Integrative multi-omics and bioinformatic analysis of MMP20 in amelogenesis imperfecta Authors: Tina Leban, Tanja Kunej Published: 2026 ABSTRACT ObjectivesTo systematically review and integrate current evidence on MMP20, a key gene associated with autosomal recessive hypomaturation amelogenesis imperfecta, using a multi-omics framework complemented by integrated bioinformatic analyses.MethodsA systematic literature search was conducted in the PubMed and Web of Science databases from database inception to March 2026 using keywords related to MMP20 and multiple omics domains. Eligible studies were systematically screened, quality assessed, and integrated with data from 16 bioinformatic databases and tools.ResultsOf 186 screened publications, 67 met the inclusion criteria. Thirteen omics layers were evaluated, with available data identified for nine layers. Integrated analysis identified 37 reported sequence variants, 173 predicted microRNA-binding sites, 11 post-translational modification sites, and one sequence variant located within a post-translational modification site. Protein–protein interaction analysis revealed a highly interconnected network comprising MMP20 and 17 other amelogenesis imperfecta-associated proteins.ConclusionsThis systematic review provides a multi-omics overview of MMP20 by integrating published evidence with bioinformatic analyses. The identified variants, regulatory elements, post-translational modifications, and interaction networks expand current understanding of the molecular landscape surrounding MMP20 and its role in enamel development. The proposed workflow may facilitate variant interpretation and serve as a framework for gene–disease investigations. Introduction Amelogenesis imperfecta (AI) is a heterogeneous group of inherited enamel defects that may occur either in isolation or in association with systemic conditions [1]. Based on clinical presentation, AI is classified into four major types: hypoplastic, hypomaturation, hypocalcified, and hypomaturation–hypoplastic with taurodontism [1]. Depending on the inheritance pattern, non-syndromic AI is associated with sequence variants in more than 18 genes involved in enamel formation and maturation, including genes encoding enamel matrix proteins, enamel proteases, ion transporters, intracellular trafficking proteins, and cell adhesion molecules [2]. Among these genes, variants in MMP20 cause autosomal recessive hypomaturation AI (type IIA2; OMIM #612529). The MMP20 gene (matrix metalloproteinase 20), also known as enamelysin, is located on chromosome 11q22.2 and encodes a zinc-dependent endopeptidase active during the secretory and early maturation stages of amelogenesis [3]. MMP20 mediates the proteolytic processing of enamel matrix proteins, including amelogenin, enamelin, and ameloblastin, thereby establishing the extracellular matrix framework required for normal enamel mineralization. This proteolytic remodeling provides the structural basis for the nucleation and growth of hydroxyapatite crystals, coordinated with tightly regulated phosphate homeostasis [4,5]. Loss-of-function variants impair matrix protein degradation, resulting in protein retention and incomplete enamel mineralization characteristic of AI type IIA2. Clinically, AI type IIA2 is characterized by enamel of normal thickness but reduced hardness and mineralization, frequently accompanied by yellow-brown discoloration and post-eruptive breakdown. Radiographically, enamel radiodensity is similar to, or only slightly greater than, that of dentin [6]. These clinical manifestations reflect the underlying retention of enamel matrix proteins and impaired hydroxyapatite crystal growth and maturation caused by defective MMP20-mediated proteolysis, resulting in structurally and mechanically compromised enamel. Consequently, the altered enamel substrate directly affects restorative outcomes. In severely compromised AI enamel, placement of restorations or build-ups may be challenging because adhesive systems generally exhibit lower bond strength to affected enamel than to healthy enamel [7,8]. Consequently, restoration longevity is frequently reduced [9]. In addition, the effectiveness of enamel deproteinization prior to the application of self-etch adhesive systems in hypomaturated enamel remains controversial, since studies have reported inconsistent effects on shear bond strength [8,10]. These findings highlight the importance of understanding the molecular and structural characteristics of AI enamel in order to optimize clinical management strategies for patients with AI. The aim of this systematic review was to collect, classify, and synthesize the available literature on the association between MMP20 and AI type IIA2. The published evidence was integrated with data from multiple omics domains and complemented by bioinformatic databases and analytical tools to provide a comprehensive overview of reported variants and their functional relevance. This integrative approach may improve the consistency of variant interpretation and support future research on MMP20-related AI. Materials and methods This study was conducted as a systematic review with integrating bioinformatic analysis. The methodological framework and reporting structure strategy followed the workflow proposed by Leban and Kunej (2025) to ensure methodological