Construction of MicroRNA-mRNA Regulatory Network in the Molecular Mechanisms of Bleomycin-Induced Pulmonary Fibrosis

Bleomycin is a common antitumor agent used to treat many different types of malignancies; however, its main side effect is pulmonary fibrosis. The mechanism of bleomycin-induced pulmonary fibrosis (BIPF) has not been fully elucidated. Therefore, to further explore the molecular mechanisms of BIPF, we screened for microRNA (miRNA) and mRNA expression obtained from BIPF samples from the Gene Expression Omnibus database. Subsequently, we identified the differentially expressed miRNAs and genes that overlapped with the differentially expressed miRNAs target genes, predicted by using the miRWalk database selected as a candidate. The candidate genes were visualized based on Gene Ontology and the Kyoto Encyclopedia of Genes and Genomes analyses. A protein-protein interaction network was constructed. Hub differentially expressed genes were selected and corresponding miRNAs to construct a miRNA-mRNA regulation network. Then, we chose three key miRNAs to study their regulatory relationship in bleomycin-induced pulmonary fibrosis. Finally, mouse lung epithelial cells TC-1 and MLE-12 were treated with bleomycin with qPCR to validate the results of three important hub genes and all key miRNAs. And dual-luciferase report experiment was carried out to verify the interaction of mmu-miR-1946a and serpina3n. The results revealed that the imbalance of matrix metalloproteinase-9 (MMP-9) and tissue inhibitor of metalloproteinase-1 (TIMP-1) plays a pivotal role in the occurrence and development of BIPF. In addition, Serpina3n and mmu-miR-1946a were proved interaction and may be involved in the regulation of the balance between MMP-9 and TIMP-1. The experimental results also verify the analysis. Our findings provide new insights into the key mediators and pathways related to the molecular mechanisms of BIPF.


Introduction
Bleomycin (BLM), a glycopeptide antibiotic isolated from Streptomyces, acts as an antitumor agent for many different types of malignancies, including lung cancers [1]. The cure rate of metastatic testicular cancer can be as high as 90% when chemotherapy regimens contain BLM [2]. However, the main side effect of BLM is pulmonary toxicity that can cause pulmonary fibrosis in approximately 10-23% of patients [3], which greatly limits the application of this drug. Some studies have explored the mechanism of BLM-induced pulmonary fibrosis (BIPF), and the experiments of animal pulmonary fibrosis with BLM have also been applicated [4,5]. Nevertheless, the specific mechanism of BIPF has not been clarified. The research on BIPF by bioinformatics method is rare despite the exploration of related genes is helpful to the individualized treatment of patients; the research on BIPF by bioinformatics method is still rare. Consequently, further studies are needed to find the potential approach to overcome this problem.
The interactions between mRNAs and microRNAs (miRNAs) are major regulators of pulmonary fibrosis. Xu et al. [6] demonstrated that miR-326 attenuates silicainduced pulmonary fibrosis by inhibiting inflammation and promoting autophagy by targeting TNFSF14 and PTBP1, whereas Tang et al. [7] found that SHIP-1 can act as a target for miR-155 to regulate the responses of endothelial cells in pulmonary fibrosis. Moreover, Wei et al. [8] found that miR-133a mediates transforming growth factor-β1, which blocks myofibroblast differentiation and pulmonary fibrosis. Although numerous miRNAs and mRNAs have therefore been found to participate in pulmonary fibrosis, the vital roles of the miRNA-mRNA regulatory network in BIPF remain unclear, which may be addressed through sequencing, bioinformatics analysis, and functional verification.
In the current study, we aimed to elucidate the interactions between miRNAs and mRNAs in a mouse model of BIPF. Our ultimate objective was to provide a theoretical basis for the prevention and alleviation of BIPF.

