Proliferative verrucous and homogeneous Leukoplakias exhibit differential methylation patterns

Abstract

Objective

Proliferative verrucous leukoplakia (PVL) is considered a clinically distinct entity from other oral leucoplakias (OLs) due to its clinical presentation and evolution. However, molecular differences between them remain unclear. We aimed to determine whether there are methylation differences between PVL and other forms of OLs.

Materials and Methods

Oral biopsies from 12 patients with PVL, eight patients with homogeneous leucoplakia (HL), and 10 healthy individuals were obtained for a genome‐wide DNA methylation analysis via the Infinium EPIC Platform.

Results

A total of 1815 differentially methylated CpGs were found between PVL and HL, with a prominent state of hypermethylation in HL patients. CpGs covered 813 genes with distinct roles, including cell adhesion, extracellular matrix organization, and cell and synaptic signaling. 43% of these genes had been previously described in cancer and associated with prognosis. We developed a multinomial logistic regression model able to differentiate HL, PVL, and control samples. The model had a cross‐validated estimate of 73% and included differentially methylated cancer‐related genes between the pathological conditions and the healthy donors, including ADNP, BRCA2, CDK13, GNB1, NIN, NUMB, PIK3C2B, PTK2, SHISA4, THSD7B, WWP1, and ZNF292. It also included CpGs covering differentially methylated genes in HL (MEN1 and TNRC6B) and PVL (ACOXL, ADH1B, CAMTA1, CBFA2T3, CPXM2, LRFN2, SORCS2, and SPN).

Conclusions

PVL and HL present differential methylation patterns that could be linked to their differential clinical behavior. Our findings show the potential of methylation markers and suggest novel diagnostic biomarkers.

1. INTRODUCTION

Oral potentially malignant disorders (OPMDs) are oral mucosal abnormalities associated with an increased risk of cancer development (Warnakulasuriya et al., 2020). Patients with OPMD may have increased susceptibility to malignancy. However, most of these OPMDs may not evolve into cancer. Instead, they represent abnormal areas where cancer occurrence is more likely than in clinically normal mucosa (Warnakulasuriya et al., 2020). The most frequent forms of OPMD are oral leucoplakias (OL), which consist of predominantly white patches (Warnakulasuriya et al., 2007). A recent systematic review and meta‐analysis estimated that the potential of OLs for malignant transformation is near 10% and some factors, such as histological findings of high‐grade epithelial dysplasia, location of the lesion on the tongue and floor of the mouth, and the existence of non‐homogeneous clinical forms are associated with a higher risk of cancer development (Aguirre‐Urizar et al., 2021; Bagan et al., 2022).

Proliferative verrucous leukoplakia (PVL) is a distinct form of multifocal OL frequently located on the gingiva that is characterized by the highest rate of cancer development and a very high recurrence frequency after treatment. The term was coined by Hansen et al. in 1985 when presenting 30 patients with a particular form of leukoplakia of unknown origin and with a high tendency to evolve into cancer (Hansen et al., 1985). Since then, this term has been widely accepted and used by the scientific community (Alkan et al., 2022; Cerero‐Lapiedra et al., 2010; Fettig et al., 2000; Herreros‐Pomares, Hervás, Bagan‐Debon, et al., 2023; Herreros‐Pomares, Llorens, Soriano, Zhang, et al., 2021; Okoturo et al., 2018; Zakrzewska et al., 1996), being included in the nomenclature and classification of OPMDs established in 2007 (Warnakulasuriya et al., 2007) and in a recent consensus report from the international seminar on nomenclature and classification convened by the World Health Organization (WHO) Collaborating Centre for Oral Cancer (Warnakulasuriya et al., 2020). It is estimated to evolve into oral cancer in 43%–71% of cases, tending to develop second primary malignancies (Bagan et al., 2020). However, although PVL is already considered a clinically distinct entity from other OLs due to its clinical presentation and evolution, the molecular differences between PVL and other OLs remain largely unexplored. As a result, various suggestions have been put forward to adjust, expand, or even abandon the term (Aguirre‐Urizar, 2011; Alabdulaaly et al., 2022; van der Waal, 2021; Villa et al., 2018). For instance, Aguirre‐Urizar advocated for replacing PVL with proliferative multifocal leukoplakia (Aguirre‐Urizar, 2011), whereas others expressed their preference to replace PVL with proliferative (erythro)leukoplakia since the verrucous areas are not always evident (Alabdulaaly et al., 2022; Villa et al., 2018). Recently, it has been suggested to consider the cases reported so far as examples of widespread/multifocal verrucous (erythro)leukoplakias (van der Waal, 2021). Identifying the various genetic and epigenetic changes that lead to the malignant conversion of the different OLs could be relevant at different levels. On the one hand, it could promote the early detection of asymptomatic carcinoma or precursor lesions, but it could also boost the differential diagnosis and treatment of PVL (Villa & Bin, 2017). It has been observed that certain OLs share some of the driver gene mutations frequently found in cancer and they also carry chromosomal instability, including telomerase dysfunction, loss of heterozygosity, and DNA aneuploidy (Odell, 2021; Odell et al., 2021; Oulton & Harrington, 2000; Sen, 2000; Siebers et al., 2013).

