Medication-Related Osteonecrosis of the Jaw: A Systematic Review and a Bioinformatic Analysis

The objective was to evaluate the current evidence regarding the etiology of medication-related osteonecrosis of the jaw (MRONJ). This study systematically reviewed the literature by searching PubMed, Web of Science, and ProQuest databases for genes, proteins, and microRNAs associated with MRONJ from the earliest records through April 2023. Conference abstracts, letters, review articles, non-human studies, and non-English publications were excluded. Twelve studies meeting the inclusion criteria involving exposure of human oral mucosa, blood, serum, saliva, or adjacent bone or periodontium to anti-resorptive or anti-angiogenic agents were analyzed. The Cochrane Collaboration risk assessment tool was used to assess the quality of the studies. A total of 824 differentially expressed genes/proteins (DEGs) and 22 microRNAs were extracted for further bioinformatic analysis using Cytoscape, STRING, BiNGO, cytoHubba, MCODE, and ReactomeFI software packages and web-based platforms: DIANA mirPath, OmicsNet, and miRNet tools. The analysis yielded an interactome consisting of 17 hub genes and hsa-mir-16-1, hsa-mir-21, hsa-mir-23a, hsa-mir-145, hsa-mir-186, hsa-mir-221, and hsa-mir-424. A dominance of cytokine pathways was observed in both the cluster of hub DEGs and the interactome of hub genes with dysregulated miRNAs. In conclusion, a panel of genes, miRNAs, and related pathways were found, which is a step toward understanding the complexity of the disease.


Introduction
Medication-related osteonecrosis of the jaw (MRONJ) encompasses osteonecrosis of the jaw associated with bisphosphonates, denosumab, and anti-resorptive agents [1][2][3][4].It is a rare but serious drug reaction associated often with receiving high doses of anti-angiogenic and anti-resorptive medication, including mTOR inhibitors [5][6][7].Anti-resorptive drugs, namely bisphosphonate and denosumab, are monoclonal antibodies that target the receptor activator of the tumor necrosis factor ligand superfamily member 11 [8,9].Under conditions of accelerated skeletal turnover, bisphosphonates are selectively absorbed at sites of active bone remodeling [10].Anti-angiogenic drugs, namely sunitinib and bevacizumab, are humanized monoclonal antibodies that impede the creation of novel blood vessels by suppressing the function of tyrosine kinases and vascular endothelial growth factor (VEGFA) [11].
MRONJ is a recognized phenomenon in nearly 1% of cancer patients and in 0.1% of those suffering from metabolic bone diseases [12].The prevalence of MRONJ reported in studies varies widely, with incidence rates ranging from as low as 0.01% following low-dose oral bisphosphonate therapy to as high as 14.4% in high dose intravenous bisphosphonate treatment [13].Dental extractions, implant procedures, oral and maxillofacial surgeries, periodontal disease, and invasive periodontal procedures have been identified as risk logical operator "OR" and the results with the logical operator "AND".The following terms were used in the search strategy: "mronj", "bronj", "aronj", "dronj", "medication-related osteonecrosis of the jaw", "bisphosphonate-related osteonecrosis of the jaw", "antiresorptive agent-related osteonecrosis of the jaw", and "denosumab-related osteonecrosis of the jaw".In addition, the search terms for transcriptomics were "microRNA" or "miRNA" and "transcriptome", "transcriptomics" or "mRNA".Proteomics was searched using the terms "proteomics" or "proteome", while metabolomics was obtained by searching for "metabolome", "metabolomics" or "metabolite".No year restrictions were applied for article publication.
Int. J. Mol.Sci.2023, 24, x FOR PEER REVIEW 3 of 20 mRNAs, microRNAs (miRNAs), proteins, and metabolites.The study merged key phrases with the logical operator "OR" and the results with the logical operator "AND".The following terms were used in the search strategy: "mronj", "bronj", "aronj", "dronj", "medication-related osteonecrosis of the jaw", "bisphosphonate-related osteonecrosis of the jaw", "antiresorptive agent-related osteonecrosis of the jaw", and "denosumab-related osteonecrosis of the jaw".In addition, the search terms for transcriptomics were "mi-croRNA" or "miRNA" and "transcriptome", "transcriptomics" or "mRNA".Proteomics was searched using the terms "proteomics" or "proteome", while metabolomics was obtained by searching for "metabolome", "metabolomics" or "metabolite".No year restrictions were applied for article publication.Further, the screening process involved manually removing duplicate results from the analysis.Two reviewers independently analyzed the titles and abstracts of the papers and evaluated the remaining articles to determine their eligibility.If a study's suitability could not be determined solely from its title and abstract, its full text was examined.Citations for the included papers were tracked using Google Scholar or PubMed.A manual review of the reference lists of the included articles was conducted to select relevant articles.The results were then summarized from the articles that satisfied the inclusion criteria.The Revised Cochrane risk-of-bias tool for randomized trials (RoB 2, Further, the screening process involved manually removing duplicate results from the analysis.Two reviewers independently analyzed the titles and abstracts of the papers and evaluated the remaining articles to determine their eligibility.If a study's suitability could not be determined solely from its title and abstract, its full text was examined.Citations for the included papers were tracked using Google Scholar or PubMed.A manual review of the reference lists of the included articles was conducted to select relevant articles.The results were then summarized from the articles that satisfied the inclusion criteria.The Revised Cochrane risk-of-bias tool for randomized trials (RoB 2, https://www.riskofbias.info/welcome/rob-2-0-tool/current-version-of-rob-2, accessed on 14 November 2023) was used to assess the risk of bias in domains related to the randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result.

