DNA methylation exploration for ARDS: a multi-omics and multi-microarray interrelated analysis

DNA methylation has been most studied and is classically associated with gene silencing via hypermethylation of CpG islands located in promoter regions of various diseases suppressor genes. In addition, DNA methylation is an attractive investigative tool for the study given that methylation is a reversible process. For instance, in myelodysplastic syndrome, demethylating agents have been tested in phase I clinical trials [14]. However, DNA methylation research on ARDS was inadequate and recent studies only focused on single genetic variations [15, 16]. Detailed studies of other novel candidates might lead to identification of unsuspected evolutionarily conserved mechanisms triggered by ARDS.

To our knowledge, this was the first study of all-encompassing analysis for DNA methylation alterations in ARDS, indicated in Additional file 9: Table S3. As furthermore research targets for improvement of therapies in ARDS, the present study uncovered 30 methylation alterations and their regulating DNA in ARDS regions. Meanwhile, as diagnostic molecules for ARDS, the clinicians might be interested in the top 10 difference methylation sites and 4 high diagnostic efficacy CpG islands variations.

Due to microarray experiments from human blood samples as the object of this study, 7 animal model studies and 10 studies not on blood samples were eliminated, leaving only 4 studies of microarray experiments and 1 study of transcriptome sequencing. In the consideration of comorbidity of hematologic malignancy existing in patients of transcriptome sequencing study, the sequencing data were eventually removed. Previous studies utilized only one single database, with a relatively small sample size. Transcriptome microarray datasets were not found healthy participants, which means in the need of searching healthy volunteers data. The data of healthy participants (generalized anxiety disorder score = 0) from one dataset related to anxiety disorder was extracted, according to matched terms of health and GPL10558 (the same platform number as included gene expression microarray of ARDS) in consideration of compatibility with microarray studies.

After merging these cohorts into a single cohort by Perl, batch correction and co-normalization were performed through ComBat normalization in the sva R package, which was a frequently-used method to decease heterogeneity among microarray studies [17‐20]. In addition, the threshold values of screening differential mRNAs were P value < 0.05 in previous studies, which might lead to partially false positives. The significance criteria of adjusted P value < 0.05 aimed for reduction of false positives [21, 22]. Strikingly, it has been reported that the most epigenetic mutations appear unrelated to biological function [23].

Since DNA methylation could directly block transcription by inhibiting the binding of specific transcription factors to their target sequences on the candidate gene in the result of an obvious variation in transcriptome, integration with methylation data and matched transcriptome data was conducted to screen meaningful methylation mutations and methylated DNA regions [24].

Until now, the most significant 10 methylation variations found through difference analyses were lack of studies, with the function of their methylated gene related to immunity and cell proliferation, which could be appealing to researchers for further studies [25‐29].

Tejera et al. [30] and Wang et al. [31] had confirmed that the mRNA level of PI3 in patients with ARDS was significantly decreased, which in turn might increase the activity of neutrophil elastase indirectly for PI3 plays a central role in controlling the excessive activity of neutrophil elastase. Liu and his colleagues [32] had showed that CX3CR1 was the crucial molecule that regulates the EGFR, Src and FAK pathways, which are crucial for epithelial and endothelial cell growth. The reduction of CX3CR1 in patients with ARDS could lead to damage to epithelial and endothelial cell regeneration. However, the mechanism of down-regulated PI3 and CX3CR1 in ARDS is unclear. Our study showed that the hypermethylation on cg24476939 of PI3 and 4 hypermethylation sites of CX3CR1 might explain down-regulation of PI3 and CX3CR1 in ARDS. Similar to PI3 and CX3CR1, FYN, DUSP6, POLB and SLC3A2, PILRA, SRPK2 and RNF1 were confirmed to be related to the pathogenesis of ARDS, including inflammation or immunity, endothelial–epithelial barrier, and coagulation function [33‐36]. The present study verified above variations on mRNA levels in ARDS. Moreover, methylation alterations in these DNA regions of molecules might be responsible for the mRNA level variation in ARDS, which might be potential therapeutic targets for ARDS.

In addition, the ROC analyses on screened methylation alterations indicated that the AUC on cg03341377, cg24310395, cg07830557 and cg08418670 was up to 0.99, which meant that these DNA methylation alterations were valuable targets to distinguish ARDS from healthy controls, Table 2.

There are several limitations to the present study. First, as a retrospective study of primarily publically available data, we are not able to control for demographics, infection, patient severity, or individual treatment, and there was no analysis combined with clinical data because the microarray data did not provide clinical information. Second, although 4 microarray datasets were merged, a larger sample size may have been better. The ideal method of transcriptome and genome interrelated analysis should measure DNA methylation and mRNA in the same samples and then perform a co-expression analysis. However, in the public database, there was no such completed microarray data regarding ARDS. Besides, for the limitation of bioinformatics, the further studies and experiments should be conducted to sequentially verify and research these methylation alterations screened by our analyses.