Background
During the past several decades, mechanical ventilation (MV) has played an essential role in the clinical management of patients with acute respiratory distress syndrome (ARDS). However, numerous studies have demonstrated that MV is imperfect and can cause ventilator-induced lung injury (VILI) [
1,
2], a common condition that pathologically manifests as an influx of neutrophils, release of inflammatory cytokines, increased alveolar exudation, and non-cardiogenic pulmonary edema [
3].
Several major mechanisms of VILI have been described including barotrauma, volutrauma, atelectrauma, and biotrauma. Barotrauna and volutrauma are caused by alveolar overdistension, and atelectrauma is due to the cyclic collapse/reopening of lung units. Biotrauma is considered to be amplification of the pro-inflammatory cascade based on a pre-existing lung injury [
3,
4]. The translocation of mediators and pathogens from the alveolar spaces into systemic circulation may result in increased alveolar–capillary permeability, pulmonary edema, or even fatal multiple organ dysfunction and death [
4].
Recognition of the importance of VILI has led to a marked conversion on the philosophy underlying the provision of MV. A series of randomized controlled trials were performed to determine feasible ventilation strategies that minimize lung injury. Recently, a lung-protective strategy was validated to reduce VILI following clinical recommendations based on the ARDS Network study [
1,
5‐
7]. This investigation was a landmark study in ventilation development, and underscored the fact that a low tidal volume (V
T) strategy with appropriate positive end-expiratory pressure (PEEP) is necessary to prevent excessive lung stretching during adjustments to MV [
7].
Despite available advances in ventilation strategies, many patients eventually die after surviving the acute phase, often with evidence of pulmonary fibrosis. A prospective cohort study by Martin et al. [
8] reported that in the 64% of ARDS patients diagnosed with pulmonary fibrosis there was a 57% fatality rate, while there were no fatalities in patients without fibrosis. Subsequently, other clinical trials were performed, which showed that ARDS patients with increased levels of transforming growth factor beta 1 (TGF-β1) and procollagen type III had extremely high mortality rates [
9‐
11] than those with lower levels. Consistent with these clinical studies, basic research studies have demonstrated that MV, either of high V
T or high peak airway pressure, can induce lung fibrosis as early as 1 week after injury [
12‐
14]. Therefore, MV is being increasingly recognized as a pivotal factor in the initiation or propagation of ARDS-associated lung fibrosis regardless of the original disease [
8,
9,
15]. Biotrauma is considered to be the major mechanism that sets the stage for the development of subsequent fibrosis. The extracellular and intercellular mediators, either directly released by injured cells or indirectly activated by pulmonary epithelial, endothelial, or immune cells through various cell-signaling pathways, are the key biological forces that drive continuation of the fibroproliferative response [
3,
9]. Recently, Lv et al. [
16] proposed that the endothelial-mesenchymal transition also contributes to abnormal modulation following mechanical injury; however, these observations are still limited in scope.
Global transcriptome analysis is an emerging, powerful tool used to reveal variability in pathophysiology on a genome-wide scale [
17]. This type of analysis is potentially a better way to address the knowledge gap of mechanism for VILI and subsequent fibrosis. In this study, we established a well-defined and standard “one-hit” mouse model of lung fibrosis, and then delineated a complete transcriptome image identifying all involved differentially expressed transcripts (DETs) as well as long non-coding RNAs (lncRNAs) using global transcriptome analysis. To the best of our knowledge, this is the first transcriptome study to reveal a broad spectrum of dysregulated transcripts and potential molecular pathways involved in fibrosis following VILI, as well as the first study to focus on dysregulated or dysfunctional lncRNAs that may be responsible for the pathogenesis of early inflammation or subsequent fibrosis. Together, the results of this study provide novel insights into the inflammatory and fibrotic responses following VILI.