Microarray Data.
We retrieved the database from the Gene Expression Omnibus (GEO; https://www.ncbi.nlm .nih.gov/gds), an open source for sequencing data [9,10]. The PRISMA flow diagram about BIPF data screening was shown in Figure 1. We then focused on the target miRNA expression profile GSE155823 ([GSE155823], Sudhiranjan Gupta, 2020, Studies on bleomycin-induced lung fibrosis in mice, GEO, GSE155823) based on the GPL28959 platform (LC Sciences Mouse miRNA Array [MRA-1002_MIR-MOUSE_22]) and mRNA expression profile GSE77326 ([GSE77326], Lv X and Liu S, 2017, Expression analysis of bleomycin-induced pulmonary fibrosis mouse model lung tissues, GEO, GSE77326) and matching the GPL15887 platform (Nimblegen Mouse Gene Expression Array [100718_ MM9_EXP]) in GEO [11]. Both datasets were sequenced from BIPF in mice or mice treated with saline solution (negative control group). In this study, the data were downloaded from the public database for analysis, and ethical approval was not required.

Predicted Target Genes of DEMs and Identified
Candidate Genes. miRWalk (http://mirwalk.umm.uniheidelberg.de) is a comprehensive miRNA target gene database that includes information on miRNA target genes in humans, mice, and other species [14] and integrates information from databases. We then entered the miRNA target genes and DEGs using Venn figure online tools (http:// bioinformatics.psb.ugent.be/webtools/Venn/) to search for overlaps between the groups.

Function and Pathways Enrichment Analyses.
After identifying the candidate genes, we analysed the function and enriched biological pathway by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG). We then evaluated GO terms of candidate genes in the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (v.6.8, https://davi.ncifcrf.gov) online database [15]; statistical significance was set at P < 0:05. We used the R ggplot2 package (https://mirror.lzu.edu.cn/CRAN/) for the visualization of GO annotations for biological processes (BPs), cell components (CCs), and molecular functions (MFs) and the online mapping software hiplot (https://hiplot.com.cn/basic) to visualise the KEGG enrichment results.

Protein-Protein Interaction (PPI) Networks and
Screening of Hub Genes. For the genes and proteins relationships, the STRING database (version: 11.0, https://string-db .org) was used [16] to obtain the PPI network, which calculated node degrees and provided different colours using the Cytoscape Network Analyzer (version 3.8.2). The MCODE and CytoHubba plug-ins were used to extract the top 10 hub genes to generate a subnetwork [17]. Then, DAVID and GO enrichment were used to evaluate the signalling pathways of the top 10 hub genes. Statistical significance was set at P < 0:05.
2.6. Construction of miRNA-mRNA Regulatory Network and Identification of Key miRNAs. Hub genes and corresponding miRNAs constructed the miRNA-mRNA interaction network, and the CytoHubba plug-in was then used to identify the top three key miRNAs. After which, the miRNAs were matched to the predicted target genes in miRWalk.

Validation of the Results by qPCR and Elisa Experiments.
To verify the results of three important hub genes and the key miRNAs, we used the following cell lines: RRID: TC-1 (ATCC Cat# CRL-2785, RRID: CVCL_4699) and MLE-12 (ATCC Cat# CRL-2110, RRID: CVCL_3751), treated with BLM (1 μg/ml) by conventional culture for 48 h to extract RNA and perform qPCR experiments (see Table 1).

Dual-Luciferase Reporter
Assays. Constructed dualluciferase reporter vectors and 293T cells (ATCC Cat# CRL-3216, RRID: CVCL_UE53) were cultured according to the above conditions. According to the instructions, the prepared DNA lip2000 complex and RNA lip2000 complex were added to the inoculated cells for cell transfection. Lysing cells for fluorescence detection. LARII (Firefly luciferase reaction working solution) and Stop & Glo® (Renilla luciferase reaction working solution) were added to detect the luciferase activity. The detection should be completed within 30 min as far as possible. BioMed Research International using the t test (unpaired t test, two-tailed). Dual-luciferase reporter assays were analyzed by a one-way ANOVA (homogeneity of variance test: P > 0:05). Meanwhile, the bioinformatics analysis was related to computer algorithm, which did not involve the author's statistical analysis.

Prediction of the Target Genes of DEMs.
We used the miRWalk database and predicted target genes of DEMs. The database identified 66 common DEMs (excluding PUC2MM in Table 2) and predicted 16,468 target genes. These target genes were then compared with DEGs, and

BioMed Research International
common genes in the two groups were identified. In total, 644 overlapping genes are identified as candidate genes, of which 431 genes are upregulated and 213 downregulated (see Figure 3).

