208 - Early Transcriptomic Predictors and Temporal Pathway Shifts in Bronchopulmonary Dysplasia

Publication Number: 208.5016

Alvaro Moreira, The University of Texas Health Science Center at San Antonio Joe R. and Teresa Lozano Long School of Medicine, San Antonio, TX, United States; Lois Randolph, The University of Texas Health Science Center at San Antonio Joe R. and Teresa Lozano Long School of Medicine, San Antonio, TX, United States; Khyzer Aziz, Johns Hopkins Children's Center, Baltimore, MD, United States; Francesca Ricci, chiesi farmaceutici S.p.A, PARMA, Emilia-Romagna, Italy; Matteo Storti, Chiesi Farmaceutici, Parma, Emilia-Romagna, Italy

Background: Bronchopulmonary dysplasia (BPD) is the most common chronic lung disease in extremely preterm neonates. The disease remains challenging due to variable definitions, complex etiology, and incomplete understanding of its mechanisms.

Objective: To identify early gene signatures predictive of BPD and uncover temporal changes in molecular pathways associated with the disease.

Design/Methods: Publicly available DNA microarray and RNA-Seq datasets were retrieved from the National Center for Biotechnology Information. Only genes common across datasets were analyzed. Primary outcomes focused on BPD prediction and pathway identification, while secondary outcomes explored the impact of sex, birthweight, gestational age, and disease severity.
For early prediction, datasets were split into training (70%) and testing (30%) sets. A predictive model was developed using greedy forward search to identify gene sets with the highest AUC, followed by LASSO regression for feature selection. Pathways enriched with differentially expressed genes (FDR < 0.05) were identified.

Temporal analysis included mixed-effects logistic regression, time-series modeling, and eigen decomposition to reduce dimensionality, account for patient variability, and explore gene-time interactions. Gene correlation networks identified clusters of co-expressed genes. All analyses were performed using R version 4.3.2.

Results: The study included 426 neonates across four DNA microarray (437 samples) and four RNA-Seq datasets (327 samples). One dataset from each platform included multiple timepoints (e.g., weeks 1, 2, and 4). A total of 4,643 genes from RNA-seq and 2,217 genes from microarray were differentially expressed, with 1,508 genes overlapping between the two platforms.

Two prediction models, using 44 and 14 genes from microarray and RNA-Seq, achieved AUCs of 0.84 and 0.91, respectively. Predictive genes were linked to immune response, angiogenesis, and gene regulation. Key pathways in week one included ion channel signaling, inflammation, and gene regulation, with later pathways involving oxidative stress, immune activation, DNA repair, and energy metabolism. Severe BPD showed increased B cell activation, Th17 differentiation, and leukocyte chemotaxis, with reduced mitochondrial processing, interferon production, and epithelial apoptosis.

Conclusion(s): Early gene signatures and temporal pathway changes predictive of BPD were identified. Predictive models demonstrated strong performance and offer insights into disease progression.

Study Characteristics
GA-gestational age; BW-birthweight

Temporal analysis of molecular pathways involved in BPD (DNA Microarray and RNASeq)
Using longitudinal transcriptomic data from peripheral blood, we identified 483 differentially expressed genes between infants with BPD (n=48, median GA 26 weeks) and those without (n=29, median GA 30 weeks) over the first month of life (NCBI GSE32472). These genes clustered into three key pathways: T cell activation, neutrophil-mediated inflammation, and oxidative stress. While both groups showed increased T cell activation over time, BPD infants had lower overall expression. In contrast, they exhibited persistently elevated neutrophil activity and oxidative stress across all time points (days 5, 14, and 28), with p<0.05 for all comparisons.

RNASeq analysis from the NCBI GSE220135 cohort (n=109; 15 with BPD, 94 without) confirmed these findings. Between weeks 1 and 2, leukocyte degranulation, neutrophil immunity, and mitochondrial biogenesis were upregulated in BPD infants, but became less prominent by weeks 2 to 4, reflecting a shift from acute immune activation to a more regulated state. Both analyses consistently highlight neutrophil-mediated inflammation, oxidative stress, and immune activation as central to BPD development, validating the role of early immune dysregulation and metabolic changes in preterm infants. These analyses offer a dynamic molecular view of BPD progression, demonstrating how evolving biological pathways over the first month of life.


Gene ontologic pathways comparing severe BPD versus none, mild, and moderate BPD across the first month of life.
Bioinformatic analysis of transcriptomic data reveals gene expression patterns distinguishing severe bronchopulmonary dysplasia (BPD) from other conditions. (Left) Venn diagram showing the distribution of differentially expressed genes across three timepoints (weeks 1, 2, and 3). Genes in the overlapping section (n=1,094) represent those consistently differentially expressed at all timepoints, highlighting a core set of genes associated with severe BPD. (Right) Gene Ontology (GO) pathway analysis of these intersection genes, with bars indicating pathways up-regulated (blue) and down-regulated (red) across all timepoints. Key pathways involve immune response regulation, T-cell differentiation, and leukocyte chemotaxis, emphasizing immune system involvement in the pathophysiology of severe BPD.