Construction of potential periodontitis-related miRNA-mRNA regulatory network


 Background

MicroRNAs (miRNAs) are found to be involved in the pathogenesis of periodontitis, a major cause of tooth loss in adults. However, a comprehensive miRNA-mRNA regulatory network has still not been established.
Methods

One miRNA expression profile and 2 gene expression profiles were downloaded from the GEO database and analyzed using GEO2R. Candidate genes commonly appeared in differentially expressed mRNAs (DE-mRNAs) and target genes of differentially expressed miRNAs (DE-miRNAs) were selected for functional and pathway enrichment analyses using Enrichr database. Multivariate Logistic regression analysis was used to screen independent variables among candidate genes. The diagnostic values of screened genes were determined by the area under the receiver operating characteristic (ROC) curve (AUC).
Results

A total of 5 DE-miRNAs (4 upregulated and 1 downregulated) and 11 candidate genes (3 upregulated and 8 downregulated) were screened. After the construction of miRNA-mRNA regulatory network, 12 miRNA-mRNA pairs were identified. In the network, the upregulated genes were significantly enriched in cellular triglyceride homeostasis and positive regulation of B cell differentiation, whereas the downregulated genes were enriched in vesicle organization, negative regulation of lymphocyte and leukocyte migration. EPCAM and RAB30 were screened as risk factors of periodontitis. The combined AUC of these 2 genes was 0.896 (GSE10334) and 0.916 (GSE16134).
Conclusion

In this study, we established a potential periodontitis-related miRNA-mRNA regulatory network, which brings new insights into the molecular mechanisms and provides key clues in seeking novel therapeutic targets for periodontitis. In the future, more experiments need to be carried out to validate our current findings.


Results
A total of 5 DE-miRNAs (4 upregulated and 1 downregulated) and 11 candidate genes (3 upregulated and 8 downregulated) were screened. After the construction of miRNA-mRNA regulatory network, 12 miRNA-mRNA pairs were identi ed. In the network, the upregulated genes were signi cantly enriched in cellular triglyceride homeostasis and positive regulation of B cell differentiation, whereas the downregulated genes were enriched in vesicle organization, negative regulation of lymphocyte and leukocyte migration.
EPCAM and RAB30 were screened as risk factors of periodontitis. The combined AUC of these 2 genes was 0.896 (GSE10334) and 0.916 (GSE16134).

Conclusion
In this study, we established a potential periodontitis-related miRNA-mRNA regulatory network, which brings new insights into the molecular mechanisms and provides key clues in seeking novel therapeutic targets for periodontitis. In the future, more experiments need to be carried out to validate our current ndings.

Background
Periodontitis is a common chronic in ammatory condition that signi cantly affects the integrity of the tooth-supporting tissues, including gingiva, periodontal ligament and alveolar bone [1]. Periodontitis is one of the major causes of tooth loss in adults, and in its severe form it is the sixth most prevalent disease worldwide, affecting 734 million people [2]. It is thought to be caused by a dysbiosis of the commensal oral microbiota, which subsequently leads to dysregulated immune-in ammatory response and the damage of periodontal tissue [3]. Although treatment of periodontitis is successful in the majority of cases, up to 30% of patients with moderate periodontitis respond poorly to treatment [2]. Further understanding of the pathogenesis of periodontitis might assist in the treatment of periodontitis.
MicroRNAs (miRNAs) are a group of small endogenous noncoding RNA molecules that target messenger RNA (mRNA), causing mRNA degradation or the inhibition of protein translation [4]. MiRNAs participate in numerous important biological processes, such as cell proliferation, migration, differentiation, and apoptosis [5]. Because of their biological importance, the dysfunction of speci c miRNAs is signi cantly associated with periodontitis. For example, miRNA-146a are found to be a regulator of in ammatory responses and can contribute to the development of periodontitis [6]. MiRNA-125b regulates osteogenic differentiation of periodontal ligament cells through NKIRAS2/NF-κB pathway in periodontitis [7]. Together, growing evidence has revealed miRNAs as one of the key players in the onset and progression of periodontitis [8].
During the last decade, advances in microarray and high-throughput sequencing technology have made it possible to quantitatively analyze miRNA expression data sets with clinical pro les, and many miRNAs have been determined in periodontitis [9; 10; 11]. However, there remain questions about how the miRNAs and their target genes interact through molecular pathways in the pathogenesis of periodontitis. Construction of potential periodontitis-related miRNA-mRNA regulatory network will bring to light a relatively all-round molecular mechanism of miRNAs' impact in periodontitis. In the present study, we analyzed three sizeable and representative microarray pro les, including one miRNA expression pro le and two gene expression pro les, and nally established a potential periodontitis-related miRNA-mRNA regulatory network to uncover potential mechanisms of periodontitis occurrence and development.

