DNA methylation changes following rapid maxillary expansion in children

Abstract

Rapid maxillary expansion (RME) is a widely used orthodontic intervention for correcting transverse maxillary deficiencies. By mechanically separating the midpalatal suture, RME improves occlusal relationships and induces tissue repair processes similar to fracture healing. Inflammation and bone regeneration following fracture are regulated by epigenetic mechanisms, such as DNA methylation. However, the effects of RME-induced mechanical stimulation on epigenetic dynamics in humans remain unclear. We evaluated genome-wide DNA methylation changes in saliva associated with RME treatment. Twenty patients undergoing orthodontic treatment with RME and four untreated controls were enrolled. Saliva samples were collected before (T0) and after treatment (T1), and genome-wide DNA methylation profiling was performed using the Illumina Infinium HumanMethylationEPIC v2.0 BeadChip. Differential methylation analysis was conducted using linear models accounting for within-subject correlations. In the RME group, 164 CpG sites were differentially methylated between T0 and T1 (false discovery rate < 0.05), whereas no significant changes were observed in controls. The differentially methylated CpG sites were annotated to genes associated with biological responses to RME, including inflammatory pathways and metabolic processes. RME may be associated with measurable changes in salivary DNA methylation, suggesting epigenetic responses detectable in saliva and providing a basis for future mechanistic investigations.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-66231-8.

Introduction

Maxillary arch constriction is a form of malocclusion observed from the primary through permanent dentition. It is influenced by environmental and genetic factors, including oral habits, perioral muscular dysfunction, and respiratory disorders^1–4. This condition is frequently associated with clinical features, such as dental crowding, crossbite, a high-arched palate, perioral muscular imbalance, and airway constriction^5–7. Furthermore, maxillary arch constriction may lead to functionally induced mandibular deviation, which contributes to temporomandibular joint symptoms, including pain, joint sounds, and restricted mouth opening⁸.

Rapid maxillary expansion (RME) is commonly used in orthodontic treatment to correct maxillary arch constriction^2,5,9. By applying intermittent orthopedic forces to the maxillary arch and inducing separation of the midpalatal suture, RME increases the transverse width of the dental arch, thereby facilitating the establishment of an appropriate occlusal relationship and improving functional and esthetic conditions, such as reduced masticatory efficiency and dental crowding^10–12. Following RME-induced separation of the midpalatal suture, bone healing progresses through biological processes similar to those observed in fracture healing¹³. Mechanical forces generated during RME are transmitted not only to the midpalatal suture but also to the surrounding craniofacial sutures and periodontal tissues, where they are sensed by local cells through mechanotransduction pathways, leading to inflammatory responses, extracellular matrix remodeling, osteoblast and osteoclast differentiation, and subsequent bone formation^14–16. To assess this healing process, occlusal radiography and cone-beam computed tomography (CBCT) were performed at different stages of treatment to evaluate suture separation, subsequent bone formation, and three-dimensional changes in palatal morphology^17–19.

Bone formation is a multistep biological process involving diverse cell types, including inflammatory cells, mesenchymal stem cells, osteoblasts, osteoclasts, chondrocytes, and endothelial cells, as well as growth factors and cytokines^20,21. Recently, epigenetic regulation, particularly DNA methylation, has been implicated in the regulation of bone metabolism–related genes (RUNX2, SP7, BMP2, SOST, ALPL, BGLAP, FZD1, TNFSF11, TNFRSF11B, LOX, ESR1, and DLX5), which govern cellular functions essential for bone formation^22–24. DNA methylation is a key epigenetic mechanism that regulates gene expression and contributes to disease-related processes. It is dynamically modified in response to genetic background and environmental factors, resulting in long-term alterations in cellular and tissue phenotypes^23,25. However, DNA methylation dynamics associated with the healing process following RME in orthodontic treatment remain poorly characterized^23,26. Elucidating DNA methylation changes before and after treatment is essential not only for advancing our understanding of epigenetic mechanisms underlying biological responses to orthodontic therapy, but also for providing novel insights into epigenetically mediated therapeutic interventions and treatment-related biological adaptations through alterations in DNA methylation of target genes^27,28. Moreover, only a limited number of studies in dentistry, including orthodontics, have examined changes in DNA methylation dynamics before and after treatment^23,26–28.