consistency, reproducibility, and traceability of the integrated multi-omics analyses. Literature search, study selection and data extraction A systematic literature search was conducted in the PubMed and Web of Science databases, covering publications from database inception to March 2026. The initial PubMed search was performed using the exact query: “amelogenesis imperfecta”[Title/Abstract] OR “AI”[Title/Abstract] AND “MMP20”[Title/Abstract] AND “genomics” OR “transcriptomics” OR “proteomics” OR “epigenomics” OR “multi-omics”. As no eligible studies were identified a broader search was subsequently performed in both PubMed and Web of Science using the exact query: “amelogenesis imperfecta” AND “MMP20”. The search was limited to English-language original research articles investigating MMP20-related AI in humans or experimental models. Methodological quality assessment The methodological quality of the included studies was evaluated using a predefined assessment framework developed for this review. Because the included literature comprised heterogeneous study designs, including human genetic studies, animal experiments, in vitro investigations, and bioinformatic analyses, no single standardized risk-of-bias tool was considered appropriate. Each study was independently evaluated using nine methodological criteria: (i) clearly defined study objective, (ii) appropriate study design, (iii) adequately described sample size, (iv) appropriate statistical and/or bioinformatic analyses, (v) independent validation cohort and/or experimental validation, (vi) public availability of data, (vii) reproducibility of the analytical workflow, (viii) biological interpretation supported by the presented data, and (ix) discussion of study limitations. Each criterion was scored as 0 (No), 1 (Partially), or 2 (Yes), yielding a maximum possible score of 18 points. Based on the total score, studies were classified as follows: 17–18 points, excellent methodological quality; 15–16 points, high methodological quality; 13–14 points, good methodological quality; and 10–12 points, moderate methodological quality. Bioinformatic analyses Gene annotations and genomic information, including chromosomal location, synteny, sequence variants, Genomic Evolutionary Rate Profiling (GERP) scores, and Gene Ontology (GO) annotations, were obtained from OMIM ( [11]; https://omim.org), Ensembl Genome Browser (Release 115; [12]; https://www.ensembl.org), HGNC ( [13]; https://www.genenames.org), and HGMD ( [14]; https://www.hgmd.cf.ac.uk). Single nucleotide polymorphisms (SNPs) were retrieved from dbSNP (Build 155; [15]). Variant rs identifiers were obtained from VarSome ( [16]; https://sso.varsome.com) and Franklin [17], and verified using Ensembl. Allele frequency data were supplemented using the Genome Aggregation Database (gnomAD; [18]; https://gnomad.broadinstitute.org/). Evolutionarily constrained elements identified in Ensembl (Release 115) across 92 mammalian species were used to assess evolutionary conservation of MMP20. MicroRNA (miRNA) binding sites within the 3′ untranslated region (3′UTR) of MMP20 were predicted using TargetScanHuman (Release 8.0; [19]; http://www.targetscan.org). CpG island prediction was performed using MethPrimer ( [20]; https://methprimer.com/) and EMBOSS CpGplot ( [21]; https://www.bioinformatics.nl/cgi-bin/emboss/help/cpgplot), applying the following criteria: CpG island length > 200 bp, GC content > 50%, and observed-to-expected CpG ratio > 0.6 within a 4500 bp upstream region of MMP20. Post-translational modification (PTM) data were obtained from dbPTM ( [22]; http://dbPTM.mbc.nctu.edu.tw), focusing on experimentally validated sites. Protein–protein interaction (PPI) networks were constructed using STRING (version 12.0; [23]; https://string-db.org) with a minimum interaction score of 0.7. The top 10 interacting proteins were visualized, and network significance was assessed using STRING PPI enrichment analysis (p < 0.05). Results and discussion The integrative analysis combined published studies and publicly available bioinformatic databases to provide a comprehensive overview of molecular insights into MMP20-related AI across multiple omics layers, including genomic, epigenomic, transcriptomic, proteomic, and phenomic data. The literature search covered publications from database inception to March 2026. Systematic literature search and multi-omics data organization The literature search identified 237 records (72 from PubMed and 165 from Web of Science). After removal of 51 duplicate records, 186 studies were screened by title and abstract. Following full-text assessment, 67 studies published between December 1997 and March 2026 met the inclusion criteria and were included in further analyses (Fig. 1).Fig. 1Flow diagram illustrating the study selection process. (*) Some studies included data corresponding to multiple omics levels. Abbreviations: AI, amelogenesis imperfecta. Flow diagram illustrating the study selection process. (*) Some studies included data corresponding to multiple omics levels. Abbreviations: AI, amelogenesis imperfecta. Because several publications addressed multiple omics approaches, some were assigned to more than one analytical category. Overall, the included literature comprised 25 genomics, one miRNomics, one transcriptomics, 17 regulomics, 