Methods
Animal protocols
Male, 8–12-week-old C57/BL6 mice were randomized to the high V
T MV group (
n = 80) or sham-operated group (sham,
n = 10). After anesthetizing mice with an intraperitoneal injection of pentobarbitone (100 mg/kg; Pfizer, Dublin, Ireland), mice were intubated using a 22 G Teflon catheter and continuously ventilated for 4 h in a volume-controlled mode with a small animal ventilator (55–7040, VentElite; Harvard Apparatus, Holliston, MA, USA). The protocol comprised the following settings: V
T of 20 mL/kg, PEEP of 0 cm H
2O, respiratory rate of 80 breaths/min, inspiratory-expiratory ratio of 1:1, and fraction of inspired O
2 of 0.4 [
18]. During the experimental period, mice were given an anaesthetic as needed; cocuronium besylate (0.6 mg/kg, H20130486; MSD Performance Products, Kenilworth, NJ, USA) was added for muscle relaxation. Anesthetized, intubated, non-ventilated animals served as the sham controls. The study was approved by the Institutional Animal Care and Use Committee of Capital Medical University (No. AEEI-2016-168; Beijing, China) and strictly conducted according to the University’s guidelines.
Specimen collection and processing
Mice were sacrificed by anesthesia overdose at 0, 1, 3, 5, 7, 14, 21, and 28 days. The left lungs were weighed, dried in an oven (60 °C for 72 h), and weighed again to determine the lung wet-to-dry weight ratio [
19]. The right upper lobes were used for hydroxyproline concentration measurement using a hydroxyproline assay kit (A030–2; Nanjing Jiancheng Bioengineering Institute, Nanjing, China), while the middle and dorsal segments were analyzed by quantitative PCR (qPCR) and RNA sequencing, respectively.
Bronchoalveolar lavage fluid preparation
Preparation of bronchoalveolar lavage fluid (BALF) consisted of three separate lavages via a tracheal catheter with 3 mL sterile, ice-cold phosphate-buffered saline (PBS). For each lavage, 0.7–0.8 mL fluids were recovered and separately centrifuged at 1500 rpm for 5 min at 4 °C. Cell-free supernatants were separated for assessment of TGF-β1 using an enzyme-linked immunosorbent assay (ELISA) kit (MB100B; R&D Systems Inc., Minneapolis, MN, USA).
Lung histopathology and immunohistochemistry
After infusion with PBS, lung samples were fixed in 10% neutral formalin for 1 week, and then embedded and sectioned for histological evaluation by hematoxylin and eosin (H&E) staining, Masson’s trichrome, and Sirius technique. Collagen deposition areas and type alterations were calculated to provide more compelling data on fibrosis. The relative expression of alpha smooth muscle actin (α-SMA) was subsequently evaluated by immunohistochemical staining with specific primary antibodies (1:100, A2547; Sigma-Aldrich, St. Louis, MO, USA). The mean density of the positive areas in the sections (at least five random microscopic fields per lung section) was calculated using Image Pro-Plus 6.0 software (Media Cybernetics Inc., Rockville, MD, USA).
Preparation for sequencing
RNA was isolated using TRIzol reagent (15,596,026; Invitrogen, Carlsbad, CA, USA), and the RNA purity and integrity were assessed using the NanoPhotometer spectrophotometer (Implen Inc., Westlake Village, CA, USA) and the Bioanalyzer 2100 RNA Nano 6000 Kit (Agilent Technologies, Savage, MD, USA), respectively. A total amount of 3 μg RNA per sample was used as the input material. After the removal of ribosomal RNA (Epicentre Ribo-Zero TM rRNA Removal Kit; Epicentre, Madison, WI, USA), sequencing libraries were generated from the rRNA-depleted RNA using the NEBNext® Ultra™ Directional RNA Library Prep Kit for Illumina® (New England Biolabs, Ipswich, MA, USA) according to the manufacturer’s instructions. PCR was performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers, and the Index (X) Primer. After clustering of the purified index-coded sampler (TruSeq PE Cluster Kit v3-cBot-HS; Illumina, San Diego, CA, USA), the libraries were sequenced on the Illumina HiSeq 2500 platform, and 125 bp (base pair) paired-end reads were produced.
