Classifying Integrated Signature Molecules in Macrophages of Rheumatoid Arthritis, Osteoarthritis, and Periodontal Disease: An Omics-Based Study

Rheumatoid arthritis (RA), osteoarthritis (OA), and periodontal disease (PD) are chronic inflammatory diseases that are globally prevalent, and pose a public health concern. The search for a potential mechanism linking PD to RA and OA continues, as it could play a significant role in disease prevention and treatment. Recent studies have linked RA, OA, and PD to Porphyromonas gingivalis (PG), a periodontal bacterium, through a similar dysregulation in an inflammatory mechanism. This study aimed to identify potential gene signatures that could assist in early diagnosis as well as gain insight into the molecular mechanisms of these diseases. The expression data sets with the series IDs GSE97779, GSE123492, and GSE24897 for macrophages of RA, OA synovium, and PG stimulated macrophages (PG-SM), respectively, were retrieved and screened for differentially expressed genes (DEGs). The 72 common DEGs among RA, OA, and PG-SM were further subjected to gene–gene correlation analysis. A GeneMANIA interaction network of the 47 highly correlated DEGs comprises 53 nodes and 271 edges. Network centrality analysis identified 15 hub genes, 6 of which are DEGs (API5, ATE1, CCNG1, EHD1, RIN2, and STK39). Additionally, two significantly up-regulated non-hub genes (IER3 and RGS16) showed interactions with hub genes. Functional enrichment analysis of the genes showed that “apoptotic regulation” and “inflammasomes” were among the major pathways. These eight genes can serve as important signatures/targets, and provide new insights into the molecular mechanism of PG-induced RA, OA, and PD.


Introduction
Arthritis is a prevalent disease. It is often defined as joint swelling, pain, and stiffness. It is a collective term that encompasses a variety of joint inflammation and bone disease conditions. The most frequent kinds of arthritis are rheumatoid arthritis (RA) and osteoarthritis (OA). RA is an autoimmune and inflammatory disease in which the body's immune system attacks its tissue, including the linings of the joints and internal organs (in extreme cases), resulting in painful swelling and ultimately leading to bone deterioration and disaggregation over time [1,2]. In contrast, OA, the most prevalent kind of arthritis, causes pain, stiffness, joint degeneration, and osteophytes. The majority of research connects OA to low-grade inflammation and mechanical stress, including accidents, aging, obesity, etc.
In the present study, we retrieved the publicly available gene expression profiles from the Gene Expression Omnibus (GEO) database containing samples for macrophages in rheumatoid arthritis (RA) synovium (GSE97779), osteoarthritis (OA) synovium (GSE123492), and Porphyromonas gingivalis stimulated macrophages (PG-SM) (GSE24897). All three data sets were screened for differentially expressed genes (DEGs) using R. The common DEGs among RA, OA, and PG-SM were selected and subjected to gene-gene correlation analysis. Additionally, an interaction network for the highly correlated DEGs was built using GeneMANIA, and the cytoHubba plugin for Cystoscope was used to calculate network centrality metrics, in order to identify important nodes in the network. Finally, functional enrichment analysis was performed for highly interacting (hub) genes using Enrichr and DAVID. The identification of hub genes shared among RA, OA, and PD, in addition to the analysis of their biological processes and pathways, may shed light on the molecular mechanism of the pathogenesis of these diseases.

Materials and Methods
In the present study, we analyzed the gene expression profiles of macrophages in RA synovium, OA synovium, and PG-SM, in order to identify common significant biomarkers/targets associated with the three conditions. Figure 1 depicts the workflow of the study.