29 proteomics, one interactomics, and 18 phenomics studies. Quality assessment of included studies Of the 67 included studies, 10 (14.9%) were classified as excellent quality, 33 (49.3%) as high quality, 17 (25.4%) as good quality, and 7 (10.4%) as moderate quality (Supplementary Table S1). Studies with higher methodological quality generally fulfilled a greater proportion of the predefined assessment criteria, particularly regarding study design, validation, reproducibility, and biological interpretation. The most frequent limitations were the lack of publicly available datasets and the absence of a discussion of study limitations. Multi-omics characterization and bioinformatic analysis The included studies were systematically integrated with data from 16 bioinformatic databases and analytical tools to characterize MMP20 across multiple molecular and regulatory layers. Genomics-DNA level and rare sequence variants The MMP20 gene, located on chromosome 11q22.2–q22.3, is catalogued under multiple identifiers, including ENSG00000111145 (Ensembl), 604629 (OMIM), 4313 (NCBI Gene), 7174 (HGNC), NM_004771.4 (RefSeq mRNA), and UniProt ID O60882. It is positioned within a cluster of matrix metalloproteinase (MMP) genes. Early studies mainly used candidate-gene approaches with PCR amplification and Sanger sequencing of coding exons and exon–intron boundaries to identify pathogenic variants [[24], [25], [26], [27], [28], [29], [30], [31], [32], [33]]. More recent studies applied next-generation sequencing (NGS), including targeted gene panels [34] and whole-exome sequencing (WES) [[35], [36], [37], [38], [39], [40], [41]], enabling broader molecular diagnostics. Hany et al. (2025) developed a custom smMIP panel targeting 19 AI-associated genes, while the GenoDENT multigene panel has also been used together with WES in AI cohorts [6]. MMP20 variants associated with AI have been reported across diverse populations, including European [30,33,40,42], Turkish [25,31,32,[35], [36], [37],39], Korean [27], Taiwanese [37,41], and cohorts from China, Pakistan, Oman, Costa Rica, and North America [37,38]. In several early studies, participant ethnicity was not explicitly reported [24,26,28,29], but the collective evidence supports a global contribution of MMP20 to AI. To date, more than 30 MMP20 variants associated with AI have been reported, predominantly missense variants, although nonsense, frameshift, and splice-site variants have also been identified (Supplementary Table S2). Despite this molecular heterogeneity, affected individuals consistently present with hypomaturation AI, and no clear genotype–phenotype correlation has been established. In some populations, haplotype analyses suggest founder effects [38]. To complement the published evidence, we performed bioinformatic analyses based on data retrieved from publicly available resources. Ensembl data revealed 12 conserved elements longer than 13 bp across 92 eutherian mammals, including primates, indicating strong evolutionary conservation of MMP20. These findings are consistent with comparative genomic analyses demonstrating that MMP20 has remained highly conserved across mammals with functional dentitions, reflecting strong evolutionary constraints on its role in enamel formation [43]. In addition, analysis of the MMP20 locus revealed several structural variants (SVs) (Fig. 2).Fig. 2Genomic organization of the chromosome 11q22.2 region showing the MMP20 gene and the co-localized lncRNA MMP20-AS1. Figure retrieved from the Ensembl Genome Browser. Genomic organization of the chromosome 11q22.2 region showing the MMP20 gene and the co-localized lncRNA MMP20-AS1. Figure retrieved from the Ensembl Genome Browser. The long non-coding RNA (lncRNAs) MMP20-AS1 (also known as RP11–817J15.2) overlaps the MMP20–201 transcript in antisense orientation (Fig. 2), suggesting a potential cis-regulatory role, as antisense lncRNAs commonly regulate neighboring or overlapping genes [44]. However, no experimental evidence currently supports a regulatory function for MMP20-AS1. GeneCards and MalaCards associate MMP20-AS1 with hypomaturation AI type IIA2; however, it remains unclear whether this reflects an independent pathogenic role or simply its genomic overlap with MMP20. To further characterize the reported MMP20 variants, we performed independent in silico variant annotation and functional consequence prediction using the Ensembl Genome Browser and Variant Effect Predictor (VEP). Most variants were located within regions encoding the four functional domains of MMP20 and overlapped the antisense transcript MMP20-AS1 (Supplementary Table S3; Fig. 3). Of the 37 reported variants (canonical transcript), 21 were missense, 6 stop-gained, 4 frameshift, 5 splice-site variants, and a synonymous variant (c.103 A>C). Overall, 29 variants overlapped MMP20-AS1, including 18 missense, 4 stop-gained, 4 frameshift, and 2 splice-site variants. Regarding clinical significance, 7 variants were classified as pathogenic, 2 as likely pathogenic, 5 as variants of uncertain significance, and 2 showed conflicting pathogenicity classifications, whereas the remaining variants lacked available clinical annotation (Supplementary Table S3).Fig. 3Structure of MMP20, MMP20-AS1, and localization of reported pathogenic variants. Schematic representation of the MMP20 