Read alignment and transcript assembly
Quality control of raw reads was performed with FastQC software (v.0.11.5). After clipping Illumina adapter sequences and trimming low-quality bases through in-house Perl scripts, high-quality clean reads were mapped against the reference genome (Ensembl, release v.87) using Tophat 2.0.9 [
20] with the default parameter settings. Then the resulting aligned reads were subjected to Scripture (beta2) and Cufflinks 2.1.1 to assemble the aligned reads into genes [
21].
Identification of lncRNAs
Cuffmerge was used to merge the assembled transcripts, and those with low expression and shorter than 200 bp were abandoned. The remaining transcripts predicted with coding potential through the CNCI [
22], CPC [
23], Pfam-scan [
24], and phyloCSF [
25] tools were also filtered out, and those without coding potential were identified as the candidate set of lncRNAs.
Quantification and analysis of differentially expressed genes
Cufflinks was used to measure the relative abundance of each transcript by calculating the fragments per kilo-base of exon per million fragments mapped (FPKM) of mRNA (coding gene) in each sample. Differential transcript expression between each pair of samples was analyzed using Cuffdiff (v 2.1.1). The calculated
P values were subjected to the Benjamini-Hochberg method to control for a false discovery rate (FDR) [
26]. Statistical significance was defined as a
P value < 0.05 and estimated absolute Log2-fold change > 1.
Target gene prediction of lncRNAs
Accordingly, we predicted the target genes for lncRNAs using co-location (
Cis) and co-expression (
Trans) analysis [
27]. The co-localization threshold was set as 100 kb upstream and downstream of each lncRNA, and lncRNA targets were identified by the expressed correlation between lncRNA and coding genes, with an absolute value of correlation greater than 0.95.
Functional enrichment analysis
KO-Based Annotation System (KOBAS) 3.0 (
http://kobas.cbi.pku.edu.cn/) is a web server for functional annotation and functional set enrichment of genes [
28]. For the DETs and predicted lncRNA targets, Gene Ontology (GO) and Kyoko Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed on KOBAS 3.0 [
29]. Statistical significance was assessed using the hypergeometric test or Fisher’s exact test with the FDR correction method (Benjamini and Hochberg), and the threshold was set as corrected
P values less than 0.05.
Regulatory network of lncRNAs and mRNAs
An lncRNA–mRNA network was built to identify the interactions between mRNA and lncRNA. The networks were built according to the target relationship between mRNA and lncRNA, as identified by co-location and co-expression analyses. These differentially expressed mRNAs and lncRNAs were retained for network construction and visualized by Cytoscape 3.3 [
30]
Quantitative PCR
Total RNA from the lung tissue was extracted using TRIzol reagent (15,596,026; Invitrogen) and cDNA was generated using the High-Capacity cDNA Reverse Transcription Kit (4,368,814; Invitrogen). The PCR reaction was executed in the iQ5 system (Bio-Rad, Hercules, CA, USA) for 44 cycles, with each cycle consisting of denaturation at 95 °C for 45 s, annealing at 60 °C for 60 s, and an extension at 72 °C for 1 min using Applied Biosystems® Power SYBR® Green (4,367,659; Invitrogen). Primers are shown in the online supplement (Additional file
1: Figure S1). The abundance of each gene was normalized to that of 18S mRNA, and the fold changes were calculated with the 2
-ΔΔCT method [
31].
Statistical analysis
All of the data were analyzed and presented using GraphPad Prism 7.0 (Prism software; GraphPad, San Diego, CA, USA). Comparisons between two groups were determined by the Student’s t-test or Mann–Whitney test, and among multiple groups using one-way analysis of variance (ANOVA). P < 0.05 was considered statistically significant.