Therefore, the aim of this study was to evaluate the effects of phase I orthodontic treatment with RME during the mixed dentition stage on DNA methylation patterns. Using saliva-derived DNA and an array-based platform, we assessed genome-wide changes in CpG site-specific DNA methylation before and after treatment.

Results

DNA methylation changes in patients treated with RME

Of the 895,978 CpG sites that passed quality control and normalization, no significantly differentially methylated CpG sites were detected between T0 and T1 in the untreated control group. In contrast, 164 CpG sites in the RME-treated group exhibited significant differential methylation (FDR < 0.05). The mean absolute change in DNA methylation (|Δβ|) across these 164 CpG sites was 0.0108 (standard deviation = 0.0122), indicating that the overall magnitude of methylation changes was modest. Among these 164 CpG sites, 73 (44.5%) showed increased methylation, whereas 91 (55.5%) showed decreased methylation, demonstrating bidirectional methylation changes following RME intervention.

Volcano plot analysis indicated that significant CpG sites were distributed in both hypermethylated and hypomethylated directions. Manhattan plot visualization showed that these CpG sites were widely distributed across multiple chromosomes, with no apparent enrichment on any specific chromosome (Fig. 1). Annotation analysis revealed that the 164 CpG sites corresponded to 128 genes (Supplementary Table S1). Notably, these genes included those involved in bone metabolism (LOX, PPARD, WNT10A), cell proliferation (APC, DDIT4, FBXO5, AURKAIP1), and immune and inflammatory responses (CARD19, CYLD, TRAF4, NOTCH4), consistent with the known biological effects of RME. To further evaluate the robustness of these findings, additional analyses were performed. First, to examine whether salivary cell composition changed during treatment, the estimated proportions of epithelial and immune cells were compared between T0 and T1 in the RME group. No significant differences were observed in the estimated proportions of epithelial cells (0.258 ± 0.053 vs. 0.235 ± 0.078, P = 0.163) or immune cells (0.721 ± 0.054 vs. 0.745 ± 0.077, P = 0.145). Second, using progressively stricter absolute Δβ thresholds, 55, 21, and 1 of the 164 CpG sites remained significant after applying |Δβ| thresholds of > 0.01, > 0.02, and > 0.05, respectively (Supplementary Table S2). Third, the differential methylation analysis was repeated using normalized M values without prior batch correction, with slide position included as a covariate in the linear model. Three CpG sites remained significant after FDR correction, corresponding to the three most statistically significant CpG sites identified in the primary analysis (Supplementary Table S3).

Fig. 1.

Visualization of CpG methylation changes in patients treated with rapid maxillary expansion (RME) using volcano and Manhattan plots. (a) Volcano plot showing genome-wide changes in DNA methylation changes at CpG sites between T0 and T1. The x-axis represents ΔM values, and the y-axis represents − log₁₀(false discovery rate [FDR]). CpG sites with FDR < 0.05 are highlighted in red. The horizontal dashed lines indicate the FDR significance thresholds. (b) Manhattan plot showing − log₁₀(FDR) values of CpG sites across all chromosomes following RME treatment. Each point represents a single CpG site. CpG sites with FDR < 0.05 are highlighted in red, and the horizontal dashed line indicates the significance threshold.