gene and the overlapping antisense transcript MMP20-AS1. Exons are shown as numbered boxes, introns as connecting lines with lengths (bp) indicated above, and shaded regions denote UTRs. Amino acid ranges encoded by individual MMP20 exons and the corresponding protein domains (signal peptide, propeptide, catalytic domain, linker region, and hemopexin-like domain) are shown below. Dashed lines indicate exon–domain correspondence. Reported pathogenic variants are shown in red and numbered according to Supplementary Tables S2 and S3; most localize to the zinc-dependent catalytic domain. Abbreviations: Sig., signal peptide; Lin., linker region. Structure of MMP20, MMP20-AS1, and localization of reported pathogenic variants. Schematic representation of the MMP20 gene and the overlapping antisense transcript MMP20-AS1. Exons are shown as numbered boxes, introns as connecting lines with lengths (bp) indicated above, and shaded regions denote UTRs. Amino acid ranges encoded by individual MMP20 exons and the corresponding protein domains (signal peptide, propeptide, catalytic domain, linker region, and hemopexin-like domain) are shown below. Dashed lines indicate exon–domain correspondence. Reported pathogenic variants are shown in red and numbered according to Supplementary Tables S2 and S3; most localize to the zinc-dependent catalytic domain. Abbreviations: Sig., signal peptide; Lin., linker region. Functional studies in animal models further support the essential role of MMP20 during enamel development, as Mmp20 knockout mice exhibit severe enamel defects [45,46]. Comparative genomic analyses additionally identified species-specific amino acid substitutions and accelerated evolution of MMP20 in the rough-toothed dolphin, suggesting a role in adaptive enamel morphology [47]. Epigenomics DNA methylation No studies specifically investigating MMP20 DNA methylation in AI were identified. In this study, analysis of approximately 4.5 kb upstream of the MMP20 transcription start site using MethPrimer and EMBOSS CpGplot did not detect CpG islands within the promoter region (Supplementary Figure S1), although epigenetically modified accessible regions (EMARs) were identified (Fig. 2). Currently, there is no direct evidence that MMP20 expression is regulated by promoter CpG methylation. However, broader studies of matrix metalloproteinases suggest that epigenetic mechanisms, including DNA methylation, histone modifications, and non-coding RNAs, may contribute to metalloproteinase regulation in specific biological contexts [48]. Non-coding RNA gene regulation A literature review identified one miRNA-related study relevant to MMP20. Yoshioka et al. [98] showed that overexpression of hsa-miR-1306–5p, hsa-miR-3195, and hsa-miR-3914 impairs ameloblast differentiation and mineralization while downregulating several AI-associated genes, including MMP20. Notably, hsa-miR-3195 reduced MMP20 expression at later stages of differentiation, suggesting stage-dependent regulation. However, direct miRNA–MMP20 interactions have not been experimentally validated. To complement the published evidence, we performed in silico miRNA target prediction using TargetScanHuman. This analysis identified 173 putative miRNA binding sites within the 3′ untranslated region (3′UTR) of MMP20, all with low evolutionary conservation. None corresponded to the miRNAs reported by Yoshioka et al. [98] (Supplementary Table S4). Transcriptomics MMP20 expression is highly tissue- and stage-specific, occurring mainly in ameloblasts, pre-ameloblasts, and odontoblasts during the secretory and late transition stages of amelogenesis [49,50]. Ectopic expression has also been reported in odontogenic tumors and tongue-derived carcinoma cell lines [50]. A literature review identified one study specifically investigating transcript-level alterations of MMP20 [36]. The study showed that pathogenic variants can disrupt exon recognition during pre-mRNA splicing, leading to aberrant transcripts, reduced levels of correctly spliced mRNA, and decreased production of functional MMP20 protein, thereby contributing to hypomaturation AI [36]. Regulomics Multiple upstream signaling pathways regulate MMP20 expression during amelogenesis. The regulatory landscape of MMP20 has been investigated in 17 studies [[51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67]]. Early in vitro studies demonstrated that TGF-β1 signaling through TGFBR1 promotes MMP20 expression in ameloblasts, highlighting the role of extracellular growth factor signaling in regulating enamel matrix protein processing [51]. Subsequent studies further showed that autocrine TGF-β1 signaling coordinately regulates MMP20 together with other enamel matrix genes [52]. In humans, pathogenic ITGB6 variants have been associated with reduced MMP20 expression and impaired enamel matrix degradation, suggesting that altered ITGB6-dependent signaling contributes to defective enamel mineralization, potentially through RUNX2-mediated regulatory mechanisms [53]. Similarly, transcriptomic analyses of dental pulp tissues from individuals with enamel–renal syndrome identified dysregulation of enamel-related genes, including MMP20, associated with pathogenic FAM20A variants [54]. Animal studies have established BMP–Smad signaling as a major upstream regulator of Mmp20 during amelogenesis. Disruption