Retrieval of Gene Expression Data
The gene expression data were retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) (date accessed: 6 Sept 2021) [34] using the following search terms: Porphyromonas gingivalis, rheumatoid arthritis, and osteoarthritis, as well as the following filters: organism (Homo sapiens) and tissue type (macrophages). Two data sets, GSE97779 and GSE24897, for macrophages in RA and PG-SM, respectively, were selected. A third RNAseq data set for OA, series ID GSE123492, was also se-

Retrieval of Gene Expression Data
The gene expression data were retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) (date accessed: 6 September 2021) [34] using the following search terms: Porphyromonas gingivalis, rheumatoid arthritis, and osteoarthritis, as well as the following filters: organism (Homo sapiens) and tissue type (macrophages). Two data sets, GSE97779 and GSE24897, for macrophages in RA and PG-SM, respectively, were selected. A third RNAseq data set for OA, series ID GSE123492, was also selected to examine the DEGs at the transcriptional level. The details of the data sets are listed in Table 1. The data sets GSE97779 and GSE24897 were generated using the Affymetrix Human Genome U133 Plus 2.0 Array platform. GSE97779 contains samples from RA synovial macrophages (nine) and control macrophages (five) [39], whereas GSE24897 contains twelve samples from four conditions: macrophages treated with saline (three), PG (three), PG-LPS (lipopolysaccharide) (three), and PG-fimbriae (three) [6]. This data set was incorporated in the study to examine the effect of PG virulence factors on adherent macrophages, in order to gain a better understanding of the intricate relationships between innate immune response in human macrophages and chronic PD. Of the total twelve samples, six sample macrophages treated with PG and saline were selected for the present analysis. The data set GSE123492 was constructed with the Illumina NextSeq 500 platform, and contains samples from synovial tissue macrophages in OA (nine), RA (two), and PD (one) [20]. OA data were subgrouped as classic OA (cOA; five) and inflammatory OA (iOA; four). For our analysis, only nine OA samples were used.

Identification of Differentially Expressed Genes (DEGs)
The GEOquery package of R software [40] was used to read the Series Matrix File(s) of the data sets GSE97779 and GSE24897, and a log2 transformation was applied to make them more symmetrical so that the parametric statistical test would produce more accurate and meaningful differentially expressed genes (DEGs). Two categories of probes have been excluded to eliminate false-positive statistical findings: (1) Affymetrix controls with the symbol "AFFX" and (2) probe sets that are likely to hybridize with several genes (Affymetrix designates these probe sets with the abbreviation x). DEGs were screened using the Linear Models for Microarray Analysis (Limma; www.bioconductor.org/packages/release/bioc/ html/limma.html) package in R (date accessed: 6 September 2021) [41,42]. A t-test in the Limma package was performed to determine the p-value of each gene symbol. The following criteria were used to identify significant DEGs: p-value ≤ 0.05 and log2-fold change (logFC) ≥ ±1 [8]. Gene annotations were retrieved from the GEO database and prob ids were transformed into gene symbols; genes with the lowest p-values were chosen in the case of duplicate genes. The Raw Count Matrix Files for GSE123492 were also read using the GEOquery package and were normalized using the EdgeR tool of R [37]. After normalizing the data, DEGs were identified using the Fisher's Exact Test between cOA and iOA. The following criteria were used to identify significant DEGs: p-value ≤ 0.05 and log2-fold change (logFC) ≥ ±1. Finally, a gene set of common DEGs were identified in RA, OA, and PG-SM.

Gene-Gene Correlation Network Construction
A correlation network was built for the common DEGs based on their expressions. Expression data for 72 common DEGs for each RA, OA, and PG-SM were extracted. Next, individual gene interactions were determined using Pearson correlation coefficients between pairs of 72 common DEGs independently for RA, OA, and PG-SM. Three co-expressionbased matrices between each of the 72 common DEGs were created using the Cladist tool [31,42]. Cladist takes expression values as input and computes an N × N co-expression matrix between each gene pair by employing the Pearson correlation coefficient (r) score given by the following: where x and y represent the sample means of the expression values of the two genes in the control and diseased state, respectively. The r score ranges from −1 to +1; the higher the r score, the greater the expression similarity between two genes. Extensive gene-gene correlation analysis produced a significantly large amount of data; therefore, non-repeating highly correlated gene-gene pairs with an r score > 0.7 were extracted individually for all three disease conditions [43].