Discussion
Currently, “one-hit” and “two-hit” animal models are commonly used to generate a mechanical-based lung injury [
33,
34]. Nevertheless, there are differing opinions on the fibrosis potentiation following ventilation. Cabrerabenítez et al. [
12] demonstrated that high lung stretch with or without hydrochloric acid instillation could result in pulmonary fibrosis on the 8th day, and peaked on 15 days after ventilation. As expected, the “two-hit” model resulted in a greater fibroproliferative response than the “one-hit” model. In contrast, Curley et al. [
13] found that VILI generated marked but transient fibrotic alternations within 24 h, and the anatomical lung structure was restored after 1–2 weeks. This experiment was also supported by Villar et al. [
14], who proposed that lung fibrosis occurred immediately after 4 h high stretch ventilation in a mouse model of sepsis-induced acute lung injury. To address this issue, we monitored our “one-hit” mice for up to 1 month. Pathological staining and measurement of hydroxyproline content and TGF-β1 indicated an overall alteration from early inflammation to fibrogensis. It is noteworthy that fibrosis signature caused by MV was most pronounced on day 7 and was not the same as fibrosis induced by bleomycin [
35]. The fibrosis induced by MV was more likely to manifest as local lesions around the lung interstitium, airway and vessels, which recovered within 1 month.
Based on a well-defined animal model, we performed whole transcriptome analysis to determine the potential molecular mechanisms in lung pathophysiology post-ventilation. Previously, transcriptomics was performed to explore the potential mechanisms of VILI, or protective ventilation use against VILI at a genetic level [
36,
37]. Different from the previous analyses, the current study did not solely focus on studying dysregulated genes or pathways involved in the early stage of VILI. Rather, this was the first transcriptomics study to reveal a broad spectrum of dysregulated transcripts and potential molecular pathways involved in the VILI fibrotic process.
From pairwise comparisons among the sham and VILI groups on day 0 and day 7, a total of 1297 dysregulated DETs were identified after ventilation. We mainly analyzed the expression patterns of the 63 commonly shared DETs, which allowed for a better understanding of the probable biological and regulatory functions of the involved transcripts. For example, the
Hspa1a, Il7r, Sult1a1, Abcg2, and
Stc1 genes exhibited increased activity at the acute injury phase, and returned to baseline levels after 7 days. The expression levels of
Wisp1, Clca1, and
Hic1 showed a rapid initial decline and then increased after 7 days, suggesting their critical role in inflammation, the innate immune response, and tissue repair and resolution. Subsequently, 79 TFs were profiled among the total DTEs, and they showed extensive regulatory functions in organ development, immune response, and homeostasis maintenance; and in lung injury, repair and regeneration. In this study, three commonly shared TFs were detected with statistical significance either in the early inflammation or subsequent phase of fibrosis. These included
MafF, a basic region leucine zipper-type TF important for cellular stress regulation [
38] that was increased 3.36-fold in mice subjected to ventilation and increased 2.17-fold after 7 days.
Hic1, a gene annotated as hypermethylated in cancer 1 that was decreased 2.55-fold in lungs at weaning and returned to baseline levels at increased with 2.33-fold 7 days later. This gene was formerly considered to be a strong candidate as a tumor suppressor gene because it is a master modulator of cell apoptosis and DNA damage survival [
39].
Fosl2, an important gene in the Fos family, was increased 2.99-fold in the VILI group on day 0 compared to the controls, and increased 2.07-fold in the VILI group on day 7 versus day 0. Studies have shown that
Fosl2 is a novel mediator of cell proliferation, differentiation, and transformation in the fibrotic pathophysiology of certain diseases, such as scleroderma and pulmonary hypertension [
40]. The identification of
Fosl2 was not unexpected; however, the two other genes (
MafF and
Hic1) were not previously found in mouse models of lung injury, fibrosis, or in ventilation research. Additionally, other factors identified in the VILI group versus sham group on day 7, such as
Hes2, Ikzf3, and
Mafb, have not been previously reported in similar diseases or models. These TFs may be potential candidate genes for predicting disease mechanisms or intervention targets.
