinical and demographic data were extracted from medical records (table 1). The study was approved by the Institutional Review Board of Changhua Christian Hospital (Approval No: 180201), and informed parental consent was obtained before enrolment.

TABLE 1.

Demographic and clinical characteristics of preterm infants on postnatal day 28, stratified by BPD diagnosis at 36 weeks postmenstrual age

	Non-BPD	BPD
Infants, n	43	25
Male sex	23 (53%)	16 (64%)
Gestational age, weeks	29.5 (27.2–31.0)	26.5 (24.5–29.3)*
Birth weight, g	1163 (869–1318)	815 (651–1175)*
Caesarean delivery	36 (83.7%)	19 (76.0%)
Antenatal steroid	35 (81.4%)	21 (84%)
Respiratory distress syndrome	24 (55.8%)	14 (56.0%)
Surfactant use	16 (37.2%)	13 (52.0%)
1-min Apgar score	7 (5–8)	5 (4–6)*
5-min Apgar score	8 (8–9)	8 (7–8)
Ventilator, days	5 (0–15)	11 (7–28)*
Supplemental oxygen exposure days	22 (19–28)	28 (28–28)*
ROP stage III–IV	6 (14.0%)	14 (56.0%)*
PDA	27 (62.8%)	20 (80.0%)
Intraventricular haemorrhage grades III–IV	1 (2.3%)	4 (16.0%)*
Maternal antibiotics in labour	7 (16.3%)	4 (16%)

Data are presented as n (%) or median (IQR). Differences between the two groups were assessed by Fisher exact test for categorical data and Mann–Whitney U-test for continuous variables. BPD: bronchopulmonary dysplasia; ROP: retinopathy of prematurity; PDA: patent ductus arteriosus. *: mean p<0.05.

Neonatal TAs and sample collection

TA samples were collected according to a standardised protocol aligned with American Thoracic Society guidelines [6, 24]. For intubated infants, respiratory therapists instilled 0.5 mL of sterile isotonic saline into the endotracheal tube, followed by manual ventilation with a bag-mask for three breaths. The aspirated fluid was then collected into a sterile mucus trap. For infants successfully weaned from mechanical ventilation by postnatal day 28, TA sampling was performed by an attending physician using direct laryngoscopy. An 8 Fr suction catheter was inserted into the trachea under direct visualisation below the vocal cords to minimise oropharyngeal contamination. Although TA samples have historically been considered susceptible to upper airway contamination, emerging evidence supports their validity for lower airway microbiome analysis. Kalantar et al. [25] demonstrated that microbial profiles from TA samples closely resemble those obtained via mini-bronchoalveolar lavage in patients with bacterial pneumonia. Furthermore, our previous studies have validated both the TA collection methodology and the downstream analytical procedures used in this cohort [6, 7, 23]. All procedures were well tolerated, with no observed adverse events such as desaturation, bradycardia or apnoea. Immediately after collection, samples were transported on ice and processed within 30 min. In the laboratory, TA specimens were centrifuged at 4°C for 10 min at 300 g to separate cellular and microbial fractions for downstream analysis.

DNA extraction and quantification of oxidative mtDNA damage from TA specimens

Genomic DNA was extracted from TA specimens using the AllPure Genomic DNA Kit (AllBio, Taiwan), following the manufacturer's protocol. DNA quality was assessed using a NanoPhotometer NP80 (Implen, Germany) and confirmed by 1% agarose gel electrophoresis. Oxidative mtDNA damage was quantified by measuring 8-OHdG levels using a ΔCt-based quantitative real-time PCR (qPCR) assay. DNA samples (5 μL) were incubated with or without 2 units of human 8-oxoguanine DNA glycosylase (hOGG1) at 37°C for 4 h. hOGG1 removes 8-OHdG lesions, creating abasic sites that inhibit PCR amplification. qPCR was conducted on a LightCycler system (Roche, Germany) using the primers mtF3212 (5′-CACCCAAGAACAGGGTTTGT-3′) and mtR3319 (5′-TGGCCATGGGATTGTTGTTAA-3′). Oxidative damage was expressed as ΔCt (Ct_hOGG1 − Ct_control), where increased ΔCt reflects higher 8-OHdG levels. The assay selectively detects oxidative lesions that impair amplification due to abasic site formation after hOGG1 digestion [26].