DNA methylation has been seen to play a crucial role in various biological processes, including the regulation of gene expression, inflammation, genomic imprinting, cell differentiation, and development. Unsurprisingly, certain alterations in DNA methylation have been strongly associated with different diseases, including cancer. For instance, research suggests that by analyzing the DNA methylation of certain genes, early‐stage oral cancer lesions may be detected (Gong et al., 2015). It is noteworthy that DNA hypermethylation at specific loci, such as CDH1, CDKN2A, and MGMT, has been discovered in oral dysplastic lesions and cancers and has been linked to tumor initiation and progression (Kato et al., 2006; Kulkarni & Saranath, 2004). Furthermore, the use of bisulfate sequencing on a 13‐gene panel has been suggested as a dependable method for detecting oral squamous cell carcinoma (OSCC) in its early stages with high precision (Morandi et al., 2017). Our group has also identified more than 4500 differentially methylated regions in patients with PVL compared to healthy donors and confirmed that the status of deregulation found for some cancer genes in patients with PVL was concordant with that of patients with oral cancer (Herreros‐Pomares, Llorens, Soriano, Bagan, et al., 2021). In addition, we recently reported that oral cancers preceded by PVL exhibit hypermethylation of the promoter region of many cancer‐related genes compared to oral cancer not preceded by PVL (Herreros‐Pomares, Hervás, Bagan‐Debon, et al., 2023).

In the era of postgenomics, high‐throughput technologies are crucial for studying the molecular biology of diseases. Currently, we cannot predict whether a particular oral mucosal abnormality will progress to oral cancer, but the molecular classification of these abnormalities may be useful in addressing this issue. Next‐generation sequencing (NGS) offers new possibilities for discovering new biomarkers that can aid in the diagnosis, prognosis, and treatment response of OLs and it can also enhance our understanding of the molecular mechanisms involved in diseases. With this in mind, we designed a methylated DNA immunoprecipitation and high‐throughput sequencing (MeDIP‐seq) study to investigate the molecular differences between PVL and classical homogeneous leucoplakia (HL).