Microarray data
One miRNA expression pro le (GSE54710) and 2 gene expression pro les (GSE10334 and GSE16134) with a sample size greater than 10 were downloaded from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). Dataset GSE54710 was Data processing GEO2R online tool (https://www.ncbi.nlm.nih.gov/geo/geo2r/) was applied to identify differentially expressed miRNAs (DE-miRNAs) and differentially expressed mRNAs (DE-mRNAs) between periodontitis and normal tissue samples. GEO2R is an interactive web tool based on the R software "LIMMA" package. We set |log 2 fold change (FC)| > 1 and P-value < 0.05 as the thresholds for identifying DE-miRNAs and DE-mRNAs.

Prediction of transcription factors (TF) of DE-miRNAs
The upstream TFs of DE-miRNAs were predicted using TransmiR database. TransmiR database (http://www.cuilab.cn/transmir) provides a valuable resource for the study of TF-miRNA regulation and can be used to analyze various processes, such as the evolution of the interactions, expression patterns and associated diseases of miRNAs [12].

Prediction of target genes of DE-miRNAs
The downstream target mRNAs of DE-miRNAs were predicted based on miRNet database (https://www.mirnet.ca/miRNet/home.xhtml), which is an easy-to use tool for comprehensive statistical analysis and functional interpretation of data from miRNAs studies.
Construction of miRNA-mRNA regulatory network Numerous evidences have supported an inverse relationship between miRNA and target gene. Thus the miRNA-mRNA regulatory network was constructed in following steps: (1) Periodontitis speci c DE-mRNAs were divided into upregulated and downregulated groups; (2) Candidate target genes were identi ed by conducting a combined analysis of upregulated DE-mRNAs and target genes of downregulated DE-miRNAs, or downregulated DE-mRNAs and target genes of upregulated DE-miRNAs; (3) The miRNAs that interacted with candidate target genes were selected for construction of the miRNA-mRNA regulatory network.

Functional and pathway enrichment analysis of candidate target genes
To determine the biological processes and pathways of candidate genes in the miRNA-mRNA regulatory networks, Enrichr database (http://amp.pharm.mssm.edu/Enrichr/) were used for Gene Ontology (GO) biological processes and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enrichment analysis. The GO analysis included three categories: biological process, molecular function and cellular component. P-value<0.05 was considered as statistically signi cant.
Evaluation of risk and diagnostic value of candidate genes Logistic regression analysis was used to calculate the odds ratio (OR) and 95% con dence interval (CI) for the association between candidate genes expression and the risk of periodontitis. Receiver operating characteristic (ROC) curve was performed, and the area under the curve (AUC) was measured to determine the diagnostic value of these candidate genes.