Network Construction
The Cytoscape [44] plugin GeneMANIA [45] v3.5.2 (https://apps.cytoscape.org/ apps/genemania) (date accessed: 6 September 2021) was used to construct an interaction network of the DEGs. Cytoscape is an open-source platform-independent network visualization software. It offers several plugins/apps for various network analyses. GeneMANIA allows users to construct a weighted composite gene-gene functional interaction network from a gene list. Functional interactions between the 47 highly correlated common DEGs were predicted by GeneMANIA. In addition to the DEGs, 20 additional genes were used to create the interaction network using Homo sapiens as a source species. The functional associations in the network were evaluated using the following terms: co-expression, colocalization, genetic interactions, pathways, physical interactions, predicted interactions, and shared protein domains.

Hub Genes Identification
Networks can be used to display a wide range of biological data, including proteinprotein interactions, gene regulation, cellular pathways, and signal transductions [10,46]. An interaction network is represented as a graph G = (V, E), where V and E are the sets of vertices (nodes/genes/proteins) and edges (links/functional associations/interactions), respectively [47]. Most biological networks have a scale-free topology and therefore are more robust than random networks. Scale-free networks have a power-law degree distribution, with a small number of highly connected nodes ("hubs") and a large number of poorly connected nodes ("non-hubs"). Hubs play a significant role in the functional and modular architecture of interactomes. As a result, they are assumed to be more vital to life than non-hub nodes, according to the centrality-lethality rule [48] The Cytoscape plugin cytoHubba [49] (https://apps.cytoscape.org/apps/cytohubba) (date accessed: 6 September 2021) was used to calculate the topological parameters of the network. CytoHubba offers 11 topological analysis methods, including six centrality measures. We selected degree centrality (DC), betweenness centrality (BC), bottleneck (BN), and closeness centrality (CC), in order to identify key/important nodes in the network. DC of a node u ∈ V(G) is defined as the number of its first neighbors. Nodes with high degrees are referred to as "hubs" [47]. BC is a measure of the number of non-redundant shortest paths that pass through a given node. Nodes with high BC are defined as "bottlenecks", as these nodes act as bridges/connecting links between dense clusters; they control the information flow among other nodes in the network. The BC of a node u ∈ V(G) is computed as follows: where σ st is the total number of shortest paths from node 's' to node 't', and σ st (u) is the number of those paths passing through u [47]. Both hubs and bottlenecks tend to be essential in protein interaction networks [48,50]. CC is defined as a measure of how fast the flow of information would be from a given node to all other nodes in a network, sequentially. Nodes with high CC are the closest to all other nodes in a network. The CC of a node u ∈ V(G) is computed as follows: where d uv is the distance (length of the shortest path) between nodes u and v, and N is the number of nodes in G [47].

Gene Enrichment Analysis
In order to analyze the role of hub genes in RA, OA, and PG-SM, a functional enrichment analysis was performed [37] using two web servers, namely Enrichr (https: //maayanlab.cloud/Enrichr/) (date accessed: 6 September 2021) [51] and DAVID v6.8 (Database for Annotation, Visualization, and Integrated Discovery database) (david.ncifcrf. gov/) (date accessed: 6 September 2021) [52]. Enrichr was first used to extract gene ontology (GO) annotations such as biological process (BP), molecular function (MF), and cellular component (CC). Enrichr was then used to perform KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway analysis, which was further confirmed using DAVID. Finally, an Enrichr feature called "Jensen DISEASES" was used to integrate evidence on disease-gene associations [53].