Construction of the PPI Network and Modular Analysis
Screening of Hub Genes. We constructed PPI network which included 314 upregulated and 90 downregulated genes. Subsequently, the MCODE plug-in Cytoscape (v3.8.2) was used to identify and extract hub genes, among which 46 upregulated and 11 downregulated hub genes are identified (see Figure 5(a)). The CytoHubba plug-in was then used to filter the top 10 nodes of the hub genes such as MMP-9 and TIMP-1, which are the common genes involved in pulmonary fibrosis (see Figure 5(b)). DAVID online database analysis and GO evaluation of the top 10 hub genes were carried out, and we found that the top 10 hub genes are mainly concentrated on the tumor necrosis factor (TNF) signalling pathway (see Table 3).

3.7.
Verification of the Binding of serpina3n and Mmu-miR-1946a. The expression of Serpina3n and mmu-miR-1946a in MLE-12 and TC-1 was confirmed to be consistent with the bioinformatics analysis results. After that, we further discussed the binding of serpina3n and mmu-miR-1946a in tool cell 293. The dual-luciferase reporter assay results showed that both mmu-miR-1946a NC group and mimic had similar fluorescence expression of intensity in the mutant type serpina3n. On the contrary, the fluorescence expression of the mmu-miR-1946a mimic group was lower than that of the NC group in the wild-type serpina3n (see Figure 8). This result demonstrated that the wild-type serpina3n was bound with mmu-miR-1946a to degrade the fluorescein, which was confirmed the interaction between serpina3n and mmu-miR-1946a.