2. MATERIALS AND METHODS

2.1. Patients and tissue samples

This study included 30 individuals who were treated at the Stomatology and Maxillofacial Surgery Department of the General University Hospital of Valencia. An evaluation of the oral microbiome of these patients is available at (Herreros‐Pomares, Hervás, Bagan‐Debon, et al., 2023). Participants were distributed into three groups according to their oral status. Group I consisted of ten healthy donors, Group II included eight patients with HL, and Group III comprised twelve patients with PVL. For the control group, samples were obtained from healthy mucosa areas adjacent to the teeth (vestibular fundus) and fresh‐frozen (FF) at −80°C until analysis. For groups II and III, two representative biopsies were taken from the same area of the lesions, including the epithelium and the underlying connective tissue between 2017 and 2021. One of each pair of specimens was analyzed with routine histopathological methods to ensure that the observed lesions met the histopathological criteria to establish the diagnosis together with the clinical data of each patient. PVL diagnosis was determined following the criteria provided by Villa et al. (Villa et al., 2018). The other sample was FF stored at −80°C until analysis. This study was approved by the Ethics Committee for Human Research of the University of Valencia (Ref. H1523722754549) and the Ethical Committee for Research of the Consortium of the General University Hospital of Valencia (30 May 2019). Informed written consent was obtained from all participants after an explanation of the nature of the study.

2.2. DNA extraction and genome‐wide DNA methylation analysis

Total DNA from clinical samples was extracted using a column‐based DNA extraction method (E.Z.N.A. DNA kit and DNeasy Blood & Tissue Kit; Qiagen, Hilden, Germany) following the manufacturer's instructions. All DNA samples were treated with RNaseA for 1 h at 45°C, quantified by the fluorometric method (Quant‐iT PicoGreen dsDNA Assay, Life Technologies, CA, USA), and assessed for purity by NanoDrop 2000 (Thermo Scientific, MA, USA) 260/280 and 260/230 ratio measurements. The DNA integrity of FF samples was checked by electrophoresis on a 1.3% agarose gel. Epigenomic studies were performed with the Infinium EPIC DNA methylation BeadChip platform (Illumina) used for the interrogation of over 850,000 CpG sites (dinucleotides that are the main target for methylation), which has been previously established as a reliable technology to detect epigenetic alterations in our and other laboratories (Moran et al., 2016). Then, 600 ng of purified DNA was randomly distributed on a 96‐well plate and processed using the EZ‐96 DNA Methylation kit (Zymo Research Corp., CA, USA) following the manufacturer's recommendations for Infinium assays. Bisulfite‐converted DNA was processed as previously described (Sandoval et al., 2011). MethylationEPIC BeadArray shares the Infinium HD chemistry Assay (Illumina Inc.) used to interrogate the cytosine markers with HumanMethylation450 BeadChip. Thus, the applicable protocol for MethylationEPIC is the same as that for HumanMethylation450, which is the Infinium HD Methylation Assay Protocol. Four microliters of tissue bisulfite‐DNA were processed following the Illumina Infinium HD Methylation Assay Protocol (Sandoval et al., 2011).

2.3. Data preprocessing

Raw data (IDATs) were normalized using the minfi R package (version 1.38) and functional normalization. CpG markers present on MethylationEPIC were classified based on their chromosome location, the Infinium chemistry used to interrogate the marker (Infinium I, Infinium II), and the feature category gene region as per UCSC annotation (TSS200, TSS1500, 5'UTR, 1st Exon, Body, 3'UTR). Additional criteria included the location of the marker relative to the CpG island (open sea, island, shore, shelf), fantom 5‐associated enhancer regions, and regulatory regions described in the ENCODE project such as transcription binding site sequences, open chromatin regions, and digital DNase I hypersensitivity clusters. Every beta value in the EPIC array was accompanied by a detection p‐value that represents the confidence of a given beta value. Probes and sample filtering involved a two‐step process in which unreliable betas with high detection p > 0.01 (1620 CpGs) and 2932 CpGs associated with SNPs were removed. Previous analyses indicated that a threshold value of 0.01 allows a clear distinction to be made between reliable and unreliable beta values (Zhou et al., 2017). Sex chromosome probes were also removed (19,681 CpGs). After filtering, the remaining 842,179 CpGs were considered valid for the study.