Microbiome analysis from TA specimens

Bacterial community profiling was performed by 16S rRNA gene sequencing of the V3–V4 regions using the Illumina MiSeq platform. Amplicon libraries were prepared using primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) on an ABI GeneAmp 9700 thermocycler (Applied Biosystems, USA). PCR conditions included initial denaturation at 95°C for 3 min, followed by 27 cycles of 95°C for 30 s, 55°C for 30 s and 72°C for 45 s, with a final extension at 72°C for 10 min. Each 20 μL reaction contained FastPfu buffer, deoxynucleotide triphosphates, primers, FastPfu DNA polymerase, bovine serum albumin and 10 ng of template DNA. Reactions were performed in triplicate. PCR products were verified by 2% agarose gel electrophoresis, purified using the AxyPrep DNA Gel Extraction Kit (Axygen), and quantified with the Promega QuantiFluor dsDNA Kit [27]. Differentially abundant taxa were initially identified using the linear discriminant analysis effect size (LEfSe) method, with a Kruskal–Wallis p-value <0.05 and a linear discriminant analysis (LDA) score >3.5 indicating a significant contribution to group differences. To further evaluate genus-level taxonomic variation and account for multiple testing, we performed pairwise Wilcoxon rank-sum tests with Benjamini–Hochberg false discovery rate correction. We used PICRUSt2 (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2) software to predict microbial metagenomes from 16S rRNA gene sequencing of TA samples, with functional pathway annotations based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database [21, 28]. Further bioinformatic processing is detailed in the supplementary Methods.

Statistical analysis

Categorical variables were compared using Chi-square or Fisher's exact test. Continuous variables were analysed using the Mann–Whitney U-test or Wilcoxon signed-rank test. Multiple group comparisons employed the Kruskal–Wallis test with Duncan's post hoc analysis. A multivariable generalised linear model with a log link and gamma distribution was used to evaluate predictors of Proteobacteria abundance. Covariates included sex, gestational age, birth weight, 1-min Apgar score, duration of mechanical ventilation, supplemental oxygen exposure, oxidative mtDNA damage (ΔCT) and maternal antibiotic use. Spearman's correlation analysis was used to examine associations between TA 8-OHdG levels and microbial community profiles. Statistical analyses were performed using IBM SPSS Statistics for Windows, Version 22.0 (IBM Corp., Armonk, NY, USA), with a two-sided p-value <0.05 considered statistically significant.

Results

Demographic and clinical outcomes

The study initially enrolled 82 VLBW preterm infants. 14 were excluded due to congenital anomalies (n=1), severe pneumonia or sepsis (n=2), loss to follow-up (n=4) or poor DNA quality (n=7), resulting in a final analytical cohort of 68 infants (supplementary figure S1). Clinical and demographic data collected on postnatal day 28 are summarised in table 1. Based on BPD diagnosis at 36 weeks PMA, infants were categorised into BPD (n=25) and non-BPD (n=43) groups. The BPD group was further stratified into mild (n=14) and moderate-to-severe (n=11) subgroups, as detailed in supplementary table S1. Compared with the non-BPD group, infants with BPD had significantly lower gestational age (median 26.5 versus 29.5 weeks) and birth weight (median 815 versus 1163 g), as well as lower 1-min Apgar scores (median 5 versus 7; all p<0.05). The BPD group also required longer mechanical ventilation (median 11 versus 5 days) and supplemental oxygen exposure (median 28 versus 22 days) and showed a higher incidence of severe intraventricular haemorrhage (grades III–IV: 16% versus 2.3%) and advanced retinopathy of prematurity (ROP stages III–IV: 56% versus 14%; p<0.05) (table 1).

Airway microbial diversity and composition in BPD progression

Airway microbial diversity and composition were compared between infants with and without BPD. Alpha diversity, which represents within-sample microbial richness, was measured using the Shannon index (figure 1a) and found to be significantly reduced in infants with moderate-to-severe BPD compared to those with mild BPD or without BPD (p<0.05). Beta diversity, which reflects between-group dissimilarity in microbial composition, was assessed using principal coordinates analysis (figure 1b). This analysis demonstrated distinct clustering between BPD and non-BPD groups. Notably, infants with moderate-to-severe BPD formed a discrete cluster, indicating a compositional shift associated with disease severity.