Interaction effects of concomitant appliance use

Interaction models were applied to evaluate the influence of clinical background factors on DNA methylation changes between T0 and T1. Among the four factors examined, only concomitant appliance use showed a significant interaction effect, with 13 CpG sites meeting an FDR threshold of < 0.05. Notably, these CpG sites did not overlap with the 164 CpG sites identified in the main-effect analysis. Among the 13 CpG sites with significant interaction effects, four loci with relatively large effect sizes and annotated gene symbols (AMACR, FOXF1, ZNF200, and HCK) were selected as representative examples (Fig. 2). At these loci, the Δβ values between T0 and T1 were smaller in patients receiving concomitant appliance therapy than in the RME-only group, indicating a reduced magnitude of DNA methylation change.

Fig. 2.

Changes in M values between T0 and T1 and distributions of Δβ according to concomitant appliance use. (a) cg03909417: AMACR (b) cg07769121: FOXF1 (c) cg09848405: ZNF200 (d) cg20547606: HCK. The left panels show longitudinal changes in M values at T0 and T1, stratified by concomitant appliance use (yes/no). The y-axis represents M values, and the x-axis represents the time points (T0 and T1). Each line connects paired measurements from the same individual. The right panels display box plots of Δβ (T1 − T0) stratified by concomitant appliance use. The y-axis represents Δβ values, and the x-axis indicates the presence or absence of concomitant appliance use.

Pathway enrichment analysis

KEGG pathway over-representation analysis was conducted for the 128 genes corresponding to 164 significant CpG sites. Although no pathways reached statistical significance after multiple-testing correction, the top 20 pathways ranked by FDR were extracted (Table 1). KEGG pathway maps of the three top-ranked pathways are presented (Supplementary Figure S1). The three pathways with the lowest FDR values were Autophagy – animal (p = 4.7 × 10^–4, FDR = 5.9 × 10^–2), Oxidative phosphorylation (p = 1.5 × 10^–3, FDR = 9.4 × 10^–2), and Cell cycle (p = 3.7 × 10^–3, FDR = 1.6 × 10^–1).

Table 1.

KEGG pathway analysis using missMethyl gometh (top 20 pathways ranked by FDR).

Description	Size	Count	P value	FDR
Autophagy—animal	157	84	4.7 × 10⁻4	5.9 × 10⁻2
Oxidative phosphorylation	118	53	1.5 × 10⁻3	9.4 × 10⁻2
Cell cycle	152	74	3.7 × 10⁻3	1.6 × 10⁻1
Cellular senescence	153	76	7.4 × 10⁻3	1.9 × 10⁻1
Ubiquitin mediated proteolysis	134	66	7.5 × 10⁻3	1.9 × 10⁻1
Nucleocytoplasmic transport	100	48	9.6 × 10⁻3	2.0 × 10⁻1
ATP-dependent chromatin remodeling	108	51	1.2 × 10⁻2	2.1 × 10⁻1
Chemical carcinogenesis—reactive oxygen species	196	84	1.4 × 10⁻2	2.1 × 10⁻1
Parkinson disease	245	102	1.5 × 10⁻2	2.1 × 10⁻1
AMPK signaling pathway	118	61	1.7 × 10⁻2	2.2 × 10⁻1
Insulin signaling pathway	131	63	4.2 × 10⁻2	4.1 × 10⁻1
Oocyte meiosis	117	54	4.4 × 10⁻2	4.1 × 10⁻1
Gastric cancer	147	70	5.3 × 10⁻2	4.1 × 10⁻1
Shigellosis	236	102	5.4 × 10⁻2	4.1 × 10⁻1
Diabetic cardiomyopathy	179	75	5.5 × 10⁻2	4.1 × 10⁻1
Ribosome	162	55	6.0 × 10⁻2	4.1 × 10⁻1
Neurotrophin signaling pathway	114	56	6.3 × 10⁻2	4.1 × 10⁻1
Huntington disease	285	115	6.5 × 10⁻2	4.1 × 10⁻1
Amyotrophic lateral sclerosis	341	134	6.5 × 10⁻2	4.1 × 10⁻1
MAPK signaling pathway	286	132	6.6 × 10⁻2	4.1 × 10⁻1

FDR, false discovery rate; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Discussion