2.4. Statistical analysis

Exploratory analysis of the methylation data was performed using principal component analysis (PCA) and heatmaps of clustered observations and variables. An elastic net penalized logistic regression model was adjusted to select the CpGs able to discriminate between groups. The penalization factor for the elastic net was selected by taking the highest lambda at one standard error from the minimum (one‐standard‐error rule) from 500 repetitions of 10‐fold cross‐validation. Then, the median of the 500 lambda values was used as the final penalization factor. The elastic net parameter alpha (regulating the mix of L1 and L2 penalization) was set at 0.4. CpGs with coefficients different from zero after the penalization were selected by the analysis and were therefore considered relevant for discriminating between groups. The performance of the model was assessed by estimating the average misclassification rate over fifty repetitions of five‐fold cross‐validation. Additionally, a rank‐based regression model was also adjusted for each CpG to assess differential methylation between groups. p‐values of the rank‐based model were adjusted for multiple comparisons by using the false discovery rate (FDR). Gene enrichment analysis was performed using ShinyGo (Ge et al., 2020) based on KEGG and GO biological processes pathways databases. All statistical analyses were performed using R (version 4.2.0) and the R packages glmnet (version 4.1–4) and Rfit (version 0.24.2).

3. RESULTS

3.1. Patients' characteristics

This study included 8 patients with a clinical diagnosis of HL, 12 patients with a clinical diagnosis of PVL, and a control group comprising 10 healthy donors. The clinicopathological and evolutionary characteristics of each of the patients with HL or PVL are shown in Table 1 and Figure 1. The mean age of patients with HL was 65.62 years [range: 54–82], 62.5% were women and 50% were smokers. HL in non‐smoker patients were idiopathic, with no apparent cause. The mean age of patients with PVL was 66.83 years [range: 50–84], 75% were women and 41.7% were smokers. In the HL group, two patients had no dysplasia (25%), five had mild dysplasia (62.5%), and one had moderate dysplasia (12.5%). In the PVL group, three patients had no dysplasia (25%), seven had mild dysplasia (58.4%), one had moderate dysplasia (8.3%), and one had severe dysplasia (8.3%). Of note, PVL were advanced cases with a long evolution time, which entails a more advanced progression of the lesions and a greater presence of epithelial dysplasia. None of the selected cases presented histopathological signs of OSCC at the time of the biopsy. None of the HL cases included in the study transformed into oral carcinoma. However, five patients with PVL underwent malignant transformation into OSCC (41.7%). All patients whose PVL transformed into OSCC previously had warty lesions with varying degrees of dysplasia. The mean time to malignant transformation was 6.8 ± 5.5 months.

TABLE 1.

Clinical pathological information of the patients with HL and PVL included in the study.

HL patients
Case	Age	Gender	Tobacco smoker	Follow‐up (years)	Number of oral locations	Size of the oral leukoplakia (cm)	Oral location	Histopathologic diagnosis	Malignant transformation
1	65	F	Yes	4.07	1	1.5	Palate	Mild dysplasia	No
2	82	F	No	3.13	1	5	Upper gingiva	Hyperkeratosis without dysplasia	No
3	64	M	No	21.17	1	3	Tongue	Mild dysplasia	No
4	68	F	No	19.75	1	2.5	Tongue	Mild dysplasia	No
5	54	M	No	2.07	1	2	Buccal mucosa	Mild dysplasia	No
6	70	M	Yes	3.32	1	3	Tongue	Hyperkeratosis without dysplasia	No
7	58	F	Yes	1.38	1	2	Floor of the mouth	Moderate dysplasia	No
8	64	F	Yes	1.48	1	2	Floor of the mouth	Mild dysplasia	No