FIGURE 1.

a) Alpha diversity (Shannon index) and b) beta diversity (principal coordinates analysis (PCoA)) demonstrate significant differences between bronchopulmonary dysplasia (BPD) and non-BPD groups, with the moderate-to-severe BPD group showing reduced alpha diversity. *: p<0.05. c) Phylum-level and d) species-level relative abundance profiles illustrate distinct airway microbiota compositions by BPD severity. Tracheal aspirates were collected at baseline (days of life (DOL) 0–3) and follow-up (DOL 28).

At the phylum level, infants with moderate-to-severe BPD exhibited a higher abundance of Proteobacteria, with relative abundance positively correlating with BPD severity (figure 1c). At the species level, microbial composition differed significantly between groups. The non-BPD group displayed a more stable and consistent airway microbiome over time, whereas infants with BPD, particularly those with moderate-to-severe disease, exhibited greater temporal variability and instability in microbial profiles (figure 1d).

Airway microbial dysbiosis in preterm infants with varying BPD severity

Ternary plot analysis of airway microbial phyla at baseline and 1-month follow-up revealed distinct clustering patterns, with infants in the BPD group exhibiting greater temporal divergence in microbial composition compared to the non-BPD group (figure 2a). LEfSe analysis at the 1-month follow-up identified significant differences in airway microbial composition between groups. Infants with BPD demonstrated increased relative abundance of Proteobacteria, particularly the Alphaproteobacteria and Gammaproteobacteria classes. In contrast, infants with non-BPD had higher levels of Firmicutes, including the Bacilli class and the Staphylococcus genus (figure 2b, c).

FIGURE 2.

a) Ternary plot illustrating respiratory microbial phyla at baseline (days of life (DOL) 0–3) and follow-up (DOL 28) reveals distinct clustering patterns, especially in the bronchopulmonary dysplasia (BPD) group. b) Linear discriminant analysis effect size analysis comparing microbial profiles in BPD and non-BPD groups at follow-up (DOL 28). c) The BPD group shows increased abundance of Proteobacteria, especially Alphaproteobacteria and Gammaproteobacteria classes, whereas the non-BPD group is enriched in Firmicutes, notably Bacilli class, Staphylococcaceae family, and Staphylococcus genus. Positive linear discriminant analysis (LDA) scores indicate microbial taxa enriched in the non-BPD group (follow-up, DOL 28); negative scores reflect enrichment in the BPD group (follow-up, DOL 28).

Genus-level analysis (figure 3) revealed distinct microbial patterns associated with BPD severity. Stenotrophomonas (figure 3a) and Acinetobacter (figure 3b) were significantly more abundant in infants with mild BPD than in those without BPD (p<0.05). Serratia was notably enriched in the moderate-to-severe BPD group compared to all other groups (p<0.05) (figure 3c). Staphylococcus abundance was significantly higher in the non-BPD (follow-up) group than in all BPD subgroups, with an additional significant difference between non-BPD (baseline) and moderate-to-severe BPD (p<0.05) (figure 3d). In contrast, Bifidobacterium and Lactobacillus showed no significant differences across groups (p>0.05) (figure 3e, f).

FIGURE 3.

Genus-level differences in airway microbial abundance between preterm infants with and without bronchopulmonary dysplasia (BPD) at baseline (days of life (DOL) 0–3) and follow-up (DOL 28). Infants with BPD had significantly higher relative abundances of a) Stenotrophomonas, b) Acinetobacter and c) Serratia compared to non-BPD infants. *: p<0.05. d) Staphylococcus was significantly more abundant in the non-BPD (follow-up, DOL 28) group than in all BPD subgroups, with an additional difference between non-BPD (baseline, DOL 0–3) and moderate-to-severe BPD. *: p<0.05. No significant group differences were found for e) Bifidobacterium and f) Lactobacillus.