Gene Ontology Enrichment Analysis
The GeneCards human gene database (https://www.genecards.org,accessed on 10 July 2023) was used to verify and revise the list of differentially expressed genes (DEGs) compiled from eligible articles.Protein accession numbers were mapped to genes using the UniProt mapping tool (https://www.uniprot.org,accessed on 14 July 2023) [23].The downloaded matching genes were then used for further investigation.
To investigate DEGs for overrepresentation in the hierarchical gene ontology (GO), the extensions of Cytoscape 3.10.0[24] and BiNGO 3.0.5 [25] were used.The enrichment analysis for cell components, molecular function, and biological process terms was performed using the Benjamini and Hochberg multiple testing procedure.The significance level was set at 0.05 (p < 0.05).

Protein-Protein Interaction Network and Module Analysis
The network of DEGs was created using the Search Tool for the Retrieval of Interacting Genes (STRING) database, which presents both insights and predictions regarding protein-protein interactions (PPIs).The network itself was constructed with the help of StringApp, Version 2.0.1 [26].

Pathway Enrichment Analysis
The ReactomeFI plug-in pathway database version 8.0.6 [29] was used for pathway enrichment analysis.
The mirPath v.4 database from DIANA Tools (https://diana-lab.e-ce.uth.gr/app/miRPathv4, accessed on 17 July 2023) was utilized to identify genes that could serve as miRNA target candidates.

Multi-Omics Network
The multi-omics data were integrated using web-based platforms such as OmicsNet (https://www.omicsnet.ca,accessed on 19 July 2023) [30] and miRNet (https://www.mirnet.ca,accessed on 21 July 2023) [31].If the multi-omics network consisted of more than 3000 nodes, we implemented the minimum network setting, i.e., the algorithms that generate the minimum network connecting all specified nodes.

Systematic Review of Screening for MRONJ
The search strategy produced 998 articles.Twenty-five articles underwent full-text review following a screening of their titles and abstracts.Twelve articles were ultimately included in the library after thirteen articles were excluded following a thorough examination of their full text (Figure 2).Exclusion criteria comprised such items as conference abstracts, letters, and review articles; non-human studies; and publications in languages other than English.Table 1 summarizes the characteristics of the studies on MRONJ that were included.Nine of the twelve studies that met the inclusion criteria were found to have an overall risk of bias of some concern, and three were found to have a high overall risk of bias.
Int. J. Mol.Sci.2023, 24, x FOR PEER REVIEW 5 of 20 abstracts, letters, and review articles; non-human studies; and publications in languages other than English.Table 1 summarizes the characteristics of the studies on MRONJ that were included.Nine of the twelve studies that met the inclusion criteria were found to have an overall risk of bias of some concern, and three were found to have a high overall risk of bias.[32] Peripheral blood mononuclear cells.Patients: MM patients with ONJ (n = 8).
RNA extraction: the Total Purification Plus Kit (Norgen Biotek Corporation, Thorold, ON, Canada).Total RNA was transcribed into cDNA through an All-in-One miRNA first-strand cDNA synthesis kit (GeneCopoeia Inc., Rockville, MD, USA).Real-Time qPCR employed a 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA).