PVL patients
Case	Age	Gender	Tobacco smoker	Follow up of PVL (years)	Number of oral locations	Percentage of oral mucosa affected with PVL lesions	Oral locations (The location of tissue sample included in the present study is underlined)	Histopathologic diagnosis of included sample	Malignant transformation
1	76	F	No	3.22	14	54	Gingiva, buccal mucosa, palate, retromolar trigone	Mild dysplasia	No
2	56	F	Yes	12.51	8	31	Gingiva, buccal mucosa, palate, floor mouth	Mild dysplasia	No
3	55	M	Yes	15.02	6	23	Gingiva, buccal mucosa, palate	Severe dysplasia	Yes
4	52	F	Yes	7.76	18	70	Gingiva, buccal mucosa, palate, lip, tongue	Hyperkeratosis without dysplasia	No
5	84	F	No	20.06	20	77	Gingiva, buccal mucosa, palate, floor mouth, tongue	Moderate dysplasia	Yes
6	67	M	Yes	8.19	9	35	Gingiva, buccal mucosa, palate, floor mouth, tongue, retromolar trigone	Mild dysplasia	Yes
7	70	F	No	5.93	14	54	Gingiva, buccal mucosa, palate, floor mouth, tongue	Hyperkeratosis without dysplasia	No
8	63	F	Yes	10.68	16	62	Gingiva, buccal mucosa, palate, floor mouth, tongue	Hyperkeratosis without dysplasia	No
9	82	F	No	7.04	12	46	Gingiva, buccal mucosa, palate, lip	Mild dysplasia	No
10	65	F	No	11.12	11	42	Gingiva, buccal mucosa, palate, floor mouth, tongue	Mild dysplasia	Yes
11	82	F	No	7.69	16	62	Gingiva, buccal mucosa, lip, tongue	Mild dysplasia	No
12	50	M	No	5.16	14	54	Gingiva, buccal mucosa, palate, tongue	Mild dysplasia	Yes

FIGURE 1.

Representative images of the oral mucosa affected by HL (Case 6) (a) and PVL (Case 10) (b).

3.2. Differential methylation analysis reveals aberrant methylation in cancer‐related genes of HL and PVL

Total DNA from clinical samples was extracted, and epigenomic studies were performed with the Infinium EPIC DNA methylation BeadChip platform (Illumina) used for the interrogation of over 850,000 CpG sites. A total of 842,179 CpGs were available after the preprocessing and quality filtering of the raw methylation data. An unsupervised PCA was performed to group specimens according to their methylation status (Figure 2a). The PCA score plot separated controls from HL and PVL cases and revealed that the methylation status of healthy donors was more homogeneous than that of patients with HL or PVL (Figure 2a). Next, we performed a differential methylation analysis between patients with HL and PVL that revealed a total of 1815 differentially methylated CpGs under an FDR < 0.05. An unsupervised hierarchical clustering analysis was used to group samples based on these CpGs (Figure 2b). Individuals were classified into two major subgroups, showing a differentiated methylation pattern between PVL and HL. Importantly, 138 out of the 1815 CpGs were hypermethylated in PVL patients compared to HL patients, whereas 1677 were hypomethylated. At the gene level, the selected CpGs corresponded to 813 genes, including 704 protein‐coding genes, 52 long non‐coding RNAs (lncRNAs), 34 microRNAs (miRNAs), 8 small nucleolar RNAs (snoRNAs), and 15 pseudogenes (Files S1 and S2). Interestingly, two enriched regions were detected on chromosome 14 (File S3). Downstream evaluation of the functional profiles specifically associated with these genes revealed significant enrichment in pathways and ontologies such as cell adhesion, actin filament‐based movement, extracellular matrix organization, cell–cell signaling, synaptic signaling, and glutamatergic, dopaminergic, serotonergic, and cholinergic synapses (Figure 2c,d; Files S4 and S5). From that point on, it is important to note that even if these CpGs are prominently hypermethylated in HL, this does not necessarily indicate that the associated genes are silenced, but that certain exons of these proteins may be silenced. In this regard, of the 704 coding genes involved in these epigenetic patterns, 186 are considered cancer‐related genes according to The Network of Cancer Genes (NCG) and 303 have shown prognostic value in different cancer types (File S2).

FIGURE 2.

Metrics for the differential methylation comparison performed between the control, HL, and PVL groups. (a) PCA score plot of samples according to the methylation levels of all the CpGs. (b) Unsupervised clustering of PVL and HL patients based on the methylation levels of the 1815 CpGs detected as differentially methylated. (c) Dot plots displaying the top 20 GO terms of the biological process category obtained from the enrichment analysis. (d) Dot plots displaying the KEGG pathways obtained from the enrichment analysis.