Airway microbial metagenome pathway analysis and BPD development

To explore the functional relevance of airway microbial changes, we analysed predicted metagenomic pathways derived from 16S rRNA sequencing data of TAs collected at birth and 1 month later. Longitudinal comparison revealed significant temporal shifts in microbial functional potential among infants diagnosed with BPD, as indicated by LDA scores (log₁₀). For this baseline versus follow-up analysis, infants with mild, moderate and severe BPD were initially pooled to evaluate overall functional trajectory. At birth, infants with BPD exhibited an increased abundance of the drug metabolism pathway. By 1 month, enriched pathways included bacterial motility proteins, circadian rhythm regulation, α-linolenic acid metabolism and the MAPK signalling pathway (figure 4a). To further investigate the impact of disease severity, subgroup analyses comparing mild versus moderate/severe BPD were conducted. In infants with moderate/severe BPD, distinct enrichment of microbial functional pathways was observed, including lipopolysaccharide biosynthesis proteins, nitrogen metabolism, phenylpropanoid biosynthesis, cyanoamino acid metabolism, and metabolism of cofactors and vitamins (figure 4b). In contrast, non-BPD infants showed minimal functional changes over time, with only the shigellosis pathway exhibiting a marginal shift (LDA score=0.11, log₁₀). This suggests that the functional alterations observed are specific to BPD development and progression. Collectively, these findings highlight dynamic, severity-associated microbial functional reprogramming during the early neonatal period.

FIGURE 4.

a) KEGG-based microbial pathways in infants with bronchopulmonary dysplasia (BPD), comparing baseline (days of life (DOL) 0–3) and follow-up (DOL 28); positive LDA scores indicate baseline enrichment, negative scores indicate follow-up. b) Subgroup analysis comparing mild versus moderate/severe BPD; positive scores indicate enrichment in mild BPD, negative scores in moderate/severe BPD. LDA: linear discriminant analysis.

Increased oxidative mtDNA damage in BPD progression

Oxidative mtDNA damage, quantified by ΔCt values reflecting 8-OHdG levels, was significantly associated with BPD severity (figure 5). At both baseline and 1-month follow-up, infants with BPD exhibited markedly higher oxidative mtDNA damage compared to infants with non-BPD (p<0.05), with the most pronounced elevation observed in the moderate-to-severe BPD group (p<0.05) (figure 5).

FIGURE 5.

Oxidative mitochondrial DNA damage in the airway of preterm infants with and without bronchopulmonary dysplasia (BPD) at baseline (days of life (DOL 0–3)) and follow-up (DOL 28). Box plots show 8-hydroxy-2′-deoxyguanosine (8-OHdG) levels in mitochondrial DNA (mtDNA) (ΔCt) across BPD severity groups and non-BPD groups. BPD infants, especially those with moderate-to-severe BPD, exhibit significantly higher oxidative DNA damage at both baseline and follow-up. *: p<0.05.

Correlation between the airway microbiome and oxidative mtDNA damage

Distinct associations were observed between bacterial phyla and oxidative mtDNA damage in infants with BPD and without BPD, assessed at birth and on postnatal day 28 (figure 6). The abundance of Firmicutes D was higher in infants with low to moderate oxidative mtDNA damage but declined significantly at high damage levels. In contrast, Proteobacteria abundance increased in parallel with elevated oxidative mtDNA damage.

FIGURE 6.

Network graphs of airway microbiota composition at the phylum level, categorised into four groups: a) bronchopulmonary dysplasia (BPD) (baseline, days of life (DOL) 0–3), b) BPD (follow-up, DOL 28), c) non-BPD (baseline, DOL 0–3) and d) non-BPD (follow-up, DOL 28). Within each group, microbiota composition was further analysed based on three levels of oxidative DNA damage (8-hydroxy-2′-deoxyguanosine (8-OHdG) in mitochondrial DNA (mtDNA), ΔCt): low, medium and high.

