Introduction
Periodontitis is a chronic infectious inflammatory disease driven by reciprocally reinforced interactions between the dysbiotic microbiome and dysregulated immunity [
1,
2]. Many studies have found that expression of specific genes play a role in the pathogenesis of periodontitis and its susceptibility [
3‐
5]. Therefore, comprehensive omics researches could provide us more information about the formation and progress of the diseases.
Noncoding RNAs (ncRNAs), including long noncoding RNAs (lncRNAs), microRNAs (miRNAs), and circular RNAs (circRNAs), participate in multiple biological processes and in the pathogenesis of diseases, such as tumors, Alzheimer's disease and other inflammatory diseases [
6]. The competitive endogenous RNA (ceRNA) hypothesis describes an intricate interplay among diverse RNA species. miRNAs complementarily pair with mRNAs through miRNA response elements (MREs), decreasing mRNA expression levels. lncRNAs or circRNAs also have MREs, to regulate mRNA expression levels through competitive binding with miRNA [
7]. Several studies have constructed lncRNA-related ceRNA networks in periodontitis [
8]. It has been reported that lncRNA MIAT and the MIAT-based ceRNA network may regulate the immune response during the progression of periodontitis [
9]. Besides, lncRNA FGD5-AS1 were reported to be involved in the pathogenesis of periodontitis through ceRNA [
10].
The complex coordination of inflammation and the immune response in the periodontal tissue is important in periodontitis. However, there are only a few studies on ceRNA related to immune regulation in periodontitis, and most studies only focus on several specific lncRNAs. Comprehensive analysis of immune-related ncRNAs and ceRNA network including lncRNAs and circRNAs has not been reported. Therefore, the objective of our study was to comprehensively profile the differential expression of mRNAs, lncRNAs, circRNAs and miRNAs in the inflamed periodontal tissue (IPT) and healthy gingiva through whole-transcriptome sequencing, and construct an immune-related ceRNA network in periodontitis, to identify key regulators which may serve as potential therapeutic targets.
Material and methods
All methods were performed in accordance with the relevant guidelines and regulations.
Sample collection
Tissue samples were collected during periodontal surgery at the Peking University Hospital of Stomatology from May 2020 to April 2021. All the participants gave written informed consent.
The inclusion criteria were as follows:
-
Patients aged 18–65 years, systemically healthy, who agreed to participate in the trial and had finished non-surgical therapy.
-
The IPT was collected from the operative site with pocket depth (PD) ≥ 6 mm and bleed index (BI) > 2 in patients diagnosed stage III and grade C periodontitis.
-
Healthy gingival tissue was collected from patients who underwent crown lengthening, PD ≤ 3 mm and BI ≤ 2, and showed no alveolar bone loss on radiography.
The exclusion criteria were as follows:
-
Patients with acute periodontal disease.
-
Patients received antibiotics or underwent periodontal surgery in the past 3 months.
-
Smokers, and pregnant or lactating women.
Clinical measurements of PD, BI, clinical attachment loss (CAL), gingival recession (REC) were recorded. The tissue samples collected were rinsed with 0.9% normal saline, immediately frozen in liquid nitrogen, and stored at − 80 °C. In total, 20 periodontitis and 20 healthy samples were collected, five from each group were used for sequencing and 15 were used for validation.
High-throughput sequencing and data processing
Total RNA was extracted from five IPT and five healthy gingivae, and sequenced using the BGISEQ-500 platform (Huada Gene Technology). Sequencing data were filtered using SOAP nuke, and clean reads of lncRNAs and mRNAs were mapped to reference the genome using HISAT2. Stringtie was used to assemble and quantify transcripts. Small RNA clean reads were aligned to reference the genome (Hg19) using Bowtie 2 and quantified with FeatureCounts. Since single prediction software programs often have certain limitations, we used two programs, i.e., CIRCexplorer2 and find_circ, to identify circRNAs, and their intersection was retained for further analysis. All expression profiles were used for principal component analysis.
Differential gene analysis and functional enrichment analysis
DESeq2 was used for differential expression analysis, mRNAs, genes with
P ≤ 0.05, FDR ≤ 0.05, and |log2FC (fold change)|≥ 1 were identified as differentially expressed genes (DEGs). GO and KEGG enrichment analyses were performed with clusterProfiler. Then, we used the MSigDB C5 GO and MSigDB C2 KEGG gene sets for gene set enrichment analysis (GSEA) [
11]. miR-Path was used to perform enrichment analysis of differentially expressed miRNAs.