Network Analysis of Protein Interaction Data
To investigate the protein interactions involved in MRONJ pathogenesis, we utilized STRING databases to analyze the 824 identified genes/proteins and created an interactive network via Cytoscape.Figure 3 depicts the resulting network consisting of 701 genes and 3993 edges, while Table 2 summarizes the network topology.The network topology was analyzed using the cytoHubba (Version 0.1) extension of Cytoscape.The highly linked hub genes were extracted from the main complex network of DEGs by using the algorithms of cytoHubba.Subsequently, 24 genes were extracted that occurred at the intersection of at least three methods: ALB, ANXA5, ATM, CCL2, CD44, CXCL8, CXCR4, EEF2, EGF, GART, HSP90AB1, HSPA4, IGF1, IL1B, IL6, ITGB1, JUN, LMNA, MMP9, PTPRC, RAB5A, RHOA, TNF, and VEGFA (see Table 3).
No relevant studies have been found regarding the metabolomics of MRONJ.

Network Analysis of Protein Interaction Data
To investigate the protein interactions involved in MRONJ pathogenesis, we utilized STRING databases to analyze the 824 identified genes/proteins and created an interactive network via Cytoscape.Figure 3 depicts the resulting network consisting of 701 genes and 3993 edges, while Table 2 summarizes the network topology.

Topological Parameters Values
The average number of neighbors The network topology was analyzed using the cytoHubba (Version 0.1) extension of Cytoscape.The highly linked hub genes were extracted from the main complex network  In complement to the cytoHubba algorithms, MCODE clustering was employed to detect the molecular complexes and the seeds-the hub genes with a high degree of connectivity.In a complex PPI network of DEGs, MCODE identified five of the intra-connected regions/clusters and five hub genes/seeds with a high degree of connectivity (see Table 4).In conjunction with cluster 1, the highest scoring MCODE clustering module, with the cytoHubba analysis results, a total of 17 hub genes were retrieved, comprising ALB, CD44, CXCL2, CXCL8, CXCR4, EEF2, EGF, IGF1, IL1B, IL6, ITGB1, JUN, MMP9, PTPRC, RHOA, TNF, and VEGFA.

GO Enrichment Analysis
A survey of how genes and gene products are represented in the biological domains concerning three aspects of molecular biology was conducted using Gene Ontology (GO).To associate GO terms with gene and protein sets, a series of enrichment analyses were performed in Cytoscape with the help of the BiNGO extension.
The dysregulated genes within the complex gene panel were linked to 422 GO terms, consisting of 333 biological process terms, 23 molecular function terms, and 66 cell component terms, as identified by GO enrichment analysis.The top GO enrichment terms associated with DEGs by p-values are shown in Table 5.The key molecular biological processes identified among these genes are those involved in regulating the immune system and the organism's immune response.Numerous genes have been implicated in protein binding.DEGs were predominantly linked to the extracellular region and the extracellular space.
A total of 918 GO terms were obtained from the analysis of seventeen hub genes.Of these GO terms, 893 were related to biological processes, 22 were associated with molecular function, and 23 were linked to cell component terms.Protein phosphorylation of amino acids constitutes a central molecular biological process.While protein binding was a common association among all hub genes, the top-ranked molecular function was cytokine receptor binding.DEGs were predominantly located in the extracellular space and region, as indicated in Table 6.

Multiomics Networks in MRONJ
To investigate the fundamental mechanisms of MRONJ regulation, OmicsNet tools to visualize multi-layered networks with a 3D-based layered layout were used.We detected a complex intrinsic network that was eventually reduced to a minimally connected network consisting of 1300 nodes (1289 genes/proteins and 11 miRNAs) and 7816 edges after merging the initial set of 22 miRNAs, 550 genes, and 292 proteins (see Figure 4).

