Comprehensive Analysis of the Profiles of Differentially Expressed mRNAs, lncRNAs, and circRNAs in Phosgene-Induced Acute Lung Injury

Phosgene exposure can cause acute lung injury (ALI), for which there is no currently available effective treatment. Mesenchymal stem cells (MSCs) which have been proven to have therapeutic potential and be helpful in the treatment of various diseases, but the mechanisms underlying the function of MSCs against phosgene-induced ALI are still poorly explored. In this study, we compared the expression profiles of mRNAs, lncRNAs, and circRNAs in the lung tissues from rats of three groups—air control (group A), phosgene-exposed (group B), and phosgene + MSCs (group C). The results showed that 389 mRNAs, 198 lncRNAs, and 56 circRNAs were differently expressed between groups A and B; 130 mRNAs, 107 lncRNAs, and 35 circRNAs between groups A and C; and 41 mRNAs, 88 lncRNAs, and 18 circRNAs between groups B and C. GO and KEGG analyses indicated that the differentially expressed RNAs were mainly involved in signal transduction, immune system processes, and cancers. In addition, we used a database to predict target microRNAs (miRNAs) interacting with circRNAs and the R network software package to construct a circRNA-targeted miRNA gene network map. Our study showed new insights into changes in the RNA expression in ALI, contributing to explore the mechanisms underlying the therapeutic potential of MSCs in phosgene-induced ALI.


Introduction
Phosgene is an indispensable mass production, used as intermediate in the manufacture of building blocks of various types of plastics, medicine, dye, and other chemical products [1]. It was reported that individuals accidentally exposed to phosgene at approximately >600 mg/m 3 ×min developed clinically significant phosgene-induced ALI [2]. Short-term exposure to phosgene leading to ALI and prolonged exposure would cause the fatal acute respiratory distress syndrome [3][4][5]. Now, exploring the potential molecular therapeutic targets for phosgene-induced ALI is needed.
Systemically, administered mesenchymal stem cells (MSCs) have the ability to selectively target sites of tissue injury or inflammation [6]. Exogenously administered MSCs have been observed to ameliorate lung injury in various animal models, including endotoxin-induced ALI [7], lipopolysaccharide-(LPS-) induced lung injury [8], and phosgene-induced ALI [9]. Several studies have demonstrated that facilitation of MSC localization to injured tissue sites can incrementally benefit ALI [10,11]. Thus, facilitation of MSC localization to target tissue sites represents a promising therapeutic strategy for ALI.
Most of the transcribed RNAs are identified as noncoding RNAs, which may be fully responsible for the complex gene expression in humans [12]. Recent years, mounting evidence has shown that lncRNAs and circRNAs play important roles in the regulation of the gene expression [13,14]. lncRNAs are a new class of regulatory RNAs over 200 nucleotides in length, and circRNAs are a class of ncRNAs that have stable structures and are resistant to the absence of 5 ′ or 3 ′ ends [15,16]. Emerging evidence suggests the involvement of lncRNAs and circRNAs in lung injury. For example, lncRNA TUG1 alleviates sepsis-induced ALI [17], whereas cir-cANKRD36 silencing can alleviate LPS-irritated human embryonic lung fibroblast cell injury [18]. To date, the profiles of ncRNAs, particularly circRNAs, and their roles in phosgene-exposed ALI have not been completely elucidated.
In our study, we conducted transcriptome sequencing to determine the expression profiles of mRNAs, lncRNAs, and circRNAs in a phosgene-exposed ALI model. Furthermore, we conducted GO and KEGG analyses and build the circRNA-miRNA coexpression networks. Our study is aimed at elucidating the molecular mechanisms of phosgeneinduced ALI after MSC treatment to identify biomarkers and new therapeutic targets for lung injury.

Experimental Animals and Sample
Collection. All experimental procedures involving animals were approved by the Animal Care and Use Committee of Jinshan Hospital affiliated to Fudan University, China. A rat model of phosgeneinduced ALI was constructed as described previously [5]. The rats were divided into three groups-air control (group A, n = 3), phosgene-exposed (group B, n = 3), and phosgene + MSCs (group C, n = 3). The rats in group A were exposed to normal room air, whereas the rats in the groups B and C were exposed to air comprising 8.33 mg/L phosgene for 5 min. Rats were intravenously injected with MSCs (10 6 cells per rat) via the tail vein. The rats' lung tissues were analyzed to determine the degree of MSC localization after 48 h.