Construction of the protein–protein interaction network
We used the STRING database to identify the protein–protein interaction (PPI) between differentially expressed mRNAs, with confidence > 0.7. The PPI network was then imported into Cytoscape for topology property analysis. Cytohubba were used to identify hub genes.
Immune cell infiltrations analysis
All expression profiles were imported to CIBERSORTx for immune cell infiltrations analysis. CIBERSORTx is widely used in immune cell infiltration analysis for bulk RNA-sequence data. The correlation coefficient between differentially expressed lncRNA/circRNA and the expression of characteristic genes in immune-infiltrated cells was calculated, and ncRNA/circRNA with correlation > 0.8 were selected as immune-related.
ceRNA network construction
miRNA–mRNA interactions were predicted using miRWalk 2.0, which uses 12 predicted algorithms, and target genes predicated by six algorithms were maintained for further analysis. The lncRNA‐miRNA interactions were extracted from miRcode and starBase, and the circRNA-miRNA interactions were retrieved from the circBank database. Expression of lncRNAs, mRNAs, and circRNAs were used for co-expression analysis. Pearson’s correlation coefficients ≥ 0.95 were considered to indicate co-expression. The ceRNA network was constructed by integrating all validated interactions and co-expression pairs. On this basis, we constructed an immune-related ceRNA network combined with the immune-related lncRNAs/circRNAs, and immune-related mRNAs from ImmPort database.
Validation of expression patterns
Total RNA from 15 inflamed periodontal tissue and 15 healthy gingivae was isolated using TRIzol® Reagent. The ABScript II cDNA Fist-Strand Synthesis Kit (ABclonal Technology Co., Ltd, Wuhan, China) was used for reverse transcription to synthesize cDNA. RT–qPCR was conducted using qPCR SYBR® Green Master Mix (Roche Holding AG, Basel, Switzerland) on an ABI Q3 system (Applied Biosystems, Foster City, CA, USA). GAPDH was used as endogenous reference. All primer sequences were synthesized by Sangon Biotech. mRNA expression was determined using the ΔΔCT method. Difference between two groups were evaluated with t tests. Statistical analyses were performed with SPSS 22.0 (IBM, Armonk, NY, USA).
Discussion
To our knowledge, this was the first study to perform whole-transcriptome sequencing on IPT and healthy tissue, and described the expression profile of mRNAs, lncRNAs, miRNAs and circRNAs at the same time. Zou et al. [
12] and Li et al. [
13] only detected lncRNAs expressions in periodontal tissues of patients with chronic periodontitis and aggressive periodontitis [
12,
13]. Yu et al. [
14] and Stoecklin-Wasmer et al. [
15] detected the expression of circRNAs and miRNAs, and mRNAs and miRNAs repectively. Besides, all samples collected in our study were from deep intrabony defect of patients with periodontitis (Stage III Grade C), which could reflect the typical transcriptome in IPT.
Based on the whole-transcriptome analysis, we comprehensively constructed the immune-related ceRNA network in periodontitis. A previous study analyzed the transcriptome of gingiva from peri-implantitis, periodontitis, and healthy patients, and compared the difference of ceRNA network between periodontitis and peri-implantitis [
16]. Lin et al. also analysis the ceRNA network in periodontitis using dataset in GEO database [
8]. These studies only described the lncRNA-related ceRNA network, but circRNAs were not included.
Function enrichment analysis showed cell chemotaxis was significantly enriched. Besides, hub genes in the PPI network were mainly of chemokines and chemokine receptors. Another interesting finding of GO analysis was that tooth mineralization, and biomineralization were significantly upregulated in IPT. It has been reported that there are kinds of stem cells in IPT, which may be related to the function of these stem cells.
