Introduction
Bronchopulmonary dysplasia (BPD), a chronic pulmonary disease, is the most common factor affecting mortality and long-term morbidity in premature infants [
1]. In recent years, with the application of prenatal glucocorticoids, postnatal alveolar surfactant (PS) replacement and early protective ventilation strategies, the mortality rate of premature infants has decreased significantly, but the incidence of BPD has shown an upward trend as the youngest and sickest patients are now surviving[
2]. Surviving children are prone to respiratory and non-respiratory comorbidities, and the long-term effects can continue to childhood, adolescence or even adulthood, which seriously endangers the health and quality of life of premature infants [
3,
4]. BPD is a major challenge for neonatal intensive care unit in China, and its related long-term complications will become a serious disease burden in China. Therefore, identifying early predictors can lead to the development of targeted therapies that may reduce the neonatal and infant mortality rate and lessen the disease burden that may last through childhood and into adulthood.
The diagnosis of BPD is mainly based on preterm oxygen dependence [
5]. The diagnosis is possibly made at 28 days, but may be up to 3 months after birth depending on what gestational age the patient was born. Due to the continuity and complexity of pathogenic factors and the particularity of diagnostic criteria, its special clinical manifestations are often lagging behind, resulting in the difficulty of early diagnosis [
6]. Therefore, early prediction and timely intervention would be greatly beneficial as this could assist in targeting specific therapies, thereby reducing the risk of developing BPD or reducing the short- and long-term burden of disease. In addition, the mechanism of hyperoxia-induced lung injury in premature infants is closely related to the imbalance of oxidative stress response, excessive activation of cytokines, nitric oxide, neutrophils, and changes in alveolar surfactant [
6,
7]. The changes of these specific protein markers already exist in the body at the early stage of BPD, which makes it possible to explore the early prediction of BPD from the perspective of proteomics.
In this study, we aimed to develop a model based on protein expression from the blood samples during their first week of life to predict the occurrence of BPD in premature infants. The predictive model could help clinicians provide timely intervention and choose appropriate therapeutic approaches for premature infants with a high risk of BPD.
Discussion
With the deepening of scholars’ research on BPD, it is found that the pathogenesis of BPD involves the complex interaction between genetic factors and environmental factors, which is based on genetic susceptibility by hyperoxia, mechanical ventilation, intrauterine and postnatal inflammation, and other factors. Previous study has confirmed that the non-synonymous gene mutation may be related to the occurrence of BPD by sequencing the gene exons of 100 children with BPD and the control group [
13]. Also, recent studies have highlighted the role of mRNA, non-coding RNA, proteins, microbiome and metabolites in BPD through omics approaches [
14]. As proteins are the direct embodiment of life phenomena, the study of protein structure and function should clarify the mechanism of life changes under physiologic or pathological conditions. Therefore, exploring the crucial proteins and related-pathways are important to understand the pathogenic mechanisms and further develop novel therapeutic strategies of BPD.
Previous studies have shown that the development of BPD is characterized by inflammation, extracellular matrix remodeling and apoptosis, and is closely related to the disorder of growth factor signal transduction [
15]. Recent studies have also confirmed the role of transforming growth factor-β (TGF-β) in BPD development, by promoting fibrosis of lung cell [
16,
17]. A large number of animal studies have shown that increased TGF-β level can be observed in animal models of BPD induced by hyperoxia exposure and prenatal inflammation, and obvious fibrosis and inhibited alveolar development can be seen in the early postnatal period (7–14 days) [
18‐
21]. In contrast, inhibiting TGF-β with neutralizing antibody alleviates these changes seen in hyperoxic BPD [
18,
19]. Pathways enrichment analysis based on differentially expressed proteins demonstrated the TGF-β signaling pathway regulated by DEPs in BPD patients, which is consistent with the conclusion of other studies that the TGF-β signaling pathway plays an important role in the development of BPD.