Firmicutes D abundance was inversely correlated with oxidative mtDNA damage (r= −0.52, p<0.01), whereas Proteobacteria showed a significant positive correlation (r=0.49, p<0.01) (supplementary figure S2). Supplementary table S2 presents a multivariable generalised linear model evaluating clinical and microbial predictors of Proteobacteria abundance. After adjusting for relevant covariates, higher oxidative mtDNA damage (mean ratio: 1.923, 95% CI: 1.050–3.520, p<0.05) and longer supplemental oxygen exposure days (mean ratio: 1.199, 95% CI: 1.065–1.349, p<0.05) were significantly associated with increased Proteobacteria abundance. Lower gestational age (mean ratio: 0.626, 95% CI: 0.404–0.970, p<0.05) and ventilator days (mean ratio: 0.923, 95% CI: 0.854–0.999, p<0.05) were also significant predictors. Other variables, including sex, birth weight, 1-min Apgar score and maternal antibiotic use, were not significantly associated. Figure 7 presents a directed acyclic graph outlining the primary causal hypothesis of this study, in which airway microbial dysbiosis and oxidative mtDNA damage are proposed to contribute to the development of BPD in preterm infants.

FIGURE 7.

Directed acyclic graph (DAG) illustrating the primary causal hypothesis: airway microbial dysbiosis and oxidative stress (mitochondrial DNA (mtDNA) damage) contribute to the development of bronchopulmonary dysplasia (BPD) in preterm infants.

Discussion

In this prospective longitudinal study, we characterised dynamic changes in the airway microbiota in relation to BPD severity in VLBW preterm infants, using serial TA samples collected within the first 3 days of life and at 28 days postnatal. Infants who developed BPD exhibited distinct microbiome alterations, including reduced alpha diversity and increased relative abundance of Proteobacteria, particularly in those with moderate-to-severe disease. Previous studies have identified Proteobacteria as a predominant phylum in the faecal microbiota of VLBW preterm infants with BPD, suggesting a potential role of the gut–lung axis in neonatal chronic lung disease pathophysiology [29, 30]. Reduced airway microbiota diversity, driven by an increased abundance of the Proteobacteria, has been correlated with the development of BPD [31, 32]. Additionally, the presence of Stenotrophomonas in TA within the first day of life has been associated with severe BPD progression [15]. Genus-level differences reflected disease severity, with Stenotrophomonas and Acinetobacter enriched in mild BPD, and Serratia significantly increased in moderate-to-severe BPD. While previous reports have associated Lactobacillus with a protective role in BPD [32], no significant differences in its abundance were observed in this study. This underscores the dynamic nature of airway microbial dysbiosis in BPD and highlights the potential of microbiome-targeted interventions to restore balance.

In preterm infants, Staphylococcus (68%) and the vaginal commensal Ureaplasma were identified as the dominant tracheal microbes during the first 3 weeks of life [33, 34]. Infants who later developed BPD exhibited lower Staphylococcus abundance, higher Ureaplasma abundance and greater microbial community instability shortly after birth [33, 35]. In this study, infants with non-BPD displayed higher airway Firmicutes levels, primarily from the Staphylococcus genus, which may be associated with the fact that nearly 80% of preterm infants were born via Caesarean delivery. These findings suggest that specific airway microbial taxa are associated with BPD severity and may serve as potential biomarkers for disease progression, emphasising the value of longitudinal microbiome monitoring in preterm infants.

This persistent increase in oxidative injury over time suggests a sustained burden of OS linked to BPD progression. These findings underscore the role of mitochondrial oxidative damage as a key pathophysiological feature of BPD and support its potential utility as a biomarker of disease severity. A key finding of this study is the significant association between increased Proteobacteria abundance and elevated oxidative mtDNA damage, as indicated by higher 8-OHdG levels in TA samples, even after adjusting for confounding clinical factors including duration of mechanical ventilation and supplemental oxygen exposure. This suggests that microbial dysbiosis may contribute to OS, exacerbating lung injury and inflammation in BPD. ROS can be generated both by hyperoxia exposure and inflammation, particularly in the presence of increased Proteobacteria abundance [32, 36]. Freeman et al. [36] demonstrated that preterm infants with severe BPD exhibited Gammaproteobacteria-predominant dysbiosis, which was associated with elevated endotoxin levels in TA. These increased endotoxin levels may compromise host antioxidant responses and were correlated with worsened lung injury in a hyperoxia mouse model, suggesting a potential link between airway dysbiosis and BPD progression [36]. Our data further support the hypothesis that specific microbial taxa, particularly Proteobacteria, may actively contribute to OS and mitochondrial injury in the developing lung.