RME imposes mechanical loading on the midpalatal suture, thereby inducing a cascade of tissue responses ranging from inflammatory reactions to bone remodeling^13–16. However, the extent to which mechanical stimulation associated with orthodontic treatment is reflected at the epigenomic level in humans remains unclear. To our knowledge, this study is the first to comprehensively evaluate epigenomic changes associated with orthodontic treatment using genome-wide DNA methylation analysis of human saliva samples. Although statistically significant DNA methylation changes were identified following RME treatment, the overall magnitude of these changes was modest (mean |Δβ|= 0.0108). Because these effect sizes approach the technical noise floor of the EPIC array platform, the observed DNA methylation changes should be interpreted cautiously despite their statistical significance²⁹. The robustness of the primary findings was nevertheless supported by additional sensitivity analyses, including progressively stricter Δβ thresholds, an alternative batch-adjustment strategy, and evaluation of salivary cell composition. Therefore, the present findings should be regarded as exploratory rather than definitive evidence of functional epigenetic regulation. Despite these limitations, significant DNA methylation changes were observed at multiple CpG sites, suggesting that RME may be associated with detectable epigenetic alterations in saliva. The presence of both hypermethylated and hypomethylated CpG sites suggests that RME is associated with bidirectional epigenetic regulation rather than uniform activation or suppression of DNA methylation. Such bidirectional changes are biologically plausible because tissue remodeling is regulated by the coordinated activity of multiple biological pathways involved in inflammation, extracellular matrix remodeling, cell proliferation, and bone metabolism³⁰. Although the functional consequences of individual CpG alterations remain uncertain, these findings suggest that RME may induce a dynamic epigenetic response that contributes to tissue adaptation during orthodontic treatment. However, whether these modest DNA methylation changes are sufficient to alter gene expression remains unknown and should be investigated in future studies.

Many of the genes corresponding to the differentially methylated CpG sites identified in this study are involved in biological processes related to tissue remodeling and inflammatory responses, which are known to be involved in the biological response to RME¹⁶. The detection of these methylation changes in saliva suggests that biological responses associated with RME may be detectable beyond the local tissue environment. However, because salivary cells are not directly derived from bone tissue, the observed methylation changes may not exclusively represent osseous responses³¹. We additionally compared the estimated salivary epithelial and immune cell proportions before and after RME using EpiDISH. No significant differences were observed in the estimated proportions of epithelial cells (0.258 ± 0.053 vs. 0.235 ± 0.078, P = 0.163) or immune cells (0.721 ± 0.054 vs. 0.745 ± 0.077, P = 0.145). These findings suggest that the observed DNA methylation changes are unlikely to be solely explained by major changes in estimated salivary cell composition. Oral hygiene status, periodontal inflammation, and other oral inflammatory conditions were not clinically assessed in the present study. Therefore, although estimated salivary cell composition was adjusted for in the statistical models, residual confounding related to oral inflammatory conditions cannot be completely excluded. The observed DNA methylation changes may also be influenced by secondary factors associated with orthodontic treatment, such as sustained mechanical stress from appliance placement, pain, or alterations in daily oral function. Indeed, previous studies have reported associations between pain or psychological stress and epigenetic modifications, raising the possibility that the multifaceted biological burden accompanying RME contributes to the methylation variability detected in saliva^32–35. Therefore, the observed salivary DNA methylation changes should be interpreted as indirect surrogate markers reflecting integrated biological responses to orthodontic treatment rather than direct evidence of tissue-specific skeletal remodeling.