Prediction of downstream target genes of DE-miRNAs
The downstream target genes of candidate DE-miRNAs were predicted by using miRNet database. We nally predicted 1858 target genes for upregulated DE-miRNAs and 318 target genes for downregulated DE-miRNA. These predicted target genes were listed in Supplementary Table S3 Construction of miRNA-mRNA regulatory network According to the predicted miRNA-mRNA pairs, the candidate miRNA-mRNA regulatory network associated with development of periodontitis were nally constructed as presented in Figure 5. We found that WASL gene was potentially targeted by 2 miRNAs, including hsa-mir-451 and hsa-mir-223. Hsa-mir-671-5p, one of the most signi cantly upregulated miRNAs, was predicted to target 3 downregulated genes, including EPCAM, FOXP2 and MAMDC2 genes. Hsa-mir-223 was also upregulated and was found to target WASL, RIF1 and PLXDC2 genes. In addition, hsa-mir-1246, the main downregulated miRNA, potentially targeted XBP1, RAB30 and SAMSN1.
GO enrichment analysis of candidate genes GO biological process analysis revealed that candidate target genes of upregulated DE-miRNAs were signi cantly enriched in vesicle organization, negative regulation of lymphocyte migration and negative regulation of leukocyte migration ( Figure 6A). For molecular function analysis, these genes were signi cantly enriched in protein homodimerization activity, prostaglandin E receptor activity and NAD+ binding ( Figure 6C). The cellular component analysis for these genes included endocytic patch, actin cortical patch and cortical actin cytoskeleton ( Figure 6E).
GO biological process analysis demonstrated that candidate target genes of downregulated DE-miRNA were signi cantly enriched in cellular triglyceride homeostasis, positive regulation of B cell differentiation and regulation of MHC class II biosynthetic process Page 5/14 ( Figure 6B). For molecular function analysis, these genes were signi cantly enriched in enhancer binding, phosphotyrosine residue binding, protein phosphorylated amino acid binding, enhancer sequence-speci c DNA binding and core promoter binding ( Figure 6D).
At last cellular component analysis for these genes revealed that they were signi cantly enriched in Golgi stack, cis-Golgi network, Golgi cisterna and trans-Golgi network ( Figure 6F).

KEGG enrichment analysis of candidate genes
KEGG pathway enrichment analysis showed that candidate target genes of upregulated DE-miRNAs were signi cantly enriched in endocytosis, pathogenic escherichia coli infection, shigellosis, adherens junction and bacterial invasion of epithelial cells ( Figure 6G).
The enriched pathways for candidate target genes of downregulated DE-miRNAs were non-alcoholic fatty liver disease (NAFLD) and protein processing in endoplasmic reticulum ( Figure 6H).
The risk and diagnostic value of candidate genes To determinate the association between candidate gene expression and the risk of periodontitis, we conducted univariate Logistic regression analysis. As shown in Table 1, candidate target genes of upregulated DE-miRNAs were associated with decreased risk of periodontitis, whereas candidate target genes of downregulated DE-miRNA were associated with increased risk of periodontitis. Multivariate Logistic regression analysis was used to screen independent variables among candidate genes, and EPCAM and RAB30 were nally screened in both GSE10334 and GSE16134 datasets. The expression levels of EPCAM and RAB30 are shown in Figure 7A-D.