of this pathway following conditional deletion of Bmp2/Bmp4 or Fam20c suppresses Mmp20 expression and results in defective enamel mineralization with AI-like phenotypes [[55], [56], [57]]. Consistent with these findings, conditional deletion of Runx2 in mouse ameloblasts likewise impairs enamel maturation by significantly reducing Mmp20 expression, identifying RUNX2 as a key downstream effector of BMP signaling and an upstream transcriptional regulator of MMP20 [58]. Furthermore, ODAM cooperates with RUNX2 to enhance MMP20 promoter activity and transcription, highlighting the importance of RUNX2-dependent regulatory networks during enamel mineralization [59]. Mmp20 expression is also modulated by additional intracellular signaling pathways, including Stim1-mediated calcium signaling [60], whereas Fam83h knockout mice exhibit significantly reduced Mmp20 expression [61]. Beyond its role in extracellular matrix proteolysis, MMP20 also contributes to enamel morphogenesis by regulating ameloblast adhesion and movement through cadherin cleavage [62]. Dysregulated Mmp20 expression disrupts ameloblast polarity [63], activates Wnt/β-catenin signaling and promotes aberrant ameloblast invasion into the developing enamel layer [64], and alters the expression of genes involved in cytoskeletal organization, cell adhesion, and directional ameloblast migration [65,66]. More recently, organoid studies have identified MMP20 as a downstream target of the Notch–DLX3 signaling axis during ameloblast maturation [67]. To complement the published evidence, we examined the regulatory landscape of the MMP20 locus using publicly available genomic resources. Data retrieved from Ensembl identified overlapping regulatory elements, including enhancer regions, CTCF (CCCTC-binding factor)-binding sites, and epigenetically modified accessible regions (EMARs) (Fig. 2 and Fig. 4), suggesting higher-order chromatin organization and potential long-range regulation of MMP20 expression.Fig. 4Regulatory landscape of the human MMP20 gene based on integrated multi-omics data. The canonical MMP20 transcript consists of ten exons encoding the enamel-specific metalloproteinase MMP20. The coding region spans exons 1–10 and is flanked by 5′ and 3′ untranslated regions (UTRs). The MMP20 locus partially overlaps the antisense transcript MMP20-AS1, indicating a bidirectional genomic organization. Two red triangles mark the overlap of the recurrent variant c.389 C>T (p.Thr130Ile) with a PTM site (amino acid 130). Regulatory features include enhancer regions (orange), epigenetically modified accessible regions (EMARs; dark green), and CTCF-binding sites (turquoise). Solid black lines indicate experimentally validated features, whereas dashed lines denote predicted regulatory interactions or mechanisms inferred from animal models and in vitro studies. A schematic also includes a proposed ITGB6/TGF-β signaling pathway potentially involved in MMP20 regulation. This pathway represents a biologically plausible regulatory model integrating evidence from human genetic studies, murine models, and in vitro experiments. In this model, activation of TGF-β signaling through TβRII/TβRI receptor complex induces phosphorylation of SMAD2/3, followed by complex formation with SMAD4, nuclear translocation, and downstream transcriptional regulation involving RUNX2. Although individual components of this pathway have been experimentally validated, the complete ITGB6–TGF-β–SMAD–RUNX2 signaling cascade has not yet been directly demonstrated in human ameloblasts. Abbreviations: PTMs, post-translational modifications; MTIs, miRNA–target interactions. Gene structures were adapted from the Ensembl, and published literature on proposed human ITGB6/TGF-β signaling mechanisms [[53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67]]. Regulatory landscape of the human MMP20 gene based on integrated multi-omics data. The canonical MMP20 transcript consists of ten exons encoding the enamel-specific metalloproteinase MMP20. The coding region spans exons 1–10 and is flanked by 5′ and 3′ untranslated regions (UTRs). The MMP20 locus partially overlaps the antisense transcript MMP20-AS1, indicating a bidirectional genomic organization. Two red triangles mark the overlap of the recurrent variant c.389 C>T (p.Thr130Ile) with a PTM site (amino acid 130). Regulatory features include enhancer regions (orange), epigenetically modified accessible regions (EMARs; dark green), and CTCF-binding sites (turquoise). Solid black lines indicate experimentally validated features, whereas dashed lines denote predicted regulatory interactions or mechanisms inferred from animal models and in vitro studies. A schematic also includes a proposed ITGB6/TGF-β signaling pathway potentially involved in MMP20 regulation. This pathway represents a biologically plausible regulatory model integrating evidence from human genetic studies, murine models, and in vitro experiments. In this model, activation of TGF-β signaling through TβRII/TβRI receptor complex induces phosphorylation of SMAD2/3, followed by complex formation with SMAD4, nuclear translocation, and downstream transcriptional regulation involving RUNX2. Although individual components of this pathway have been experimentally validated, the complete ITGB6–TGF-β–SMAD–RUNX2 signaling cascade has not yet been