In the pathway enrichment analysis, pathways related to autophagy, oxidative phosphorylation, and cell cycle—key processes involved in cellular stress responses, energy metabolism, and regulation of cell proliferation—were ranked among the top candidates³⁶. Although none of these pathways reached statistical significance after multiple-testing correction, they are biologically plausible in the context of tissue remodeling associated with RME, including inflammatory responses and remodeling of the midpalatal suture¹⁶. In particular, mitochondrial metabolism and autophagy play critical roles in maintaining cellular homeostasis in response to mechanical stimulation, suggesting that these pathways may reflect general cellular stress responses associated with mechanical stimulation rather than RME-specific biological mechanisms^15,37. Nevertheless, because none of the pathways remained statistically significant after multiple-testing correction, these findings should be regarded as hypothesis-generating and require validation in future studies.

Interaction analysis examining how epigenomic responses to RME vary according to treatment conditions revealed multiple CpG sites showing significant associations between concomitant appliance use and DNA methylation changes from T0 and T1. Notably, the magnitude of these methylation changes was consistently smaller in patients receiving concomitant appliances than in those treated with RME alone. These findings suggest that alterations in the oral environment associated with different treatment conditions may modulate the transmission of mechanical forces induced by RME, which in turn could influence cellular stress responses and inflammatory sensitivity in bone and surrounding tissues¹⁵. These findings suggest that mechanical conditions during orthodontic treatment may influence cellular responses and epigenetic regulation, providing a novel perspective for understanding biological responses to orthodontic interventions beyond RME^15,38. Furthermore, the observation that the magnitude of applied orthodontic force and appliance selection may influence epigenetic changes has considerable clinical relevance.

In the present study, DNA methylation changes associated with RME were detectable in saliva, a noninvasive biological sample. Although saliva comprises a heterogeneous mixture of cell types, previous studies have demonstrated that saliva shares concordant DNA methylation patterns with blood at many CpG sites and is a reliable specimen for genome-wide DNA methylation analysis^39,40. Recent studies have further demonstrated the utility of salivary DNA methylation profiling across diverse fields, including systemic and psychiatric disorders, supporting the notion that saliva can capture systemic biological responses beyond the oral cavity^41,42. Taken together, these findings support the feasibility of saliva as a noninvasive biospecimen for detecting epigenetic responses associated with RME, although further validation in target tissues is warranted.

This study has some limitations. First, the overall sample size was relatively small, particularly in the untreated control group, limiting the statistical power to detect subtle temporal changes. Second, the untreated control group was small, slightly older than the treatment group, and consisted exclusively of female participants, limiting our ability to distinguish RME-specific methylation changes from normal developmental or temporal variation. Although sex was included as a covariate in the statistical models, residual confounding by sex cannot be completely excluded. Therefore, the observed methylation changes cannot be exclusively attributed to RME and should be interpreted in the context of potential temporal and developmental influences. In addition, because only three patients were treated with RME alone, the effects of RME could not be reliably distinguished from those of concomitant orthodontic appliances. Therefore, the observed DNA methylation changes should be interpreted as reflecting the combined effects of RME and adjunctive orthodontic appliances rather than the isolated effect of RME alone. Third, because all participants were Japanese children with a relatively homogeneous ethnic background, the influence of ethnic heterogeneity was likely limited. However, individual genetic background and family history were not evaluated in the present study. Because DNA methylation patterns are influenced by genetic variation, these unmeasured factors may have contributed to the inter-individual variability observed in DNA methylation responses. Future studies incorporating genomic information and family history are warranted to clarify the contribution of genetic factors to orthodontic treatment-associated epigenetic responses. Fourth, the extent to which the observed methylation changes reflect osseous responses at the midpalatal suture or represent systemic changes shared across other tissues, such as blood, remains unclear. Quantitative correlations between radiographic findings and DNA methylation changes were beyond the scope of the present study and were therefore not evaluated. Future studies integrating radiographic assessments with genome-wide DNA methylation analyses may help clarify the relationship between skeletal remodeling and epigenetic responses following RME. Finally, because the observed DNA methylation changes were generally modest in magnitude, the biological significance of individual CpG alterations remains uncertain and requires validation in larger independent cohorts. Representative CpG sites should also be validated using orthogonal methods, such as pyrosequencing, to confirm the reproducibility of the observed DNA methylation changes.