Pathway Enrichment Analysis
To identify the pathways involved in medication-induced osteonecrosis of the jaw, we analyzed the associated genes/proteins of the complex panel of DEGs using the ReactomeFI tool within Cytoscape.Table 7 and Figure 6a indicate that DEGs were notably enriched in the top ten signaling pathways, specifically in the Innate immunity system pathways.These pathways form the nonspecific part of immunity and include functions such as Neutrophil degranulation (R-HSA-6798695) and regulation of the complement cascade (R-HSA-6803157).Similarly, pathway enrichment analysis was performed on the significantly dysregulated Reactome signaling pathways using the set of 17 hub genes/proteins identified in MRONJ (Table 8, Figure 6b).

Pathway Enrichment Analysis
To identify the pathways involved in medication-induced osteonecrosis of the jaw, we analyzed the associated genes/proteins of the complex panel of DEGs using the ReactomeFI tool within Cytoscape.Table 7 and Figure 6a indicate that DEGs were notably enriched in the top ten signaling pathways, specifically in the Innate immunity system pathways.These pathways form the nonspecific part of immunity and include functions such as Neutrophil degranulation (R-HSA-6798695) and regulation of the complement cascade (R-HSA-6803157).Similarly, pathway enrichment analysis was performed on the significantly dysregulated Reactome signaling pathways using the set of 17 hub genes/proteins identified in MRONJ (Table 8, Figure 6b).Table 7. Pathway enrichment analysis of the Reactome signaling pathways dysregulated in MRONJ ranked according to their p-values.The full set of dysregulated genes/proteins was considered in the analysis.To map the signaling pathways of MRONJ and to identify potential Reactome molecular pathway targets of miRNAs, the associated miRNAs (hsa-mir-16-1, hsa-mir-21, hsa-mir-23a, hsa-mir-145, hsa-mir-186, hsa-mir-221, and hsa-mir-424), obtained with the OmicsNet tools (Figure 5), underwent an in silico analysis with the DIANA Tools mirPath v.4 database.MiRNA-centric analysis of hsa-mir-16-1, hsa-mir-21, hsa-mir-23a, hsa-mir-145, hsa-mir-186, hsa-mir-221, and hsa-mir-424 was carried out with the help of the TarBase v8.0 database and miTarBase2022 as a secondary target source, searching for the direct miRNA target genes.The union of Reactome pathways option was used in the analysis.The results of the analysis are shown in Table 9.The Reactome pathways Interleukin-4 and interleukin-13 signaling (R-HSA-6785807) and signaling by interleukins Table 8.Pathway enrichment analysis of Reactome signaling pathways dysregulated in MRONJ ranked according to their p-values.The set of seventeen hub-dysregulated genes/proteins was considered in the analysis.To map the signaling pathways of MRONJ and to identify potential Reactome molecular pathway targets of miRNAs, the associated miRNAs (hsa-mir-16-1, hsa-mir-21, hsa-mir-23a, hsa-mir-145, hsa-mir-186, hsa-mir-221, and hsa-mir-424), obtained with the OmicsNet tools (Figure 5), underwent an in silico analysis with the DIANA Tools mirPath v.4 database.MiRNA-centric analysis of hsa-mir-16-1, hsa-mir-21, hsa-mir-23a, hsa-mir-145, hsa-mir-186, hsa-mir-221, and hsa-mir-424 was carried out with the help of the TarBase v8.0 database and miTarBase2022 as a secondary target source, searching for the direct miRNA target genes.The union of Reactome pathways option was used in the analysis.The results of the analysis are shown in Table 9.The Reactome pathways Interleukin-4 and interleukin-13 signaling (R-HSA-6785807) and signaling by interleukins (R-HSA-449147) show the most comprehensive association of miRNAs with gene targets consisting of IL1B, VEGFA, and CXCL8.