3.3. Methylation patterns can effectively differentiate HL, PVL, and control patients

To further investigate the feasibility of creating a model that can differentiate HL, PVL, and control samples using methylation patterns, we conducted an elastic net penalized logistic regression analysis. We identified 163 CpGs that cover 99 genes and can effectively differentiate between groups (Files S6 and S7). The performance of the model was assessed by estimating the average accuracy rate over fifty repetitions of five‐fold cross‐validation. The apparent accuracy rate of the model was 100%, whereas the cross‐validated estimate was 73%. The results of the selected CpGs are shown in an unsupervised heatmap where groups are clearly separated by their methylation status (Figure 3a). HL and PVL samples showed higher methylation levels than controls in 76 of the CpGs. This differential methylation between healthy and pathologic conditions involved 52 genes, 12 of which were cancer‐related genes: ADNP (cg12310925), BRCA2 (cg22145805), CDK13 (cg02035448, cg17413540), GNB1 (cg12134886), NIN (cg16455234), NUMB (cg07185684), PIK3C2B (cg23296861), PTK2 (cg26890815), SHISA4 (cg18059457), THSD7B (cg01439023), WWP1 (cg25223327), and ZNF292 (cg15443043) (Figure 3b; File S8). In contrast, control samples displayed higher methylation levels in four CpGs that corresponded to KIAA1549L (cg22068574) and LZTFL1 (cg09355011). HL samples had higher methylation levels in 17 CpGs and lower methylation levels in two when compared to the control and PVL groups. This aberrant methylation pattern covered 12 genes, including MEN1 (cg18759725) and TNRC6B (cg15031003). In the case of PVL, 62 CpGs were hypomethylated, whereas only two were hypermethylated. The hypomethylated regions involved 38 genes, including 8 cancer‐related genes: ACOXL (cg12966876), ADH1B (cg06612891), CAMTA1 (cg07946652, cg11112634), CBFA2T3 (cg05240948), CPXM2 (cg05767088, cg27289181), LRFN2 (cg00285537), SORCS2 (cg06570432), and SPN (cg22543378) (Figure 3b; File S8).

FIGURE 3.

Results of the logistic regression model adjusted to select the CpGs able to differentiate HL, PVL, and control samples. (a) Unsupervised clustering according to the methylation status of the 163 CpGs selected. (b) Percentage of methylation of the CpGs covering tumor suppressor genes and candidate cancer drivers for each group.

4. DISCUSSION

In this study, a differential methylation analysis was performed between patients with classical HL and patients with PVL according to the diagnostic criteria provided by Villa et al. (Villa et al., 2018). Our analyses showed that HL and PVL are different molecular entities with clearly differentiated methylation patterns. These epigenetic modifications affected more than 700 protein‐coding genes that were associated with pathways and ontologies such as cell adhesion, extracellular matrix organization, and cell or synaptic signaling. Importantly, more than 25% of the differentially methylated genes were cancer‐related genes according to the NCG and more than 43% have been reported to have prognostic value in different cancer types. To date, some studies have reported DNA methylation markers in oral precancer lesions (Shridhar et al., 2016). Among them, the most reported hypermethylated gene biomarkers are p16 (methylation‐positive rate, 17.5%–87.5% in cases vs. 0–14.3% in controls), p14 (3.8%–73.4% in cases vs. null in controls), and MGMT (4.0%–72.7% in cases vs. 0–14.3% in controls) (Shridhar et al., 2016). Unfortunately, most of the studies are small, cross‐sectional studies with poorly defined control groups and lacking validation. A larger study performed by Cheng and colleagues showed the efficacy of hypermethylated ZNF582 and PAX1 genes as biomarkers for oral dysplasia and cancer detection in patients with OPMDs and OSCC (Cheng et al., 2016; Cheng et al., 2018). However, to the best of our knowledge, this is the first study addressing the methylation differences between HL and PVL.