Identification of DEGs
The data sets with the series ID GSE97779, GSE123492, and GSE24897 were subjected to DEG analysis using R. A total of 10890, 1276, and 2340 DEGs were identified in RA, OA, and PG-SM, respectively, based on the following statistical parameters: p-value ≤ 0.05 and log2-fold change (logFC) ≥ ±1. An MA-plot (M, log-ratio; A, mean average) for each data set is shown in Figure 2. A total of 72 common DEGs were identified among RA, OA, and PG ( Figure 3).  based on the following statistical parameters: p-value ≤ 0.05 and log2-fold change (logFC) ≥ ±1. The red and blue dots indicate genes that have been considerably up-and down-regulated, respectively, whereas the black dots indicate genes with no significant difference between the diseased and control groups.

Gene-Gene Correlation Network Construction
A total of 72 genes common in RA, OA, and PG-SM were selected for the gene-gene correlation network (Supplementary File: Table S1). The expression values of each sam-

Gene-Gene Correlation Network Construction
A total of 72 genes common in RA, OA, and PG-SM were selected for the gene-gene correlation network (Supplementary File: Table S1). The expression values of each sample for the 72 common DEGs in RA, OA, and PG-SM were extracted individually. Three gene-gene correlation matrices were constructed using the Cladist tool [40,43]. Forty-seven genes out of seventy-two common DEGs were highly correlated based on their Pearson correlation coefficient (r-value) values. Table 2 shows the r-values, logFCs, and p-values for each of the 47 highly correlated DEGs in RA, OA, and PG-SM. Genes were selected based on a Pearson correlation coefficient (r score) > 0.7. A gene-gene correlation grid map for the overlapping highly correlated DEGs in RA, OA, and PG-SM is shown in Figure 4, and a Venn diagram representing the overlapping highly correlated gene set is shown in Figure 5. The genes were separated into up-and down-regulated genes based on logFC ≥ ±1. Two genes, IER3 and RGS16, were found to be up-regulated in all three data sets; as a result, they were utilized for additional investigation. These genes may be crucial in the shared pathogenic pathways that underlie the states of these disorders. Table 2. List of highly correlated (r score > 0.7) differentially expressed genes (DEGs) in rheumatoid arthritis (RA), osteoarthritis (OA), and Porphyromonas gingivalis (PG) macrophages, along with their log2-fold change (logFC) ≥ ±1 and p-value ≤ 0.05.

Network Construction
The GeneMANIA plugin was used to predict interactions between the 47 highly correlated common DEGs and 20 additional related genes in the network, using Homo sapiens as the source organism. The network comprised 85 nodes and 399 edges (Supplementary File: Table S2). The pathway interaction information from GeneMANIA showed that the majority of the genes were found to be participating in "Metabolism", "Immune System", "Signaling Pathways", "Class I MHC mediated antigen processing & presentation", "Inflammasomes", "Metabolism of lipids", "Nucleotide-binding domain, leucine rich repeat containing receptor (NLR) signaling pathways", and "The NLRP3 inflammasome". However, our work was focused on physical and genetic interactions to better understand the physical and molecular mechanisms associated with the hub genes. Therefore, all the nodes with less than two edges were removed, and pathway interactions were omitted, leaving a network with 53 nodes and 271 edges ( Figure 6). Among the 53 nodes, 33 were DEGs, and the remaining 20 were related genes.

Network Construction
The GeneMANIA plugin was used to predict interactions between the 47 highly correlated common DEGs and 20 additional related genes in the network, using Homo sapiens as the source organism. The network comprised 85 nodes and 399 edges (Supplementary File: Table S2). The pathway interaction information from GeneMANIA showed that the majority of the genes were found to be participating in "Metabolism", "Immune System", "Signaling Pathways", "Class I MHC mediated antigen processing & presentation", "Inflammasomes", "Metabolism of lipids", "Nucleotide-binding domain, leucine rich repeat containing receptor (NLR) signaling pathways", and "The NLRP3 inflammasome". However, our work was focused on physical and genetic interactions to better understand the physical and molecular mechanisms associated with the hub genes. Therefore, all the nodes with less than two edges were removed, and pathway interactions were omitted, leaving a network with 53 nodes and 271 edges ( Figure 6). Among the 53 nodes, 33 were DEGs, and the remaining 20 were related genes.