Discussion
In patients with malignancies treated with BLM, pulmonary fibrosis remains a major problem in clinical practice. Therefore, further work is needed to explore the molecular mechanisms and identify potential therapeutic targets of BIPF. Accordingly, in this study, we used bioinformatics analysis to provide new insights into key mediators and pathways related to the molecular mechanisms of BIPF. KEGG analyses showed enrichment pathways involving the top 10 hub genes and the top three key miRNAs were consistent with the pathophysiological processes of pulmonary fibrosis, including inflammatory and immune responses [18]. These findings suggested that the top 10 hub genes, top three key miRNAs, and their target genes obtained in this study may be involved in BIPF.  BioMed Research International miRNAs are endogenous small RNAs that regulate gene expression after transcription [19]; these molecules mainly act on the base pairing of the 3′-untranslated region (UTR) of mRNA to posttranscriptionally regulate the mRNA [20]. Specific miRNAs have been reported to be closely associated with pulmonary fibrosis [8]. We found that mmu-miR-1946a is one of the top three key miRNAs after analysing DEMs in BIPF and normal lung tissues. In a study by Fan et al. [21], mmu-miR-1946a is predicted to bind to the transforming growth factor (TGF-β1) 3 ′ UTR in mouse lung fibroblasts, thereby inhibiting its transcription. And they also found that alcohol can suppress mmu-miR-1946a and induce TGF-β1, which was reported to worsen fibroproliferation following BIPF in alcohol-fed mice. These findings are consistent with the results of our study, which demonstrated that mmu-miR-1946a may play vital regulatory roles in BIPF.
In this study, among the DEGs identified between the pulmonary fibrosis and control groups, three of the top 10 hub genes (serpina3a, serpina3n, and serpina10) and two 9 BioMed Research International target genes of the three key miRNAs genes (serpina3n and serpina10) were serine protease inhibitor (serpina) family genes. Serpina belongs to the serpin family, which is a widely distributed protease inhibitor. Among the serpina family, the serpina3 (alias: alpha-1-antichymotrypsin, α-ACT) subfamily is a typical acute phase protein that is significantly upregulated in response to inflammation in the circulatory system. Moreover, serpina3 participates in inflammatory responses and connective tissue renewal [22]. However, overexpression of serpina3 affects inflammation [23]. Previous studies have shown that serpina3 is associated with malignant melanoma and other diseases [24]. Our results show that serpina3n is a potential target gene of mmu-miR-1946a, and the expression of serpina3n is upregulated as mentioned above. Although regulation of mRNA translation and stability by miRNAs is achieved by degrading mRNA in most studies, various studies found that miRNAs can activate gene transcription abnormally as an enhancer trigger [25], suggesting that miR-NAs have a dual regulatory mechanism on mRNA. In this study, upregulated serpina3n is the target gene of upregulated mmu-miR-1946a, which may be a positive regulatory mechanism. The upregulated expression of mmu-miR-1946a may be achieved through the positive regulation of miRNA, but the specific regulatory mechanism remains to be confirmed. To the best of our knowledge, no reports have described to date the relationships between serpina3 and pulmonary fibrosis. The above results suggest that the ser-pina3 subfamily may play important roles in BIPF in mice.
In the top 10 hub genes, we found that the tissue inhibitor of metalloproteinase-1 (TIMP-1) expression was increased. TIMP-1 is a member of the metalloproteinases (TIMPs) family, with the main function of regulating the hydrolysis of cell matrix and surface proteins. TIMP-1 expression is also used as a biomarker to evaluate the prognosis of cancer patients [26] and specifically inhibit the activity of matrix metalloproteinases (MMPs). The expression of serpina3 and TIMP-1 is increased in the development of chronic obstructive pulmonary disease, which indicates that serpina3 and TIMP-1 may show the same expression trend in the disease [27], which is consistent with the relationships observed in our study. In addition, we used the gene expression profile interaction Atlas (GEPIA) database [28] for gene correlation analysis and found that the expression of serpina3 was correlated with the gene expression of TIMP-1 (see Figure 9(a)), which supported our results.
MMPs are an important family of proteases that degrade the extracellular matrix and its components. Among them, MMP-9, which exists in our top 10 hub genes, is the largest enzyme in the family [29]. MMP-9 has been found to be involved in the process of pulmonary fibrosis [30], and TIMP-1 is a specific inhibitor of MMP-9 [31]. Under normal physiological conditions, MMP-9 and TIMP-1 maintain a delicate balance in vivo, but the imbalance of the MMP-9/ TIMP-1 ratio is related to the progression of many diseases. Liu et al. [32] found that Icariside II attenuates cerebral ischemia/reperfusion-induced blood-brain barrier dysfunction in rats via regulating the balance of MMP-9/TIMP-1. In the clinical study of pulmonary fibrosis, an abnormal MMP-9/TIMP-1 ratio indicates a dysfunctional matrix degradation system, which leads to the progress of pulmonary fibrosis [33]. Wu et al. [34] found that IL-33 promoted the process of pulmonary fibrosis by inducing the imbalance between MMP-9 and TIMP-1. All the above studies showed that the imbalance of the MMP-9/TIMP-1 ratio can lead to the occurrence of many diseases, especially pulmonary fibrosis. Our study further demonstrates that the expression of MMP-9 decreases and the expression of TIMP-1 increases in mice three weeks after BIPF (see Figure 9(b)). Additionally, serpina3 is a specific activator of MMP-9 [35]. a3 expression increased, while MMP-9 expression decreased. In this study, the expression of serpina3 was increased, and then, the expression of MMP-9 decreased. The expression trends of these two genes are consistent with the biological effect of serpina3, which in turn supports the findings of this study. We also verified it through experiments in vitro. Therefore, the imbalance of the MMP-9/TIMP-1 is involved in the disease progression of BIPF.
From the internal threat to the validity of this study, we draw a conclusion through a large number of scientific and recognized approaches to screened and analyzed data. And then, we verified the results through in vitro experiments, 10 BioMed Research International so it contains a fairly high availability. However, the threat of our research is that we did not verify the results using functional experiments in vivo, and the effectiveness needs to be further performed in animal models. From the external threat to the validity, our validation experiment in vitro had been repeatedly verified, and the same trend and results had been obtained three times. So, it is also scalable for other similar or identical studies. Despite its limitation, this study is the first to use bioinformatics analysis to explore the hub genes and key miRNAs related to BIPF from the perspective of miRNA-mRNA regulatory network and the possible molecular mechanism to the best of our knowledge. Therefore, in consideration of our research findings being preliminary, we will continue to explore further mechanisms in the future.

Conclusions
In conclusion, our results indicate that the imbalance of MMP-9/TIMP-1 plays a critical role in BIPF in mice. Ser-pina3 and mmu-miR-1946a may be involved in the regulation of MMP-9 and TIMP-1. Our findings provide new insights into key mediators and pathways related to the molecular mechanisms in BIPF.

Data Availability
The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Ethical Approval
Ethical approval is not applicable.

Consent
Consent is not applicable.

Disclosure
Liuyin Chen, Zhiling Shi, and Lin Deng are co-first authors.