directly demonstrated in human ameloblasts. Abbreviations: PTMs, post-translational modifications; MTIs, miRNA–target interactions. Gene structures were adapted from the Ensembl, and published literature on proposed human ITGB6/TGF-β signaling mechanisms [[53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67]]. Collectively, the published evidence together with the regulatory features identified in this study support a biologically plausible ITGB6–TGF-β–SMAD–RUNX2 signaling model. This conceptual model integrates evidence from human genetic studies, murine models, in vitro experiments, and bioinformatic analyses. However, the complete signaling cascade has not yet been directly demonstrated in human ameloblasts (Fig. 4). Proteomics A literature review identified 29 proteomic studies investigating the structure, biochemical activity, and functional role of MMP20 [25,27,32,49,62,[68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83], [84], [85], [86], [87], [88], [89], [90], [91]]. MMP20 is a zinc-dependent metalloproteinase with the conserved domain architecture characteristic of the MMP family, including a catalytic domain with a zinc-binding active site [68,69]. The human MMP20 protein comprises 483 amino acids and is organized into a signal peptide (Met1–Ala22), a prodomain (Ala23–Asn107), a catalytic domain (Tyr108–Gly271), a linker region (Pro272–Leu295), and a C-terminal hemopexin domain (Cys296–Cys483) [68]. Comparative analyses demonstrate high amino acid conservation across species, supporting its functional importance [69]. As a key component of the enamel proteome, MMP20 initiates the proteolytic processing of enamel matrix proteins during the secretory stage of amelogenesis, thereby facilitating proper enamel matrix organization and subsequent mineralization [70]. Its major substrates include amelogenin [71,72] and ameloblastin [73]. Subsequent structural and functional studies further characterized cleavage sites within enamel matrix proteins and demonstrated that MMP20-mediated processing prevents protein occlusion within developing mineral crystals, thereby facilitating normal enamel prism formation [[74], [75], [76]]. In addition to extracellular matrix proteolysis, MMP20 regulates ameloblast adhesion through cadherin cleavage and associated β-catenin signaling, thereby contributing to ameloblast differentiation and enamel morphogenesis [62,[77], [78], [79]]. It also contributes to the regulation of hydroxyapatite crystal growth and morphology during enamel biomineralization [80,81]. Proteomic studies and comprehensive reviews indicate complementary, stage-specific functions of MMP20 and KLK4, with MMP20 acting predominantly during the secretory stage to initiate enamel matrix protein processing, whereas KLK4 functions primarily during the maturation stage to promote the degradation the removal of residual matrix proteins, thereby enabling complete enamel mineralization [[82], [83], [84], [85]]. The essential role of MMP20 during the secretory stage is further demonstrated by severe enamel defects observed in Mmp20-null mice, including impaired matrix processing, disrupted prism architecture, reduced mechanical strength, rough enamel surfaces, and compromised dentino-enamel junction integrity [88, 89, 90, 91). Although Mmp20-processed amelogenin is sufficient to restore enamel rod decussation, complete enamel development requires coordinated processing of additional enamel matrix proteins, KLK4 activity during maturation, and normal ameloblast function [90]. Moreover, preserved MMP20 activity in an ADAM10 (A Disintegrin and Metalloproteinase 10)-deficient mouse model indicates that ADAM10 regulates enamel development through mechanisms distinct from MMP20-mediated matrix proteolysis [91]. Human genetic studies show that variants affecting the catalytic domain markedly reduce enzymatic activity, whereas variants in the hemopexin domain impair substrate recognition and protein stability [25,27,32]. Together, these findings establish a direct link between MMP20 dysfunction and hypomaturation AI. Post-translational modifications (PTMs) To investigate post-translational modifications of MMP20, we retrieved PTM annotation data from the dbPTM database. Eleven PTM sites were identified, including phosphorylation sites at amino acid positions 121, 123, 124, 127, 129, 130, 132, 134, and 135, and acetylation sites at positions 128 and 140. Integration of PTM annotation with reported MMP20 variants identified a single overlap between a AI-associated variant and a PTM site. The recurrent missense variant c.389 C>T (p.Thr130Ile) coincided with a predicted phosphorylation site at amino acid position 130 (Fig. 5).Fig. 5Missense variant c.389 C>T (p.Thr130Ile) in MMP20 located at a predicted phosphorylation site. Data and figure adapted from the Ensembl Genome Browser (lower pannel) and the dbPTM database (upper pannel). Missense variant c.389 C>T (p.Thr130Ile) in MMP20 located at a predicted phosphorylation site. Data and figure adapted from the Ensembl Genome Browser (lower pannel) and the dbPTM database (upper pannel). Interactomics The literature review identified one interactomic study involving MMP20 [71]. At the interaction level, MMP20 exhibits selective binding to enamel matrix proteins, particularly amelogenin. Tanimoto et al. [99] demonstrated