Discussion
To explore the pathological mechanisms of MRONJ, we utilized gene profiling datasets, proteins, and miRNAs.A variety of analytical strategies were employed to investigate the molecular mechanisms underlying MRONJ, including PPI network analysis, GO enrichment, and Reactome pathway enrichment analysis.
Pathway enrichment analysis is a valuable tool for gaining a mechanistic comprehension of the intricate gene, miRNA, and protein inventories resulting from omics experiments.It assists in the interpretation of biomedical data to reveal the molecular basis of disease [48].Our analysis identified immune dysfunction-related pathways associated with MRONJ as the main reason for developing and progressing osteonecrosis.When analyzing the entire pool of genes and proteins, it is evident that the Reactome signaling pathways that are significantly dysregulated are primarily dominated by the Innate immune system and Neutrophil degranulation pathways, as indicated by the p-values.Reducing the set of hub genes to 17, which includes ALB, CD44, CXCL2, CXCL8, CXCR4, EEF2, EGF, IGF1, IL1B, IL6, ITGB1, JUN, MMP9, PTPRC, RHOA, TNF, and VEGFA, resulted in the dominance of cytokine signaling pathways in the immune system.Moreover, simultaneously analyzing hub genes with miRNAs (hsa-mir-16-1, hsa-mir-21, hsa-mir-23a, hsa-mir-145, hsa-mir-186, hsamir-221, and hsa-mir-424) using network-based multiomics analysis revealed dysregulated pathways in the immune system's cytokine signaling, specifically signaling by interleukins.
There is increasing evidence that inflammatory osteoimmunology is critical to the development of osteonecrosis [49].Cytokines that regulate inflammatory responses contribute to the onset and progression of osteonecrosis [17,50,51].In individuals with osteonecrosis, necrotic cells produce cytokines that attract inflammatory cells, triggering both local and systemic immune responses [17,52,53].
The pathogenesis of inflammatory bone disease is significantly influenced by T cells and B cells [56].There have been intense discussions regarding new roles for B cells and a potential role for peripheral blood γδ T cells [57,58].Γδ T cells are innate lymphocytes with a crucial role in regulating immune homeostasis [59].Kalyan et al. investigated the potential predictive role of peripheral blood γδ T cells in osteonecrosis of the jaw.The authors propose that the loss of γδ T cells caused by bisphosphonates may be involved in the development of osteonecrosis [57].Moreover, the proliferation of macrophages and γδ T cells promotes inflammation in zoledronic acid-induced jaw necrosis, as the authors of the study [60] concluded.
The understanding of biological systems is facilitated by the objective study of PPIs.An effective approach to characterizing system-wide PPIs is the use of PPI networks.These networks are constructed from pairwise protein interactions and serve as an efficient tool for describing PPI landscapes [61].To investigate protein functions and biological processes based on predicted PPIs and to gain new insights into diseases, the DEG PPI network was analyzed in this study.For osteonecrosis of the jaw, 17 hub genes with aberrant expression were selected.They included ALB, CD44, CXCL2, CXCL8, CXCR4, EEF2, EGF, IGF1, IL1B, IL6, ITGB1, JUN, MMP9, PTPRC, RHOA, TNF, and VEGFA.
Various signaling molecules, such as VEGFA, EGF, MMP9, and TNF, contribute to angiogenesis by stimulating the proliferation and migration of vascular endothelial cells [62].VEGFA is a highly potent pro-angiogenic factor that plays a critical role in the healing of microvascular wounds associated with bisphosphonate administration [63].A statistically significant increase in VEGFA gene expression was also demonstrated in response to zoledronic acid [64].The physiological processes of angiogenesis and vascular remodeling involve the regulation of non-coding RNAs, specifically miRNA-based regulation (as noted by reference [65]).Objective evaluation of these processes is necessary for accurate understanding.
MicroRNAs are small endogenous RNA molecules (∼22 nt) that were recently discovered.Disorders such as cancer or heart disease have demonstrated the diagnostic potential of circulating miRNAs [66,67].MiRNA-mediated RNA interference, a unique mechanism that binds miRNAs to different direct targets, controls both post-transcriptional gene expression and protein expression [68].
The PPI network revealed that CXCL8 [71] is among the factors related to the progression of osteonecrosis of the jaw, and our analysis identified it as a direct target of miR-23a.Earlier studies reported a significant reduction of miR-23a-3p during osteogenic differentiation of human mesenchymal stem cells of bone marrow origin [72].Furthermore, inhibition of miR-23a in a rat model resulted in a lower incidence of osteonecrosis [73].
Serum microRNAs, including miR-21, miR-23a, and miR-145, were observed to be dysregulated in BRONJ rats [37].Furthermore, candidate microRNA expressions were confirmed in human samples.During the progression of BRONJ, there was an upregulation of circulating miR-21, which corresponds to the alteration of miR-21 in proosteoclastogenesis [37].