Future studies involving larger cohorts and parallel analyses across multiple tissues, including blood, gingival tissue, and saliva, are needed to clarify the tissue specificity and systemic nature of the epigenetic alterations associated with RME. Such investigations may also inform the potential application of DNA methylation as a biomarker in orthodontic treatment.

This exploratory study suggests that RME may be associated with detectable changes in saliva-derived DNA methylation patterns. Although the biological significance of these epigenomic alterations remains to be established, this study provides the first genome-wide evidence that RME is associated with detectable DNA methylation changes in saliva. These findings provide a foundation for future investigations into the epigenetic mechanisms underlying orthodontic treatment and may support the future development of saliva-based epigenetic biomarkers for monitoring biological responses to orthodontic therapy.

Methods

Study population

Japanese pediatric patients in the mixed dentition stage, who underwent orthodontic treatment with RME at the university-affiliated orthodontic clinics, were enrolled in this study. The RME group comprised 20 patients (mean age, 9.6 ± 0.9 years), including 9 males and 11 females. As a control group, four Japanese pediatric patients with no history of orthodontic treatment were included (mean age, 11.0 ± 1.1 years; 0 males and 4 females). Among patients treated with RME, some received adjunctive orthodontic appliances, including expansion plates, bihelix appliances, and face mask appliances. The mean treatment duration in the RME group was 129.4 days (Table 2). Patients with systemic diseases, congenital anomalies, or syndromic conditions, such as cleft lip and/or palate, as well as those with a history of orthodontic treatment, were excluded from the study⁴³.

Table 2.

Participant characteristics.

Rapid maxillary expansion (RME) group: 9 boys and 11 girls
n = 20	Sex	Age (years)	Treatment duration (days)	Additional appliances
A	Male	10	112	Yes
B	Female	11	189	No
C	Male	8	132	Yes
D	Female	9	153	Yes
E	Female	10	105	Yes
F	Female	10	139	Yes
G	Male	11	126	Yes
H	Male	9	109	No
I	Female	10	115	Yes
J	Male	10	154	No
K	Female	9	147	Yes
L	Female	10	139	Yes
M	Female	10	133	Yes
N	Female	10	105	Yes
O	Female	10	146	Yes
P	Male	9	114	Yes
Q	Male	10	119	Yes
R	Female	9	105	Yes
S	Male	8	112	Yes
T	Male	9	117	Yes
Mean		9.6 years	129.4 days	yes: 17/no: 3

Control group: 4 girls
n = 4			Sex	Age (years)
U			Female	10
V			Female	12
W			Female	12
X			Female	10

Saliva samples were collected from all participants using the Oragene® DNA kit (DNA Genotek Inc, Canada) for the evaluation of genome-wide DNA methylation. In the RME-treated group, samples were collected at two time points: before placement of the RME appliance (T0) and at 12 ± 5 weeks after completion of maxillary expansion (T1). The T1 sampling time point corresponded to the early retention period following maxillary expansion, during which active tissue remodeling and bone remodeling within the midpalatal suture are expected to occur, making this period appropriate for evaluating early epigenetic responses to RME. The expansion protocol consisted of one turn per day (0.20 mm per turn). To prevent relapse following expansion, the RME appliance was retained intraorally from completion of expansion until T1⁴⁴. Successful separation of the midpalatal suture was confirmed in all patients before the retention period using either CBCT (n = 2) or maxillary occlusal radiography (n = 18) (Fig. 3). No patients without confirmed suture separation were included in the study. In the untreated control group, saliva samples were similarly collected at two time points, with the second collection performed 12 ± 5 weeks after the initial sampling. All saliva samples were collected between 09:00 and 17:00, prior to any treatment procedures, to minimize the risk of blood contamination.

Fig. 3.

Confirmation of midpalatal suture separation using occlusal radiographs or cone-beam computed tomography (CBCT) images.