We also performed a logistic regression that identified a methylation signature able to distinguish between HL, PVL, and control specimens. The signature comprised 163 CpGs, covering some cancer‐related genes and tumor suppressor genes. Of note, 46.6% of the selected CpGs displayed higher methylation levels in HL and PVL samples than in controls, whereas only 2.5% of the CpGs had lower methylation levels. The CpGs differentiating pathologic conditions from controls included some genes that have been widely associated with head and neck and oral tumors, such as BRCA2, NUMB, or WWP1. Regarding BRCA2, a recent study reported that low levels of BRCA2 in peripheral blood lymphocytes are associated with an increased risk for head and neck squamous cell carcinoma (HNSCC) (Das et al., 2023). Similarly, a retrospective study of 75 patients with HNSCC concluded that BRCA2 is among the five most frequently altered genes (Wilson et al., 2021). Genetic alterations in BRCA2 have also been implicated in all toxicity parameters evaluated in patients who received radiation therapy for primary or locally recurrent HNSCC (Sumner et al., 2021). In oral cancer, observations have shown that knockdown of BRCA2 using small interference RNA suppression affects the sensitivity to 5‐fluorouracil therapy (Nakagawa et al., 2014) and a retrospective study including 60 OSCC cases found a significant correlation between cytoplasmic BRCA2 expression and histological grade (Irani & Rafizadeh, 2020). Numb is an adaptor protein capable of suppressing malignant transformation and is targeted by miR‐31/96/182 (Chou et al., 2018). In HNSCC, increased migration and invasion of cells have been associated with the exogenous expression of miR‐31/96/182 and this was reversed by the expression of Numb (Chou et al., 2018). Mechanistically, the miR‐31‐NUMB cascade has been reported to modulate monocarboxylate transporters to increase oncogenicity and lactate production in OSCC cells (Chou et al., 2021). In addition, Numb has also been associated with the inhibition of epithelial‐mesenchymal transition via the RBP‐Jκ‐dependent Notch1/PTEN/FAK signaling pathway in tongue cancer (Li, Huang, et al., 2019). Overexpression of WWP1, a WW domain‐containing protein that also plays an important role in the regulation of a wide variety of cellular functions such as protein degradation, transcription, and RNA splicing, has been reported to significantly increase the proliferation and invasion of laryngeal cancer cells (Li, Sun, et al., 2019).

The differential methylation pattern of HL cases included the hypermethylation of CpGs covering the tumor suppressors MEN1 and TNRC6B. Inactivating germline mutations in MEN1 have been associated with multiple neoplasias type 1 (MEN‐1 syndrome), which are characterized by the simultaneous occurrence of at least two of the three main related endocrine tumors: parathyroid, enteropancreatic, and anterior pituitary (Suárez et al., 2006). Variations in this gene have been associated with several tumor types and, in esophageal squamous cell carcinoma (ESCC), MEN1 has been included in a diagnostic model based on expression levels developed and validated in two independent cohorts of patients (Sun et al., 2020). Although less is known about TNRC6B, recent data suggest that LINC01207 promotes HNSCC via TNRC6B upregulation (Chen et al., 2020).