Reference Genome Mapping and Transcriptome
Assembly. Raw reads generated during high-throughput sequencing were FASTQ format sequences. High quality reads obtained can be used for subsequent analysis, and these raw reads needed to be filtered further in terms of quality. Trimmomatic software was first used to remove adapters, after which low-quality bases, N-bases, and low-quality reads were filtered out. Finally, we obtained high-quality clean reads. The Q30 (Q score of 30) and GC contents of the clean data were then measured. The samples were assessed by genomic and gene alignment using HiSAT2 to align clean reads to the reference genome of the experimental species.
2.4. Identification of lncRNAs. Candidate lncRNA sets were subjected to the following rigorous screening steps for subsequent analysis. (1) The merged transcripts were compared with a known reference gene model using Cuffcompare software, and "I," "u," "x," and "o" transcripts were retained. (2) Transcripts of >200 bp and with ≥2 exons were selected. (3) The obtained transcripts were predicted using CPC2, CNCI, PLEK, and Pfam databases, and the transcripts from the intersection of these four databases were screened to obtain candidate lncRNAs. (4) The predicted lncRNA sequences were compared to known lncRNA sequences through BLAST software and were thus identified as known lncRNAs. For species without known lncRNAs, the predicted lncRNA sequences obtained were directly used for quantitative analysis.
2.5. Identification of circRNAs. CIRI software is highly sensitive and can be used to perform multiple screenings to reduce false positives; therefore, it is an authoritative software for circRNA prediction. In the present study, the CIRI software was based on the new alignment algorithm BWA-MEM comparison results, and the specific prediction process was as follows: (1) SAM files were obtained by the BWA-MEM comparison of clean reads with the genome of the reference species. (2) Balanced junction reads were detected based on xS/HyM (upstream) or xMyS/H (downstream) paired chiastic clipping signals.

GO Analysis.
In order to investigate the functions of the abnormally expressed circRNAs, lncRNAs, and mRNAs, GO annotation enrichment analyses were carried out. GO analysis classified differently expressed genes on the basis of three aspects (biological processes (BP), cellular components (CC), and molecular functions (MF)). Of the genes with aberrant mRNA targets between groups A and B, 1,558 were related to BP, 273 with CC, and 732 with MF ( Figure 4(a), Table S10). Of the genes with aberrant mRNA targets between groups B and C, 289 were related to BP, 51 with CC, and 68 with MF such as channel regulation ( Figure 5(a), Table S13). Of the dysregulated lncRNAs between groups A and B, 913 were related to BP, 182 with CC, and 250 with MF ( Figure 4(b), Table S11); of those between groups B and C, 435 were related to BP, 120 with CC, and 126 with MF ( Figure 5 Table S14). Of the dysregulated circRNAs between groups A and B, 296 were associated with BP, 87 with CC, and 101 with MF ( Figure 4(c), Table S12); of those between groups B and C, 113 were associated with BP, 49 with CC, and 57 with MF ( Figure 5(c), Table S15). GO analysis showed that aberrant lncRNA targets are mainly associated with regulation of the interleukin-4-mediated signaling pathway, symbiontcontaining vacuole membranes, and STAT family protein binding. Differentially expressed circRNA genes were found to mainly participate in immune system processes, lysosome activity, and enzyme inhibitor activity (Figures 4 and 5).