Discussion
MiRNAs are critical regulators of immune-in ammatory response and play important roles in the pathogenesis of periodontitis [8].
MiRNAs usually function by regulating target genes within the miRNA-mRNA regulatory network [13]. The dysregulation of miRNA-mRNA regulatory network has been reported to lead to a variety of human diseases, such as pancreatitis [14], chronic obstructive pulmonary disease [4] and cancer [15]. During the past few years, many studies have intensively demonstrated that alteration of miRNAs and their downstream target genes expression levels is closely associated with the development of periodontitis [8]. However, to our knowledge, up to now, a comprehensive miRNA-mRNA regulatory network in periodontitis has still not been created. In the present study, we conducted a potential periodontitis-related miRNA-mRNA regulatory network using miRNA and mRNA data from GEO database.
Based on this regulatory network, 4 upregulated DE-miRNAs and 1 downregulated DE-miRNA were nally identi ed. Although hsa-miR-3917 was initially identi ed as one of the most signi cantly upregulated miRNAs, it was excluded from the miRNA-mRNA regulatory network, because none of target genes of hsa-miR-3917 was screened as the candidate target genes in this regulatory network. In addition, most of DE-miRNAs that we screened was identical with previous studies. For example, hsa-miR-223 and hsa-miR-671-5p were found to be signi cantly upregulated in periodontitis tissue [9; 11]. Hsa-miR-1246 expression is lower in periodontitis than that in normal tissues [11]. Previous studies have showed that several DE-miRNAs, such as hsa-miR-223, could modulate the host immune response [8], and the expression of proin ammatory cytokines [16]. Thus, we speculated that these DE-miRNAs could play a potential regulatory role in the immune-in ammatory response in periodontal disease.
As reported in previous studies, the expression of miRNA can be modulated by TFs [17; 18]. We therefore predicted the TFs in relation to these DE-miRNAs. Pax4, a TF located on chromosome 7q32, was predicted as the TF that could potentially regulate expression of a majority of upregulated DE-miRNAs. Pax4 has been demonstrated to serve as a vital player in modulating miRNA expression and function. For instance, PAX4 was reported to promote migration and invasion in human epithelial cancers by decreasing miR-144 and miR-451 expression levels. NOTCH3, a signaling receptor involved in cell differentiation, was also predicted as one of the TFs in relation to upregulated DE-miRNAs. A recent investigation has demonstrated a critical role for NOTCH3 signaling in the modulation of tissue damage during in ammation [19]. In the future, the roles of these predicted TFs in periodontitis need to be further investigated. identi ed, including 3 up-regulated genes and 8 down-regulated genes. The majority of these genes were rstly identi ed to act as potential modulators in periodontitis. Among them, XBP1 has been reported to play a critical role in in ammatory diseases [20]. The role of XBP1 in mammalian host defenses and the innate immune response has also been uncovered by previous studies [21].
Subsequent functional enrichment analysis revealed that the downregulated genes were enriched in 56 biological processes, such as "negative regulation of lymphocyte migration" and "negative regulation of leukocyte migration", whereas the upregulated genes were enriched in 106 biological processes, such as "positive regulation of B cell differentiation" and "regulation of immunoglobulin secretion". Therefore, we speculate that these candidate genes may be of great importance in the immune-in ammatory response in periodontitis.
To evaluate the diagnostic value of candidate genes, multivariate Logistic regression analysis was performed on 11 DE-mRNAs, and EPCAM and RAB30 were nally screened. EPCAM is a type I transmembrane protein that regulates cell cycle progression and differentiation [22]. The extracellular domain of EPCAM can signi cantly enhance osteogenesis of mesenchymal stem cells under differentiation conditions [23]. RAB30 is a novel anti-bacterial autophagic regulator which plays an important role in the autophagy regulated immune response [24]. Using ROC analyses, we found that the aforementioned two genes could be considered biomarkers for the diagnosis of periodontitis. More interestingly, combining these two genes resulted in an extremely high diagnostic value, indicating that the two-gene diagnostic model had a good performance for the clinical detection of periodontitis.
Although a potential periodontitis-related miRNA-mRNA regulatory network has been constructed in the present study, some limitations should be recognized. First, the sample size is still small for miRNA expression pro le; Second, the direct relationship of the miRNA-mRNA pairs in the constructed network lacks of experimental validation. Third, miRNA and mRNA expression pro les are from the GEO database but different tissue origination, which may affect the accuracy of results. Finally, EPCAM and RAB30 were found to be gene biomarkers of periodontitis in this study, but the evaluation of periodontitis activity using gingival tissue specimens seems to be infeasible. In the future, the biomarkers associated with the two genes in the saliva and gingival crevicular uid need to be explored in periodontitis.

Conclusion
In summary, we identi ed a number of DE-miRNAs and DE-mRNAs between periodontitis tissues and normal tissues. Based on these DE-RNAs, we constructed a potential periodontitis-related miRNA-mRNA regulatory network and revealed important miRNA-mRNA regulatory axes, which may contribute to the nding of molecular mechanisms underlying the initiation and development of

Declarations
Ethics approval and consent to participation Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
The detailed data supporting the present study can be obtained upon reasonable request.

Competing interests
Gene symbol     The construction of miRNA-mRNA regulatory network in periodontitis.  The ROC curve of EPCAM and RAB30 in GSE16134.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryTableS1TheexpressionlevelsofDEmiRNAs.pdf SupplementaryTableS2.pdf SupplementaryTableS3.pdf SupplementaryTableS4.pdf