Metagenomic pathway predictions provided functional insights into microbial activity over time [21, 37]. Lal et al. [21] reported that TA collected within 6 h of birth from infants who later developed BPD showed alterations in fatty acid metabolism and steroid hormone biosynthesis, which can impact lung development and injury susceptibility. This study examined the relationship between the airway microbiome, microbial metagenomic signatures and BPD risk. Functional metagenomic pathway analysis, predicted from 16S rRNA sequencing of TA at birth, revealed an increased abundance of drug metabolism pathways, which could reflect microbial adaptation to OS. At birth, infants with BPD exhibited an increased abundance of the drug metabolism pathway, possibly reflecting microbial adaptation to OS and representing an early compensatory mechanism in BPD pathogenesis. By 1 month of follow-up, microbial enrichment shifted towards pathways involved in bacterial motility proteins, circadian rhythm, MAPK signalling and particularly α-linolenic acid metabolism. Bacterial motility may facilitate microbial invasion and modulate host immune responses, contributing to persistent airway inflammation. Disruption of host circadian signalling by the microbiota has been implicated in immune dysregulation and lung injury in preterm infants. Additionally, activation of the MAPK pathway is known to mediate proinflammatory cytokine production and OS, both key features of BPD pathogenesis [21, 38]. Enrichment of α-linolenic acid metabolism pathways may reflect microbial-derived lipid signalling that enhances arachidonic acid turnover and the release of inflammatory mediators, potentially exacerbating airway injury in susceptible neonates [23, 39, 40]. Specifically in infants with moderate/severe BPD, the enrichment of lipopolysaccharide biosynthesis proteins suggests an increased microbial capacity to induce pro-inflammatory responses via TLR4 signalling, which may contribute to sustained airway inflammation and lung injury [41]. Nitrogen metabolism pathways, often linked to bacterial nitrification and denitrification processes, can result in the production of reactive nitrogen species, further exacerbating OS in the immature lung [42]. These metabolic changes may indicate host–microbe interactions that contribute to disease progression, highlighting potential therapeutic targets for modulating microbial metabolism in BPD.

Despite these novel insights, the study has several limitations. The challenges of obtaining suitable lower airway samples in preterm infants and processing low-biomass specimens have complicated lung microbiome research. In this study, some TA samples were excluded due to suboptimal DNA quality, further emphasising the difficulties in lung microbiome analysis. Most human microbiome studies are observational, allowing for the identification of correlations between microbial composition and host disease states, but they do not establish direct causation. Factors such as mode of delivery and nutritional interventions, which influence microbiome development, require further investigation to clarify their impact on BPD risk. An important limitation of this study is the potential influence of antibiotic exposure on the airway microbiome. Although maternal antibiotic use was included as a covariate in the multivariable analysis and did not significantly differ across BPD severity groups (supplementary tables S1 and S2), the possibility of residual confounding cannot be entirely excluded. To reduce the impact of infection-related disruption of the microbiome, two infants with severe pneumonia or sepsis requiring systemic antibiotics were excluded from the analysis. Nonetheless, the cumulative effects of antibiotic exposure on microbial diversity and community structure—particularly in relation to BPD severity—remain an important area for future research. Furthermore, mechanistic studies and interventional trials are needed to validate these findings and evaluate the therapeutic potential of microbiome modulation in BPD.

In summary, this study demonstrates a strong association between respiratory microbiome alterations, OS and BPD progression in preterm infants. The identification of microbial signatures linked to disease severity supports the potential utility of microbiome-targeted interventions, such as probiotics, prebiotics and antioxidants. Furthermore, microbial biomarkers associated with disease progression could facilitate early risk stratification, allowing personalised therapeutic approaches in at-risk infants.

Supplementary material

Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.

Supplementary material

00874-2025.SUPPLEMENT

Data availability

Due to Institutional Review Board restrictions at Changhua Christian Hospital, some data are not publicly available. However, all raw sequence data generated in this study have been deposited in a public repository and can be accessed via the following link: https://www.ncbi.nlm.nih.gov/sra/PRJNA1254474. All other relevant data supporting the findings of this study are included in the manuscript and its supplementary files.