Identification of Hub Genes
The Cytoscape plugin cytoHubba was used to calculate the network centrality parameters. In order to identify key nodes in the interaction network, the parameters DC, BC, BN, and CC were selected. The top 15 nodes with high DCs were selected to create a subnetwork of hub genes (Figure 7). Among the top 15 hub genes, 6 genes, namely API5, ATE1, CCNG1, EHD1, RIN2, and SQSTM1, were common DEGs, while the remaining (ATP6V1E1, BAG3, FLNB, PSMD12, STK39, TPM1, TXNIP, TXNL1, XPNPEP1) were related genes added by GeneMANIA (Table 3). Figure 6. The interaction network of highly correlated common differentially expressed (DEGs) as predicted using the GeneMANIA plugin of Cytoscape. The red and green nodes sent common DEGs and related genes, respectively. The red, green, purple, and grey edges sent physical interactions, genetic interactions, co-expression, and consolidated pathways, r tively.

Identification of Hub Genes
The Cytoscape plugin cytoHubba was used to calculate the network centrali rameters. In order to identify key nodes in the interaction network, the parameter BC, BN, and CC were selected. The top 15 nodes with high DCs were selected to cr subnetwork of hub genes (Figure 7). Among the top 15 hub genes, 6 genes, namely ATE1, CCNG1, EHD1, RIN2, and SQSTM1, were common DEGs, while the rem (ATP6V1E1, BAG3, FLNB, PSMD12, STK39, TPM1, TXNIP, TXNL1, XPNPEP1) we lated genes added by GeneMANIA (Table 3).

Gene Ontology (GO) and Pathway Enrichment Analysis of Hub Genes
In order to gain insights into the biological roles of the hub genes, we performed

Gene Ontology (GO) and Pathway Enrichment Analysis of Hub Genes
In order to gain insights into the biological roles of the hub genes, we performed GO enrichment analysis using Enrichr. We collected the GO annotations for 15 hub genes, including biological process (BP), molecular function (MF), and cellular process (CC). Furthermore, significant GO terms with a p-value ≤ 0.05 were selected.
Of the 15 hub genes, 11 were significantly (p-value ≤ 0.05) enriched in 36 GO-BP terms (     FLNB) were the CC terms that were enriched with more than one hub gene.

Discussion
Rheumatoid arthritis (RA), osteoarthritis (OA), and periodontal disease (PD) are chronic inflammatory diseases that are globally prevalent and pose a public health concern. However, most of the research is concentrated on one disease at a time [54,55]. Recent advances in the study of RA and PD, OA and PD, RA and OA, as well as RA, OA, and PD separately, have been made by utilizing the data generated from high-throughput technologies (DNA microarrays and RNAseq), and Omics-based techniques [5,6,8,16,20,37,39,[56][57][58]. However, their applicability is limited when it comes to identifying integrated signature molecules for RA, OA, and PD together. Macrophages play a significant role in disease progression in RA, OA, and PD [6,17,26,29,30,39,45,50,59,60]. A recent study revealed that PG can interfere with the host immune system by preventing intracellular apoptosis and inflammasome activation in macrophages [29]. This is done by modifying the lipopolysaccharide (LPS) composition in order to evade pattern recognition by phagocytes [19]. Apoptosis is involved in the removal of pathogens, diseased or damaged cells, as well as the development of inflammatory diseases via the PI3K/AKT/p53 signaling pathways. Bacterial resistance mechanisms cause anti-apoptosis proteins to invade the host defense system, implying that apoptosis inhibition is an important survival and pathogenicity mechanism [61]. PG or its constituents, such as fimbriae and LPS, act as danger signals, stimulating cell surface receptors such as Toll-like receptors (TLRs) [62] as well as Nod-like receptors (NLRs) [63]. Furthermore, lipids derived from PG can impair osteoblastic function, initiate inflammation, and ultimately alter osteoblast physiology [19,28,64].
OA alters the articular cartilage, synovium, and bone. Articular cartilage is a highly specialized connective tissue found in joints that are composed of chondrocytes. Chon-