In the case of PVL, 62 CpGs were hypomethylated when compared to patients with HL and controls. The hypomethylated regions involved 38 genes, including ADH1B, CAMTA1, and CBFA2T3. ADH1B gene polymorphisms influence the risk of HNSCC through modulation of acetaldehyde metabolism and propensity to alcohol intake (Matejcic et al., 2017). In particular, a multistage genome‐wide association study of ESCC including more than 3000 individuals concluded that the risk allele rs1229984 of ADH1B is highly associated with disease development (Tanaka et al., 2010). On its behalf, the protein encoded by CAMTA1 is a transcription factor that is believed to act as a tumor suppressor. In glioma cells, CAMTA1 overexpression was demonstrated to inhibit cell growth, migration, and invasion in vitro and attenuate tumor formation and growth in vivo (He et al., 2021). Of note, a recent case report on an oral composite hemangioendothelioma characterized by a YAP1‐MAML2 fusion reported that the neoplastic cells were negative for CAMTA1 (Koutlas et al., 2021). CBFA2T3 is a tumor suppressor gene that mediates transcriptional repression. The expression of CBFA2T3 has been found to be significantly reduced in several breast cancer cell lines and primary breast tumors (Kochetkova et al., 2002). Importantly, the reintroduction of CBFA2T3 into different breast tumor‐derived cell lines reduced colony growth (Kochetkova et al., 2002). CBFA2T3 is also involved in several translocations associated with myeloid malignancies. The t(16;21)(q24;q22) translocation is a rare but recurrent chromosomal abnormality associated with therapy‐related myeloid malignancies (Salomon‐Nguyen et al., 2000). The translocation produces a chimeric gene made up of the 5′‐region of RUNX1 fused to the 3′‐region of this gene in adult therapy‐related acute leukemia. CBFA2T3‐GLIS2 fusion has also been described in pediatric acute myeloid leukemias (Le et al., 2022).

All these findings demonstrate the methylation differences existing between PVL and HL that could be linked to their differential clinical behavior. However, this preliminary study has limitations. On the one hand, the number of cases included in the study is small and the results reported lack validation in independent cohorts. Thus, further studies with larger sample sizes are desirable. On the other hand, we compared PVL cases with a group of classic HL cases. We have already started sample collection to perform a new study including a group of classic leukoplakias of the non‐homogeneous clinical type. DNA methylation is frequently linked with downregulation of gene expression. However, it has also been reported that DNA methylation is positively associated with gene expression, suggesting a more diverse mechanism of epigenetic regulation (Rauluseviciute et al., 2020). Such additional complexity could have important implications for understanding oral leukoplakias but has not been studied at a genome‐wide scale.

5. CONCLUSION

The present study confirmed the presence of differential epigenetic alterations in PVL and HL. Abnormal DNA methylation patterns found in OPMDs involve multiple genes associated with cancer, including potential cancer drivers and tumor suppressor genes linked to prognosis, suggesting DNA methylation as a mechanism for regulating the expression of these genes. Based on our initial research, it seems that genome‐wide methods could be valuable in the study of OPMDs. Our study has limitations due to the small sample size. Larger sample sizes are necessary for further research into the clinical and molecular differences of patients.

AUTHOR CONTRIBUTIONS

José Bagan: Conceptualization; funding acquisition; writing – review and editing; supervision; resources; project administration. Alejandro Herreros‐Pomares: Investigation; methodology; data curation; validation; writing – original draft; visualization; conceptualization; writing – review and editing. David Hervás: Data curation; methodology; validation; formal analysis; visualization; software; writing – review and editing. Leticia Bagán: Methodology; writing – review and editing; resources. Alex Proaño: Resources; writing – review and editing.

FUNDING INFORMATION

This document is the result of a research line funded by the Instituto de Salud Carlos III (Spain) through the project “PI19/00790”, Principal investigator: Jose Bagan, co‐funded by the European Regional Development Fund/European Social Fund “A way to make Europe/Investing in your future” and the Ministerio de Ciencia e Innovación (Spain) through the project “PID2022‐138398OB‐I00”, Principal investigator: Jose Bagan.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no competing interests.

Supporting information

Additionalfile1

Additionalfile2

Additionalfile3

Additionalfile4

Additionalfile5

Additionalfile6

Additionalfile7

Additionalfile8

Herreros‐Pomares, A. , Hervás, D. , Bagán, L. , Proaño, A. , & Bagan, J. (2025). Proliferative verrucous and homogeneous Leukoplakias exhibit differential methylation patterns. Oral Diseases, 31, 137–147. 10.1111/odi.15028

DATA AVAILABILITY STATEMENT

Methylation data have been deposited in the ArrayExpress database at EMBL‐EBI (www.ebi.ac.uk/arrayexpress) under accession number E‐MTAB‐12202.