Discussion
In the present study, we used transcriptome sequencing to compare the expression profiles of mRNAs, lncRNAs, and circRNAs in the lung tissues of rats in the air control,     14 BioMed Research International phosgene-exposed, and phosgene + MSC groups. Furthermore, we conducted GO and KEGG analyses and constructed regulation networks. Our study is aimed at elucidating the molecular mechanisms of phosgene-induced ALI after MSC treatment to identify biomarkers and new therapeutic targets for lung injury. Previous studies on gene regulation have focused on protein-coding genes. However, in recent years, with the discovery of many ncRNAs, such as microRNAs, lncRNAs, and circRNAs, this view has changed [19]. The roles of lncRNAs in the progression and treatment of lung diseases have been reported. For example, MALAT1 was reportedly related to acute respiratory distress syndrome related to lung injury [20], downregulation of SNHG14 had protective effects against LPS-induced ALI [21], and CASC2 improved ALI by reducing lung epithelial cell apoptosis [22]. A potential relationship between lung injury and circRNAs has been demonstrated [18], revealing that circRNAs might have an important role in lung injury. In this study, we analyzed the abnormal expression profiles of lncRNAs, circRNAs, and mRNAs for the first time in phosgene-induced ALI after MSC treatment.
We determined the expression profiles of mRNAs, lncRNAs, and circRNAs using transcriptome sequencing. A total of 22,601 mRNAs, 10,187 lncRNAs, and 7,231 cir-cRNAs were identified, and a total of 109 circRNAs, 393 lncRNAs, and 560 mRNAs were observed to be differentially expressed in the three groups with a fold change of ≥2.0, P value of <0.05, and FDR of <0.05. The majority (56.26%) of the lncRNAs were antisense, and approximately 94.03% of the circRNAs were exonic. Compared with that observed in group A, 27 circRNAs, 133 lncRNAs, and 233 mRNAs were increased, and 29 circRNAs, 65 lncRNAs, and 156 mRNAs were decreased in group B. Compared with that observed in group B, 9 circRNAs, 55 lncRNAs, and 14 mRNAs were upregulated, and 9 circRNAs, 33 lncRNAs, and 27 mRNAs were reduced in group C. In our study, we found that fatty acid-binding protein 4 (Fabp4) mRNA and lncRNA plasmacytoma variant translocation 1 (PVT1) were downregulated in group C than that in group B. Previous studies have shown that Fabp4 and PVT1 were reported to be participated in the immune response. For example, Fabp4 inhibitors suppress inflammation and oxidative stress in murine and cell models of acute lung injury [23], and suppression of Fabp4 protects against rhabdomyolysis-induced acute kidney injury [24]. In chronic obstructive pulmonary disease patients, the PVT1 expression positively correlated with the GOLD stage and levels of TNF-α, IL-6, IL-8, and IL-17 [25]; PVT1 exacerbates the inflammation and cell-barrier injury during asthma by regulating miR-149 [26]. These findings suggested that Fabp4 and PVT1 might play the role in the therapeutic potential of MSCs in phosgene-induced ALI, and we will research the potential role of them in the phosgene-induced ALI in the following research. The differentially expressed    Figure 6: Continued. 16 BioMed Research International lncRNAs, mRNAs, and circRNA may play roles in phosgeneinduced ALI and may prove to be important in the treatment of ALI using MSCs. GO and KEGG analysis indicated that the main mechanisms of lung injury included single-organism processes, drug metabolism, and immune system processes. We also found that the lncRNA TCONS_00026162 (GO: 0060487) was associated with lung epithelial cell differentiation, TCONS_00026162 (GO: 0060441) was associated with epithelial tube branching involved in lung morphogenesis, TCONS_00026162 (GO: 0030324) was associated with lung development, and XR_001839103.1 and XR_593920.2 (GO: 0055114) were associated with the oxidation-reduction process. Meanwhile, circRNA_4627 (GO: 2000791 and GO: 0048286) was associated with the negative regulation of MSC proliferation involved in lung development and lung alveolus development, circRNA_5485 (GO: 0055114) was associated with the oxidation-reduction process, and circRNA_4178 (GO: 0002526) was associated with the acute inflammatory response. Lung epithelial cell differentiation has been reported to play key roles in various models of lung injury [27,28]. Melittin exerts beneficial function on paraquat-induced lung injury by regulating oxidative stress and apoptosis [29]. Cordycepin suppresses LPScaused ALI by preventing inflammation and oxidative stress [30]. Puerarin prevents LPS-induced ALI via inhibition of the inflammatory response [31]. These results suggest that the above mentioned lncRNAs and circRNAs may be involved in ALI.
Increasing evidence shows that natural endogenous cir-cRNAs are inherently resistant to exonucleolytic RNA decay, and that they contain selectively conserved miRNA target sites. Therefore, circRNAs can function as efficient miRNA sponges, interacting with miRNA to regulate the gene expression [14,32]. For example, the circRNA PVT1 facilitates osteosarcoma metastasis through regulation of the miR-526b/FOXC2 axis [33], and circRNA-33186 participated in the pathogenesis of osteoarthritis through functioning as a sponge of miR-127-5p [34]. In the present study, the potential target miRNAs were predicted, and the R network software package was used to establish a circRNA-targeted miRNA gene network map. circRNA-3871 contains binding sites for miR-339, miR-320-3p, miR-346, and miR-345-3p, and circRNA-2246 comprises binding sites for miR-149-5p, miR-296-3p, and miR-3593-5p. Compared with that observed in group A, the circRNA-3871 expression was upregulated in group B, and its target miRNA-miR-339-reportedly attenuates inflammation and inhibits pulmonary microvascular endothelial cell apoptosis in mice with severe acute pancreatitisassociated ALI [35]. Meanwhile, the circRNA-3235 expression       BioMed Research International was lower in group B than that in group C, and the expression of its target miR-320 was increased in ALI induced by cardiopulmonary bypass [36]. These findings suggest that circRNA-3871 functions as a miR-339 sponge and circRNA-3235 functions as a miR-320 sponge. Both these circRNAs may participate in phosgene-induced ALI progression and prove to be important in the treatment of ALI using MSCs. In our next study, we will verify this conjecture further.
In conclusion, we compared the expression profiles of mRNAs, lncRNAs, and circRNAs in the lung tissues of rats in three groups. Furthermore, we conducted GO and KEGG analyses and constructed coexpression networks. In addition, we used a database to predict target miRNAs interacting with circRNAs and the R network software package to establish a circRNA-targeted miRNA gene network map. Our study is aimed at elucidating the molecular mechanisms of phosgeneinduced ALI after MSC treatment to identify biomarkers and new therapeutic targets for lung injury.

Data Availability
We provide our data in the Supplementary Information files that we submit alongside our manuscript. The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
The authors declare no conflict of interest.

Acknowledgments
This work was funded by the National Natural Science Foundation of China (82002027) and the Science and Technology Committee of Jinshan District, Shanghai (2019-3-07).