Discussion
Rheumatoid arthritis (RA), osteoarthritis (OA), and periodontal disease (PD) are chronic inflammatory diseases that are globally prevalent and pose a public health concern. However, most of the research is concentrated on one disease at a time [54,55]. Recent advances in the study of RA and PD, OA and PD, RA and OA, as well as RA, OA, and PD separately, have been made by utilizing the data generated from high-throughput technologies (DNA microarrays and RNAseq), and Omics-based techniques [5,6,8,16,20,37,39,[56][57][58]. However, their applicability is limited when it comes to identifying integrated signature molecules for RA, OA, and PD together. Macrophages play a significant role in disease progression in RA, OA, and PD [6,17,26,29,30,39,45,50,59,60]. A recent study revealed that PG can interfere with the host immune system by preventing intracellular apoptosis and inflammasome activation in macrophages [29]. This is done by modifying the lipopolysaccharide (LPS) composition in order to evade pattern recognition by phagocytes [19]. Apoptosis is involved in the removal of pathogens, diseased or damaged cells, as well as the development of inflammatory diseases via the PI3K/AKT/p53 signaling pathways. Bacterial resistance mechanisms cause anti-apoptosis proteins to invade the host defense system, implying that apoptosis inhibition is an important survival and pathogenicity mechanism [61]. PG or its constituents, such as fimbriae and LPS, act as danger signals, stimulating cell surface receptors such as Toll-like receptors (TLRs) [62] as well as Nod-like receptors (NLRs) [63]. Furthermore, lipids derived from PG can impair osteoblastic function, initiate inflammation, and ultimately alter osteoblast physiology [19,28,64].
OA alters the articular cartilage, synovium, and bone. Articular cartilage is a highly specialized connective tissue found in joints that are composed of chondrocytes. Chondrocytes maintain the structural and functional integrity of cartilage by synthesizing an extracellular matrix (ECM). This process is known as cartilage homeostasis. Hence, chondrocyte preservation is critical to joint health because articular cartilage lacks blood vessels and nerves and has a limited capacity for intrinsic repair. In response to OA stimuli, however, chondrocytes lose their ability to maintain cartilage integrity and their survival [65]. Inflamed OA synovium and damaged cartilage produce pro-inflammatory cytokines such as interleukin (IL)-1, tumor necrosis factor (TNF)-α, and IL-6 via paracrine or autocrine pathways, and together they activate NF-κB signaling pathways [65,66]. Due to restricted blood supply, limited ECM synthesis, and low cell density, cartilage regeneration is difficult and diminishes during prolonged inflammation [67]. Thus, overexpression of apoptotic inhibitors by PG or its component can be one of the possible mechanisms that explains the interplay of PG in OA. On the other hand, reduced expression of an apoptotic inhibitor results in an increase in apoptosis. Consistently circulating apoptotic debris can induce autoreactive cells and develop immunological complexes [68], which may play a significant part in RA. Since synovial fibroblasts of RA are the most severely impacted by apoptosis, it leads to gradual degradation of articular cartilage [18]. As apoptosis is increased during the early stages of infection and gradually decreases during the later stages of infection, we can assume that PG may be involved in RA during the early stages of infection and that the prolonged infection with PG may result in OA. Hence, analysis of PG-infected host macrophage and synovial macrophage gene expression profiles are required in order to identify the subcellular activities affected by this bacteria.