Conclusions
The emergence of high-throughput platforms for the comprehensive analysis of genes, proteins, and other biological molecules has afforded an exceptional capability for the recognition of novel, valid signatures of disease-related processes.
In conclusion, our systematic review study indicates specific alterations in proteins, genes, and microRNAs and thus unravels novel insights into the molecular mechanism behind the MRONJ disease.The identified dysregulated genes in MRONJ are mostly linked to the regulation of immune system processes and the immune response of the organism.These dysregulated genes significantly enrich pathways related to the Innate immunity system, a crucial component of the nonspecific part of immunity.Particularly important are the 17 hub genes, which exert dominance in the cytokine signaling pathways within the immune system.Additionally, the interaction network between these hub genes and dysregulated miRNAs uncovered pathways associated with the cytokine signaling in the immune system, particularly the signaling by interleukins pathway.Subsequent miRNA analysis showed a set of highly connected miRNAs with direct targeting of multiple genes such as IL1B, VEGFA, CXCL8, and CD44.
This study has potential limitations that should be noted.There was considerable heterogeneity observed between studies, which may impact the interpretation of the results.Factors such as patient selection variability, differences in the origins and causes of MRONJ, and variations in the material and methodology used could contribute to this heterogeneity.To ensure more reliable results, it would be beneficial to establish strict inclusion/exclusion criteria based on the disease state and MRONJ treatment in future studies.Additionally, our objective was to present a comprehensive overview of the pathophysiological processes of MRONJ in humans using multi-omics techniques.Expanding the research to encompass other types of non-coding RNAs could prove helpful in filling this gap.Furthermore, there is a shortage of data for analysis due to the limited amount of metabolomics research.
Despite the limitations of this study, the panel of proteins, genes, and microRNAs presented, along with their associated pathways, constitutes a significant advancement toward comprehending the intricate cause of MRONJ.

Figure 1 .
Figure 1.PRISMA flow diagram of literature for a systematic review on the screening of medicationinduced osteonecrosis of the jaw.

Figure 1 .
Figure 1.PRISMA flow diagram of literature for a systematic review on the screening of medicationinduced osteonecrosis of the jaw.

Figure 3 .
Figure 3.The STRING network formed from the dysregulated genes/proteins in MRONJ.

Figure 3 .
Figure 3.The STRING network formed from the dysregulated genes/proteins in MRONJ.

Figure 5 .
Figure 5.The multi-omics network of MRONJ.Input genes-large pink circles; input miRNAs-large blue squares; connector genes-small pink circles; connector miRNAs-small blue squares.

70 19 R 5 Figure 6 .
Figure 6.Pathway enrichment analysis of significantly dysregulated Reactome signaling pathways ranked according to the p-values; (a) the full set of dysregulated genes/proteins, (b) the set of hub genes/proteins of MRONJ.

Figure 6 .
Figure 6.Pathway enrichment analysis of significantly dysregulated Reactome signaling pathways ranked according to the p-values; (a) the full set of dysregulated genes/proteins, (b) the set of hub genes/proteins of MRONJ.

Table 1 .
Characteristics of included studies on medication-induced osteonecrosis of the jaws.

Table 1 .
Characteristics of included studies on medication-induced osteonecrosis of the jaws.

Table 2 .
The most relevant topological parameters of the STRING network.

Table 2 .
The most relevant topological parameters of the STRING network.

Table 3 .
Hub genes identified by the cytoHubba plug-in algorithms.The crucial hub genes that occurred at the intersection of at least three methods are highlighted in red.

Table 4 .
MCODE-interconnected clusters generated from the Cytoscape-derived gene interaction network.The seed node with the highest score within the cluster is marked with an asterisk.

Table 5 .
Gene ontology enrichment analysis performed in Cytoscape using the BiNGO extension.The full set of dysregulated genes/proteins was considered in the analysis.The most enriched gene ontology terms based on the respective p-values are depicted.

Table 6 .
Gene ontology enrichment analysis in Cytoscape using the BiNGO extension.A set of seventeen hub-dysregulated genes/proteins was considered in the analysis.The most enriched gene ontology terms based on the respective p-values are depicted.

Table 7 .
Pathway enrichment analysis of the Reactome signaling pathways dysregulated in MRONJ ranked according to their p-values.The full set of dysregulated genes/proteins was considered in the analysis.