The results also determined substantially enriched pathways that were dysregulated in the respective phases of inflammation and fibrosis. For example, mTOR signaling is a central regulator activated in response to growth factors, nutritional status, and stress signals [
41]. JAK/STAT signaling pivotally regulates cell growth, proliferation, differentiation, migration, and apoptosis [
42]. cAMP signaling mainly functions in cell chemotaxis, immune mediator induction, and inflammatory response regulation, with a correlation to microbial and cardiovascular pathogenesis [
43]. In the current study, these three pathways were all significantly implicated in the regulation of early inflammation. With an intensive focus on fibrosis-promoting pathways, the TGF-β, HIF-1, TLR, and NF-κB signaling pathways were found to be key in fibrogenesis. Accordingly, it is well recognized that overactivation of the TGF-β signaling pathway is one of the most commonly characterized events in the regulation of fibrosis. In addition, TGF-β1 is considered the most central mediator in the activation, proliferation, and differentiation of epithelial cells, myofibroblasts, excessive production of extracellular matrix (ECM), and inhibition of ECM degradation. Apart from canonical (SMAD-based) signaling pathways, non-canonical (non-SMAD-based) pathways such as MAPK, p53, and Notch signaling have also been explored in bleomycin-induced lung fibrosis as well as fibrosis in other organs [
44]. HIF-1 and TLR signaling function as master regulatory signaling pathways in cellular and systemic homeostatic responses to hypoxia as well as natural and acquired immune responses to pathogens, both of which have been increasingly implicated in pulmonary fibrosis [
45]. However, these pathways have not been associated with the pathogenesis of VILI fibrosis.
Importantly, the results of this study contribute to a consistent emerging picture of critical non-coding transcripts involved in lung pathophysiology. LncRNAs are an abundant class of transcripts with no coding function and typical lengths of > 200 nt [
46‐
48]. Over the last decade, many lung diseases have been associated with dysregulated lncRNAs, such as idiopathic pulmonary fibrosis, chronic obstructive pulmonary disease or pulmonary hypertension [
49,
50]. However, to the best of our knowledge, no studies have focused on dysregulated or dysfunctional lncRNAs that may be responsible for the pathogenesis of VILI or subsequent fibrosis. Thus, there is a need to identify functional non-coding transcripts from a vast transcriptome, to provide a better perspective of the molecular mechanisms and therapeutic targets of the disease.
In total, this study profiled 332 differentially altered lncRNAs and well-established lncRNA-mRNA networks in the comparisons between each pair, nearly a quarter of which were novel genes without annotations. Here, 14 common DE lncRNAs that may serve as signals, decoys, guides, or scaffolds for regulating gene expression in lung injury or fibrosis are listed. For example,
1200007C13Rik, a gene that reportedly plays a master function in organ development and repair, may also be an important candidate lncRNA with potential implications in fibrogenesis following ventilation. Other genes, such as
LNC_000027, which is located on chr4 from chr1:74175576–74,180,761 with a predicted target gene of
Cxcr2, is considered to be a master mediator for neutrophil migration and activation during the inflammatory response [
51]. Through target prediction and enrichment analysis, we found that the identified DE lncRNAs likely participate in the processes of cellular and biological regulation through the mTOR, FOXO, MAPK, and cAMP signaling pathways. We speculated that DE lncRNA likely regulates fibrosis through signaling pathways such as PI3K/Akt, TLR, HIF-1, and Wnt signaling. Wnt signaling is a classical developmental pathway required for proper organ development. The overactivation of both canonical and non-canonical signaling has been identified and established in a variety of fibrotic diseases [
52]. Previously, Villar and his co-workers [
18,
53] showed that the Wnt/β-catenin signaling pathway is modulated very early by MV in VILI, and in lung repair without pre-existing lung disease. This suggests an attractive candidate for the prevention and/or management of VILI. However, the majority of identified pathways have not been considered in MV-associate lung fibrosis, and no studies have associated these pathways with the regulatory role of lncRNAs. Therefore, these observations may provide a novel perspective into potential molecular mechanisms for further research.
This study had several limitations. First, our “one-hit” mouse model did not fully represent the real-life “two-hit” conditions seen in humans. It should also be recognized that even if a mouse model used for VILI or fibrosis is validated, it is not fully representative of the classical recommended parameters of ventilation seen in humans. Therefore, bias in comparison to what occurs in humans may develop. Second, our initial results were interpreted mainly based on bioinformatics and literature analysis. It may be necessary confirm our findings using functional or mechanistic studies at the protein level. Nevertheless, the results of this study provide novel perspectives into the potential molecular mechanisms underlying VILI and subsequent fibrosis, providing a foundation for future research studies.