In this study, we identified DEGs by using gene expression profiles from RA and PG-SM, as well as RNAseq data from OA. A total of 72 common DEGs (RA and PG; Standard Error = 0.012) were found across the data sets, with 47 of them being highly correlated based on their r-values. A GeneMANIA interaction network for the highly correlated DEGs comprised 53 and 271 nodes and edges, respectively ( Figure 6). Among the 53 nodes, 33 were DEGs, while the remaining 20 were related genes added by GeneMANIA (Table 3). Network centrality analysis identified 15 hub genes, of which 6 genes (API5, ATE1, CCNG1, EHD1, RIN2, and SQSTM1) were DEGs while the remaining 9 (ATP6V1E1, BAG3, FLNB, PSMD12, STK39, TPM1, TXNIP, TXNL1, and XPNPEP1) were related genes.
The differential gene expression analysis showed that API5, ATE1, CCNG1, EHD1, IER3, and RGS16 were significantly dysregulated in all three macrophage conditions. Two DEGs, RGS16 (RA, logFC = 1.141; OA, logFC = 2.358; PG, logFC = 1.065) and IER3 (RA, logFC = 1.583; OA, logFC = 1.611; PG, logFC = 3.691), were considerably elevated in all macrophage conditions as compared to other DEGs which were either up/down-regulated. The GeneMANIA interaction network showed that RGS16 is linked to hub genes CCNG1 and TXNIP, whereas IER3 is linked to hub genes ATE1, EHD1, SQSTM1, and BAG1. Both RGS16 and IER3 are associated with the IL-18 signaling pathway, which is known to influence the activity of osteoclasts and osteoblasts/chondrocytes, and may operate as a physiological regulator of bone development [69]. Therefore dysregulation of the IL-18 signaling pathway plays a major role in RA [69] and OA [70]. The IL-18 signaling pathway also plays an important role in pro-inflammatory cytokine production in PD, which is the main cause of alveolar bone loss in the presence of PG [71]. Therefore, even though IER3 and RGS16 are non-hub genes, it is crucial to understand their role in three macrophage conditions. Interestingly, hub genes SQSTM1, FLNB, TXNL1, and ATP6V1E1 had previously been related to one or more of the three diseases. SQSTM1 (sequestosome1) is already known to play a critical role in the evolution of OA via microRNAs (miRNAs). It also inhibits cytokine expression in activated macrophages [72,73]. FLNB (filamin B) plays a role in the spondylocarpotarsal synostosis (SCT) syndrome, Larsen syndrome, atelosteogenesis types I and III (AOI and AOIII), and Piepkorn osteochondrodysplasia (POCD), which are also associated with the bone disorders [74]. FLNB is largely expressed in macrophages, and governs skeletal development by promoting the migration, proliferation, and secretion of pro-inflammatory cytokines as well as the synthesis of metallopeptidase-9 (MMP-9) and extracellular signal-regulated kinase (ERK) activity. Dysregulation of FLNB leads to various bone diseases. Recently, FLNB has been linked to periodontal ligament and bone marrow development [74][75][76][77][78][79]. TXNL1 (thioredoxin-like 1) is a redox-scavenger gene involved in the management of oxidative stress, which is a critical factor in the development of RA, OA, and PG [79]. ATP6V1E1 (ATPase H+ transporting V1 subunit E1) is associated with osteoclastic activity, which destroys joints and affects inflammatory responses such as phagocytosis, cytokine secretion, and neutrophil granule exocytosis [80]. This shows that the related genes added by GeneMANIA are not random, but play a crucial role in the aetiology of RA, OA, and PD.