that an AI-associated amelogenin variant (P41T) reduces amelogenin–MMP20 binding, potentially impairing enamel matrix processing and maturation. To complement the published evidence, we explored protein–protein interactions (PPIs) involving MMP20 using STRING. The resulting network comprised 11 nodes and showed significantly more interactions than expected by chance (p = 4.83 × 10⁻¹¹) (Fig. 6).Fig. 6Protein–protein interaction (PPI) network of MMP20, generated using the STRING database. The network includes both experimentally validated and predicted interactions, encompassing direct (physical) and indirect (functional) associations. Nodes represent proteins, and edges indicate interactions supported by curated databases, experimental evidence, or computational predictions. Protein–protein interaction (PPI) network of MMP20, generated using the STRING database. The network includes both experimentally validated and predicted interactions, encompassing direct (physical) and indirect (functional) associations. Nodes represent proteins, and edges indicate interactions supported by curated databases, experimental evidence, or computational predictions. MMP20 showed strong interactions with the major enamel proteins AMELX, AMBN, ENAM, as well as with the protease KLK4, supporting its central role in enamel formation and maturation. Additional interactions with ODAM, AMTN, ODAPH, and WDR72 suggest broader involvement in enamel development. In contrast, HPX appeared as an isolated node without supported functional associations within the network. To place MMP20 within a broader molecular context, we expanded the interactome by including 17 genes previously associated with non-syndromic AI (ENAM, AMELX, ARHGAP6, AMBN, KLK4, FAM83H, WDR72, ODAPH, SLC24A4, GPR68, FAM20A, LAMB3, COL17A1, ACP4, ITGB6, AMTN, and DLX3) [92]. First-shell interactors were subsequently added using the STRING “More” option. The resulting 23-protein network showed significant interaction enrichment (p < 1.0 × 10⁻¹⁶), with all but three proteins forming a connected cluster and MMP20 occupying a central position (Fig. 7). Additional interactors, including TUFT1, DSPP, FAM20C, ODAM, and LAMC2, participate in enamel or dentin development, further supporting the integration of MMP20 within a broader amelogenesis-associated protein interaction network.Fig. 7STRING-derived protein–protein interaction (PPI) network of MMP20 and 17 genes associated with non-syndromic amelogenesis imperfecta (AI). Additional first-shell interacting proteins retrieved using the STRING “More” option (TUFT1, DSPP, FAM20C, ODAM and LAMC2) were included to provide a broader molecular context. Nodes represent proteins, and edges indicate known or predicted functional associations supported by curated databases, experimental evidence, or computational predictions. STRING-derived protein–protein interaction (PPI) network of MMP20 and 17 genes associated with non-syndromic amelogenesis imperfecta (AI). Additional first-shell interacting proteins retrieved using the STRING “More” option (TUFT1, DSPP, FAM20C, ODAM and LAMC2) were included to provide a broader molecular context. Nodes represent proteins, and edges indicate known or predicted functional associations supported by curated databases, experimental evidence, or computational predictions. These findings are consistent with emerging evidence that AI-associated genes function within interconnected molecular and regulatory networks governing amelogenesis. For example, a KDF1 missense variant altered Hippo–YAP signaling and MMP20 expression, resulting in enamel defects [93]. Phenomics The literature review identified 18 studies investigating MMP20-associated enamel phenotypes and genotype–phenotype correlations [24,25,27,[30], [31], [32], [33],36,37,39,41,42,45,46,65,89,94,95]. Collectively, these studies demonstrate that biallelic variants in MMP20 cause autosomal recessive hypomaturation AI type IIA2. Affected individuals are typically homozygous or compound heterozygous, whereas heterozygous carriers are clinically unaffected, supporting haplosufficiency [24,25,28]. Clinically, AI type IIA2 is characterized by enamel of normal thickness but reduced mineralization and hardness, resulting in post-eruptive breakdown [24,25,27,30,33,36]. Enamel discoloration ranges from white or yellow to brown and often darkens with age [27,33]. Radiographically, affected teeth typically show reduced enamel–dentin contrast, which may range from mildly decreased to nearly absent [27,31,33]. Additional findings include anterior open bite, rough enamel surfaces, mild tooth-shape anomalies, increased dental sensitivity, and, less frequently, root abnormalities, delayed eruption, and pulp chamber alterations [24,25,30,33,37]. Phenotypic severity has been associated with missense variants affecting the catalytic and hemopexin domains of MMP20 [32,33]. Nevertheless, considerable phenotypic variability exists, likely reflecting the influence of genetic modifiers [94]. Recent evidence suggests that the phenotypic spectrum of MMP20-related AI may also include hypoplastic–hypomaturation presentations, thereby further expanding the recognized clinical variability associated with MMP20 deficiency [41]. This concept is supported by reports of digenic inheritance involving MMP20 together with LAMA3 or COL17A1 [39,42]. Animal models further support these observations. Mmp20-null mice exhibit severe enamel hypomineralization, reduced hardness, post-eruptive wear, and disrupted enamel rod organization, whereas heterozygous mice display a normal phenotype [45,46,95]. Moreover, altered Mmp20 expression disrupts coordinated ameloblast movement and enamel patterning [65], while combined heterozygous variants in Mmp20 and Klk4 produce an enamel phenotype not observed in single heterozygotes, providing additional evidence for digenic effects during enamel development [89]. Limitations of the study This study is subject to limitations related to the heterogeneity of published data in terms of study design, sample size, and consistency of reporting. In addition, a substantial proportion of evidence derives from in vitro and animal studies, which may limit direct clinical translation. Inconsistent annotation of genetic variants across studies may further affect comparability. Some of the bioinformatic predictions presented here have not been experimentally validated and should therefore be interpreted with caution. Furthermore, the interpretation of MMP20 regulatory mechanisms is constrained by the reliance on murine models, as direct experimental evidence in human ameloblasts remains limited. Future direction of the study Although substantial progress has been made in understanding MMP20-related AI in humans and animal models, several questions remain unresolved. Future studies should focus on strengthening genotype–phenotype correlations and clarifying the contribution of non-coding regulatory mechanisms, including the potential roles of the antisense transcript MMP20-AS1 and miRNA-mediated regulation of MMP20 expression. Particular emphasis should be placed on determining how these regulatory mechanisms contribute to the stage-specific regulation of MMP20 during the secretory and maturation phases of amelogenesis. Given that MMP20 expression peaks during the secretory stage and declines during the transition to maturation, epigenetic regulation and miRNA-mediated post-transcriptional mechanisms may contribute to its temporal regulation. However, these predicted regulatory mechanisms require experimental validation using luciferase reporter assays, qRT-PCR, western blotting, RNA sequencing of stage-specific ameloblasts or ameloblast organoids, and chromatin accessibility assays (e.g., ATAC-seq) to confirm the functional relevance of the predicted miRNA binding sites and regulatory elements identified in this study. Furthermore, the identification of the recurrent p.Thr130Ile variant at a phosphorylation site highlights the need for functional studies investigating whether disease-associated variants can disrupt PTMs and thereby alter MMP20 function. Integration of emerging multi-omics data, expansion of molecular databases, and inclusion of MMP20 in comprehensive AI gene panels may improve diagnostic accuracy and support future research on MMP20-related AI. Ultimately, a better understanding of MMP20-associated molecular networks could contribute to more precise diagnostics, personalized clinical management, and the development of targeted therapeutic approaches for patients with AI. In the longer term, these insights may also support the development of biomimetic approaches for mineralized tissue regeneration and enamel repair [96,97]. Conclusions By integrating published evidence with data retrieved from 16 bioinformatic databases and tools with evidence across nine omics layers, this study provides a comprehensive overview of MMP20-related AI. The integration of published findings with bioinformatic analyses improves variant interpretation and data comparability. Interaction network analysis involving MMP20 and 17 additional AI-associated proteins further highlighted its central role within the molecular network governing amelogenesis. Rather than simply compiling database annotations, this study integrates dispersed evidence across multiple molecular layers into a unified framework linking genetic variation with regulatory mechanisms, protein function, interaction networks, and clinical phenotypes. Several regulatory features presented here represent in silico predictions and should therefore be regarded as candidate mechanisms requiring experimental validation. This integrative approach identifies current knowledge gaps, provides a structured framework for future investigation of MMP20-related AI, and highlights priorities for experimental validation. Beyond summarizing current knowledge, this work proposes a conceptual framework for the analysis and reporting of molecular data in other AI-associated genes. Standardized annotations, consistent transcript references, and unified reporting criteria may improve the reliability of future studies and support more accurate diagnosis and individualized management of AI. Declaration of Generative AI and AI-assisted technologies in the writing process During the preparation of this manuscript, the authors used ChatGPT (OpenAI) to assist with language editing, improve readability, and refine the presentation of the text. The authors reviewed, revised, and verified all generated content and take full responsibility for the accuracy, integrity, and originality of the manuscript. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.