In our study, whole-transcriptome expression data and CIBERSORTx were used for Immune cells infiltration analysis, revealing the proportion of immune cells in periodontal tissue. It has showed that the proportion of neutrophils was not high and the highest proportion immune cell are lymphocytes. The results are consistent with the immune cell infiltration analysis by Li et al. [
17]. Single-cell sequencing could directly obtain the proportion of different kinds of cells in the tissue, which is more accurate than immune cell infiltration analysis. Recently single-cell sequencing analysis of periodontal tissue also showed that neutrophils accounted for only 0.2% in healthy, and 0.8% in periodontitis samples, far lower than plasma cells (about 10%-30%) and T lymphocytes (about 20%-25%) [
18]. Previous studies have confirmed that the main types of immune cells in advanced periodontitis were B cells and plasma cells, but the role of naive B cells in periodontitis has received little attention. We found the proportion of naive B cells was higher in IPT, although previous flow cytometry analysis showed few naive B cells (< 8%) in periodontal tissues [
19]. However, a previous study found no significant difference in the number of naive B cells between chronic periodontitis, aggressive periodontitis and healthy group by flow cytometry [
20]. In fact, the role of naive B cell in periodontal tissue is still unclear and more research is needed.
JUN, FOS, KLF2, THBS1 and
WIF1 were identified as key regulator in periodontitis and validated to be highly expressed in IPT through RT-qPCR. We identified these 5 genes as key regulators according to function enrichment analysis and PPI network. We identified 10 hub genes in the PPI network, however, most of them are cytokines and chemokines, apart from
JUN and FOS, which have been studied a lot. JUN and FOS are subunits of transcription factors AP-1. Activation of toll-like receptors (TLRs) stimulates the production of multiple cytokines, and eventually activates AP-1 [
21]. M1 macrophages regulate TLR4/AP1 and promote alveolar bone destruction in periodontitis [
22]. A deep learning-based autoencoder predicted
FOS and
JUN to be critical immunosuppression genes and mediate immune suppression in periodontitis [
23].
WIF1, 29-fold up-regulated in IPT, could directly interacts with Wnt ligands and may be key to the inhibition of Wnt signaling in IPT.
KLF2 plays a key role in the activation of immune cells and participates in inflammatory diseases by regulating the NF-κB pathway. Besides,
KLF2 regulates osteoclast generation and inhibits PDSLCs osteogenic differentiation, thus participating in bone destruction [
24].
THBS1 is an important mediator involved in the chemotactic function of neutrophils and monocytes [
25].
We identified 20 immune-related lncRNAs, and
EGOT and
PRKCQ-AS1 were the top two.
EGOT modulates the PI3K/AKT, MAPK, and NF-κB pathways to activate inflammation [
26]. After LPS and TNF-α stimulation, the expression of EGOT in THP-1 and CD4, CD8 + T cells increased.
PRKCQ-AS1, has not been studied in periodontitis, and our analysis showed that PRKCQ-AS1 may act as sponge of
miR-141, miR-6512, miR-513c, and regulate the expression of
CXCL1, PTGS2 (COX-2),
THBS1 and
PRKCQ.
As the central molecules of the ceRNA network, miRNAs participate in maintaining periodontal homeostasis.
miR-146a and
miR-17 regulate the osteogenic differentiation of PDSLCs in an inflammatory environment and miR-34a, miR-146a, and miR-223 inhibit osteoclast differentiation [
27]. Topological analysis of the ceRNA network showed that
miR-141-3p, miR-1304-3p, miR-1293, and
miR-33a-5p were hub miRNAs. Among them, the plasma derived
miR-1304-3p exosome was down-regulated in periodontitis, and returned to normal after periodontal treatment, suggesting
miR-1304-3p may be involved in the regulation of periodontal inflammation [
28].
miR-1293 directly bind to
IL-6 mRNA and inhibit
IL-6 expression [
29].
miR-33a-5p can target the NF-κB pathway and Wnt/β-catenin pathway to regulate immune responses and affect cell proliferation, migration, and other biological processes [
30].
circRNAs may play a key role in periodontal homeostasis. However, the related research in its infancy. Most studies on the role of circRNAs in periodontitis have mainly focused on osteogenic differentiation, or cell proliferation of PDLCs. In our study, 518 circRNAs were differentially expressed, and 146 were immune-related. Next, we will conduct further validation to clarify how these circRNAs regulate the periodontal inflammatory process.
Our study identified the above key regulators in periodontitis through bioinformatics analysis. Further experimental studies are needed to verify the exact role of the identified ceRNA network in IPT. In addition, the sample size included in our study was only 5, in the future, with the development of technology and reduction of cost, we would further expand the sample size for more accurate ncRNA expression data in IPTs.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.