Occlusal radiographs and CBCT images were acquired using radiology equipment at the participating institutions. Occlusal radiographs were obtained using an ALULA system (Kyoto, Japan) with an exposure setting of 70 kV, 6 mA, and an acquisition time of 0.32 s. CBCT images were acquired with a KaVo OP 3D Vision system (Biberach, Germany) at 120 kV, 5.0 mA, a voxel size of 0.3 mm, and an acquisition time of 17.8 s. Additional occlusal radiographs were obtained using a MORITA V080 system (Kyoto, Japan) at 70 kV, 7.0 mA, and an acquisition time of 0.32 s.

The present study was conducted in accordance with the Declaration of Helsinki and followed current standards for reporting observational epidemiological studies (i.e., STrengthening the Reporting of OBservational studies in Epidemiology checklist). Written informed consent was obtained from the parents of all participants, and informed assent was obtained from the children prior to participation. This study was approved by the Ethics Committees of Kanagawa Dental University (Approval No. 1079), Kanazawa University (Approval No. 114855–1), and the National Center for Child Health and Development (Approval No. 2024–259). The privacy rights of human subjects were observed.

In this study, sex was defined as the biological sex assigned at birth (male or female) based on medical records. Gender identity was not assessed. Also, no sex-stratified analyses were performed due to the limited sample size.

Because of the exploratory nature of this genome-wide study and the limited availability of eligible participants, no formal sample size calculation was performed, and the sample size was determined based on feasibility.

DNA isolation and DNA methylation arrays

Genomic DNA was extracted from saliva samples using an Oragene DNA Kit (DNA Genotek Inc., Canada) according to the manufacturer’s instructions. DNA concentration was quantified using a Qubit™ Flex Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). Bisulfite conversion of genomic DNA was performed using the EpiTect® Plus DNA Bisulfite Kit (QIAGEN, Hilden, Germany). Following bisulfite treatment, DNA samples were processed on the Illumina Infinium Human Methylation EPIC v2.0 BeadChip (Illumina Inc., San Diego, CA, USA) according to the manufacturer’s standard protocol, which included hybridization and fluorescent labeling. Arrays were scanned using an iScan System (Illumina), and raw fluorescence intensity data were obtained in IDAT format for downstream analysis.

Bioinformatics analysis

Data preprocessing and quality control

Raw IDAT files were imported into R (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria) using the Minfi package (version 1.54.1). Data preprocessing and normalization were performed using the ENmix package (version 1.44.3) through a multi-step pipeline⁴⁵. Briefly, background correction based on out-of-band signals, dye-bias correction using the RELIC method, and correction of Type I/II probe design bias using regression on correlated probes were applied. Subsequently, quantile normalization was conducted to standardize signal intensity distributions across all samples.

Probe-level quality control and filtering were performed using the champ.filter function in the ChAMP package (version 2.38.1). Probes with detection P-values > 0.01, bead counts < 3, multi-hit probes, and probes affected by single-nucleotide polymorphisms were excluded from further analysis⁴⁶. Sex chromosome–associated probes were selectively filtered based on annotations from the Illumina HumanMethylationEPICv2anno.20a1.hg38 database (version 1.0.0) in accordance with the analytical objectives of the study^47,48.

Normalized β values were extracted using the getBeta function and subsequently transformed into M values using a logit transformation (M = log₂[β/(1 − β)]). Principal component analysis (PCA) was performed on M values to assess sources of technical variation. PCA revealed a substantial influence of array slide position on signal variability; therefore, slide position was treated as a batch factor and corrected using the removeBatchEffect function in the limma package (version 3.64.3) in R, minimizing systematic technical bias⁴⁹.