Animal and in vitro studies found that the protective effect of FGF-7 decreased significantly after gene silencing inhibition of AKT [
22,
23]. In addition, in vitro experiments found that PI3K inhibitor could counteract the cell proliferation induced by FGF-7, accompanied by a decrease in PI3K and AKT level [
24]. Therefore, it is speculated that PI3K-Akt signaling pathway may be involved in the occurrence and development of BPD. In the present study, pathways analysis based on protein levels among patients showed PI3K-Akt signaling pathway was significantly altered in BPD. Additionally, HIF-1 signaling pathway, Cytokine-cytokine receptor interaction, Rap1 signaling pathway, Ras signaling pathway and Toll-like receptor signaling pathway differed significantly between BPD and NO-BPD groups, which suggested that these biological function changes played an important role in the occurrence and development of BPD. In subsequent BPD studies, we should focus on these pathways.
WGCNA is a system biological analysis method used in genomics, proteomics and metabolomics to describe the expression correlation of corresponding genes, proteins, or metabolites, which can cluster molecules with similar expression patterns and analyze the association between modules and specific shapes or phenotypes [
25]. Therefore, it is widely used in the study of disease and other traits and molecular correlation analysis [
26‐
28]. Here we combined WGCNA and DEPs screen method for data dimension reduction and hub proteins identification. And 59 hub proteins were found, which indicated these proteins were involved in the evolution of BPD. Current diagnostic approaches mainly based on clinical definitions, imaging modalities, and biomarker data are limited by subjectivity, high radiation exposure, and patient cooperation [
5]. Interestingly, blood-based biomarker tests are cost-effective and can be assayed easily and quickly with minimal sample volume. Thus, they have the potential to greatly aid the disease diagnostics.
In the present study, we selected 8 predictors for BPD prediction through LASSO regression, a method which has been widely used in feature selection and model development [
29,
30]. Our model could achieve an of AUC (95% CI) 1.00 (0.99-1.00) and 0.96 (0.90-1.00) in training cohort and test cohort, respectively. These results suggested that these proteins have important value in BPD development. Finally, we identified 8 proteins in BPD patients, including Cystatin M, a1 Antichymotrypsin, Elafin, sE Selectin, TARC, SRCN1, Protein C and Persephin. Studies have characterized the protective role of activated Protein C in anti-inflammatory, anti-apoptotic activities and endothelial barrier stabilization [
31]. Ina Rudloff et al. have found that Protein C could reduce the lung structural damage induced by BPD, with significant effect in IL-1b, IL-1Ra, IL-6 decrease [
32]. Elafin is a serine protease inhibitor that could bind to extracellular matrix (ECM) proteins thereby protecting against injury induced by sustained inflammation [
33]. Animal study has found that the Elastin gene-deficient mice die soon after birth from developmental defects including loss of alveolar separation and airway branching [
34,
35]. Wenli Han et al. have reported Elafin could inhibit elastase and activation of the TGF-β1 signaling cascade, hence ameliorating apoptosis, inflammation and the elastin organization in the alveoli based on O
2-induced lung injury model in neonates [
36]. Elevated serum levels of sE-selectin have been found in a variety of other inflammatory conditions including asthma and septic shock, which was closely associated with disease severity and outcome [
37‐
39]. Also, study has revealed that higher cord blood levels of sE-selectin in the tracheal aspirate at birth are associated with increased risk of BPD development in preterm infants [
40]. These results suggested that the protein model composed of these proteins for BPD prediction was interpretable.
In our work, some limitations need to be acknowledged. First, the gestational age differed significantly in the no-BPD and BPD groups, thus the effect of time on the expression of proteins could affect the generalizability. Thus, the robustness and generalizability of the eight-protein model requires further optimization and validation in large-scale prospective multi-center cohorts. Secondly, although experimental data of the correlation between some of the proteins identified in this paper and BPD have been reported in academic literature, more evidence is needed to elucidate the inherent correlation and specific function mechanism between the eight-protein and the occurrence and development of BPD. Despite these drawbacks, our results demonstrated valuable information on the importance and significance of the eight-protein model in early prediction and evaluating BPD in high risk premature infants.
In conclusion, our study revealed significantly altered and hub proteins in the blood of premature infants with BPD through WGCNA and differential protein analysis, and identified significantly enriched pathways in the disease, which could provide potential targets for the treatment of BPD and novel clues for understanding the pathogenesis of the disease. As such, we proposed a blood-based reliable model for BPD prediction, which could efficiently and conveniently predict and evaluate BPD in high risk premature infants and further provide pathways to target to prevent or lessen the severity of BPD.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.