ATE1 (Arginyltransferase 1) is down-regulated in RA (logFC = −0.683), OA (logFC = −0.909), and PG (logFC = −1.018). ATE1 catalyzes the post-translational conjugation of arginine to the N-terminal aspartate or glutamate of a protein required for ubiquitin-mediated degradation. Furthermore, down-regulation of ATE1 allows bacteria to develop in the presence of stress stimuli, such as oxidative stress or high nitric oxide concentrations, which can induce cell growth inhibition or cell death, as well as chronic inflammation and resistance. Numerous studies have shown that sufficient arginylation is required for anti-apoptotic action in the presence of an apoptotic inhibitor as well as a protein that destroys pro-inhibitors or similar proteins, such as G-coupled protein regulators [85,86] EHD1 (EH Domain Containing 1) is a member of a conserved gene family that produces EH domain-containing proteins, and is overexpressed in RA (logFC = 2.131) and PG (logFC = 2.084), but underexpressed in OA (logFC = −2.591). It controls phosphophagocytosis, cytokine secretion, cell proliferation, and motility, as well as macrophage receptor activation and signaling [87]. EHD1 also regulates protein-protein interactions, intracellular sorting, and IGF1 receptor endocytosis. EHD1 mediates IGF1 receptor endocytosis through an adaptor protein complex. IGF1 is important in cartilage metabolism; it promotes chondrocyte proliferation, matrix protein production, terminal differentiation, and mineralization, as well as raising cellular calcium levels [88]. EHD1 dysregulation impairs IGF1-mediated signaling, which results in RA and OA [79]. PG is also known to increase IGF1 expression [89], and one possible explanation could be EHD1 dysregulation, which leads to RA and OA.
CCNG1 (Cyclin G1) is a topologically significant gene with the highest bottleneck value (BN: 7) and is underexpressed in all three macrophage conditions (RA, logFC = −0.712; OA, logFC = −1.613; PG, logFC = −0.784). CCNG1 regulates cell proliferation, apoptosis, migration, invasion, and cell cycle arrest. Inhibiting CCNG1 activates p53 and causes apoptosis [90]. PG can regulate cell cycle disruption and apoptosis through the involvement of p53 [91]. p53 regulates CCNG1 expression, and its dysregulation is linked to osteosarcoma (bone cancer) [92]. Furthermore, CCNG1 regulates mesenchymal stem cell (MSC) differentiation, and the absence or loss of p53 causes abnormal osteogenic differentiation of MSCs, which results in imbalanced bone remodeling [93]. As a bottleneck gene, CCNG1 may be an important link in explaining the role of PG in RA and OA.
For RIN2 (Ras and Rab interactor 2), Enrichr-Jensen DISEASES analysis associated RIN2 with gingival overgrowth, and literature studies associate it with endocytosis and bone formation [94]. RIN2, a widely expressed protein, interacts with Rab5 to regulate endocytic trafficking, as well as to target and fuse endocytic vesicles to early endosomes [95].
The activation of Rab5 via TNF-α increases the ability of PG to invade human gingival epithelial cells [96].
We also examined the importance of the related gene STK39 (serine/threonine kinase 39), which appears to be significant based on topological parameters (DC:19, BC:139.88; BN:2; CC:34.33). It is involved in the stress response pathway, actively regulating the p38 MAP kinase pathway and TCR signaling in naive CD4+ T cells. It is well-known for its role in inflammatory diseases, macrophage activation, and bacterial growth [97,98]. According to research, STK39 dysregulation affects MMP2 (matrix metallopeptidase 2) and MMP9 (matrix metallopeptidase 9), thus actively participating in osteosarcoma [99]. As a result, it may play an important role in other bone-related diseases such as RA, OA, and PD.
The major limitation of our study is that the data we collected are from different years and each data set has a methodological constraint of being generated using different protocols. Despite these differences, the variances are reasonably low. Our results should be interpreted in terms of variance among the data.

Conclusion
The results of the present study proposed 8-gene signatures including API5, ATE1, CCNG1, EHD1, IER3, RGS16, RIN2, and STK39 by employing gene expressions and networkbased studies. These gene signatures could aid in the early diagnoses of RA, OA, and PD; however, experimental studies are needed to confirm these findings.