Cell-type composition estimation and adjustment

To account for differences in cellular composition across saliva samples, cell-type proportions were estimated using the EpiDISH package (version 2.24.0). Reference-based deconvolution was performed using the centEpiFibIC.m reference matrix to estimate the relative proportions of the three major cellular components present in saliva: epithelial cells, fibroblasts, and immune cells⁵⁰. When incorporating cell type proportions into the linear models, fibroblasts were used as the reference category to avoid perfect multicollinearity; therefore, only the proportions of epithelial and immune cells were included as covariates. To assess whether salivary cell composition changed during treatment, the estimated proportions of epithelial and immune cells were compared between T0 and T1 in the RME group using paired t-tests.

Linear model design and differential methylation analysis

Differential DNA methylation was analyzed using normalized and batch-corrected M values. In the present study, the RME treatment and untreated control groups were analyzed separately, and DNA methylation changes between T0 and T1 were evaluated within each group.

For paired samples obtained from the same individuals at T0 and T1, intra-subject correlation was accounted for using the duplicateCorrelation function in the limma package⁴⁹. In the linear models, age, sex, co-use of orthodontic appliances, treatment duration, and estimated salivary cell type proportions were included as covariates for the RME treatment group, whereas age, sex, and cell type proportions were included as covariates for the untreated control group.

Model fitting was performed using the lmFit function in limma, followed by contrast specification using the contrasts.fit function and variance stabilization via the empirical Bayes method implemented in the eBayes function⁴⁹. Resulting P values were adjusted for multiple testing using the Benjamini–Hochberg procedure, and a false discovery rate (FDR) < 0.05 was considered statistically significant. Statistically significant CpG sites were visualized using volcano and Manhattan plots and annotated to genes and genomic regions based on the Illumina HumanMethylationEPICv2anno.20a1.hg38 annotation database. As a post hoc sensitivity analysis, progressively stricter absolute Δβ thresholds (|Δβ|> 0.01, > 0.02, and > 0.05) were applied to the CpG sites identified as significantly differentially methylated (FDR < 0.05), and the number of CpG sites meeting each threshold was summarized. In addition, differential methylation analysis was repeated using normalized M values without prior batch correction, with slide position included as a covariate in the linear model.

Interaction models and functional enrichment analysis

Within the RME treatment group, interaction models were constructed to examine whether changes in DNA methylation from T0 to T1 were modified by clinical background factors. The clinical variables evaluated included orthodontic appliance co-usage, sex, treatment duration, and age. For each clinical variable, interaction terms with the time factor (T0 vs. T1) were incorporated into linear models to assess their effects on DNA methylation changes over time. Moderated t- and F-tests were performed using the limma package. Resulting P values were adjusted for multiple testing using the Benjamini–Hochberg procedure, and an FDR < 0.05 was considered statistically significant, enabling a comprehensive assessment of the modifying effects of clinical factors on biological responses to RME^49,51.

Functional enrichment analysis was conducted using genes annotated to statistically significant CpG sites. To correct for inherent bias due to CpG probe distribution on DNA methylation arrays, KEGG pathway over-representation analysis was performed using the missMethyl package (version 1.42.0), which incorporates CpG density correction. Biologically relevant pathways were subsequently evaluated for interpretation⁵².

Author contributions

T.I.: Conceptualization, Methodology, Formal analysis, Investigation, Visualization, Writing – original draft. M.T.: Conceptualization, Methodology, Investigation, Writing – review & editing, Funding acquisition. H.K.: Investigation, Data curation. T.K.: Investigation, Data curation, Writing – review & editing. A.T.: Methodology, Formal analysis, Writing – review & editing, Supervision. T.Y.: Conceptualization, Writing – review & editing, Supervision, Project administration, Funding acquisition. All authors read and approved the final manuscript.

Funding

This work was supported by the Kanagawa Dental University 2024 Project Research Grant and JSPS KAKENHI (Grant Numbers 21K10195, 24K20091, 24K13185 [FY2024–FY2026], and 24K02109). The funders had no role in study design, data collection, analysis, or decision to publish.

Data availability

The microarray data have been deposited in the GEO under accession number GSE317740 and will be publicly available upon publication.

Declarations

Competing interests

The authors declare no competing interests.