Aberrant DNA methylation of SOX1 and PAX1 as epigenetic biomarkers and prognostic indicators in oral squamous cell carcinoma: a single-centre retrospective study

Abstract

Background

Aberrant DNA methylation is a cancer hallmark with diagnostic and prognostic potential. This study aimed to investigate the methylation status of SOX1 and PAX1 in oral squamous cell carcinoma (OSCC) tissues and adjacent normal tissues, and explore their associations with clinicopathological features and patient prognosis.

Methods

A single-center retrospective cohort of 164 OSCC patients was analyzed. Methylation levels of SOX1 and PAX1 were detected using quantitative methylation-specific PCR (Q-MSP) in formalin-fixed paraffin-embedded (FFPE) tumor tissues and 88 matched normal tissues. Associations with clinicopathological parameters (tumor size, lymph node metastasis, clinical stage) and survival outcomes.

Results

OSCC tumor tissues exhibited significantly higher methylation indices (M-index) for both SOX1 (273.32 vs. 93.57, P = 0.039) and PAX1 (720.92 vs. 108.52, P < 0.0001) compared to adjacent normal tissues, with methylation positivity rates of 56.71% and 73.78%, respectively. Stratified analysis revealed SOX1 methylation positivity was strongly associated with larger tumor size (T stage, χ²=8.04, P = 0.045), lymph node metastasis (N stage, χ²=4.27, P = 0.039), and advanced clinical stage (χ²= 8.33, P = 0.040). Kaplan-Meier survival analysis showed patients with SOX1-methylated tumors had significantly shorter DFS (hazard ratio [HR] = 0.53, 95% CI: 0.31–0.92, P = 0.03), whereas PAX1 methylation status did not correlate with DFS or OS. Combined detection of SOX1 and PAX1 methylation improved sensitivity to 81.71% for OSCC diagnosis.

Conclusion

Promoter hypermethylation of SOX1 and PAX1 was a frequent event in OSCC, with SOX1 methylation specifically linked to aggressive clinicopathological features and poorer disease-free survival. These findings highlighted SOX1 as a potential prognostic biomarker and PAX1 as a candidate diagnostic marker, warranting further validation in multi-center cohorts to inform epigenetic-targeted strategies in oral oncology.

Introduction

Oral squamous cell carcinoma (OSCC), the seventh most prevalent malignancy worldwide, accounted for 4.5% of global cancer diagnoses and 4.6% of cancer-related deaths, posing a significant public health burden [1]. Despite advancements in surgical resection, radiotherapy, and systemic therapies, patient outcomes remained suboptimal, with a 5-year survival rate stagnating at 50–60% due to late-stage presentation, aggressive tumor biology, and high rates of local recurrence or metastasis. Early detection and identification of robust prognostic biomarkers are critical to improving clinical management, as current staging systems (e.g., TNM classification) and histological grading lack sufficient precision to stratify patients at risk of poor outcomes.

DNA methylation, a key epigenetic modification mediated by DNA methyltransferases, involved the addition of methyl groups to cytosine residues within CpG islands, primarily in gene promoter regions [2]. This process regulated transcriptional activity without altering the genomic sequence, influencing cellular differentiation, genome stability, and tumor microenvironment dynamics [3]. In cancer, aberrant methylation patterns frequently arised, including hypermethylation-driven silencing of tumor suppressor genes and hypomethylation-mediated activation of oncogenes, driving oncogenesis and disease progression [4]. Emerging evidence highlighted DNA methylation as a promising biomarker for early cancer detection and prognosis, with methylomic profiling enabling the identification of dysregulated pathways in various solid tumors, including head and neck cancers [5, 6]. However, the methylation landscape of OSCC, particularly the role of specific regulatory genes in tumor progression, remains incompletely characterized.

The SOX (sex-determining region Y-box) and PAX (paired box) gene families encode transcription factors essential for embryonic development and tissue homeostasis, with dysregulation linked to oncogenic transformation. In a prior study, we identified associations between methylation of SOX1 and PAX1 and cervical lesion progression, leveraging their shared mucosal tissue origin and morphological similarities with OSCC [7]. This rationale prompted us to investigate their methylation status in OSCC, hypothesizing conserved epigenetic mechanisms across mucosal malignancies.

SOX1, a member of the SOXB1 subgroup, was critical for neural stem cell maintenance and neuronal differentiation, suppressing glial lineage commitment to ensure central nervous system integrity [8, 9]. Beyond neurodevelopment, SOX1 functioned as a tumor suppressor in breast cancer by inhibiting the Wnt/β-catenin signaling pathway, reducing cell proliferation and invasion [10–13]. Recent whole-genome bisulfite sequencing in oral dysplastic keratinocytes identified SOX1 as a differentially methylated tumor suppressor gene, suggesting its involvement in oral carcinogenesis [14]. However, the role of SOX1 methylation in OSCC pathogenesis and its association with clinicopathological features remain undefined.

PAX1, a key regulator of skeletal and epithelial development, exerted tumor suppressive effects in cervical, ovarian, and colorectal cancers by downregulating oncogenic kinase pathways (e.g., MAPK/ERK and SRC), maintaining phosphatase-kinase homeostasis, and inhibiting epithelial-mesenchymal transition [15–20]. In OSCC, PAX1 hypermethylation has been detected in exfoliated oral cells, correlated with tumor presence; however, its functional significance in tumor tissues and prognostic value remain unclear.

Given the unmet need for molecular biomarkers in OSCC and the emerging role of SOX1 and PAX1 in mucosal cancer epigenetics, this study aims to characterize the methylation status of SOX1 and PAX1 in OSCC tissues versus adjacent normal mucosa, and to investigate associations between their methylation and clinicopathological features, including tumor stage, lymph node metastasis, and histological differentiation, and finally to evaluate their prognostic utility in predicting DFS and OS. By leveraging FFPE tissues with confirmed pathological diagnoses, this study seeks to clarify the epigenetic landscape of SOX1 and PAX1 in OSCC, providing foundational insights for biomarker development and targeted therapeutic strategies.

Materials and methods

Study population and clinical data collection

A retrospective cohort of 164 patients with histopathologically confirmed OSCC was enrolled from the First and Seventh Medical Centers of PLA General Hospital between January 2019 and December 2020. The flowchart was shown in Fig. 1.Inclusion criteria comprised: (1) primary tumor located in the oral cavity (oral tongue, lip, buccal mucosa, gingiva, floor of mouth, or oropharynx); (2) absence of concurrent systemic diseases or severe comorbidities (e.g., uncontrolled diabetes, cardiovascular failure); (3) no prior neoadjuvant radiotherapy or chemotherapy before surgical resection; and (4) definitive OSCC diagnosis confirmed by two board-certified pathologists via hematoxylin and eosin (H&E)-stained sections. Eighty-eight matched adjacent normal mucosal tissues (≥ 1 cm from tumor margins, histologically free of dysplasia or malignancy) were collected as controls.

Fig. 1.

The flowchart of patients

Clinical data, including demographics (age, gender), smoking history (≥ 100 lifetime cigarettes vs. non-smokers), alcohol consumption (≥ 10 g ethanol/week vs. non-drinkers), tumor location, histological differentiation (high, moderate, low), and TNM staging (8th edition AJCC/UICC), were abstracted from electronic medical records. The study was approved by the Ethics Committee of the seventh Medical Center of Chinese PLA General Hospital (Approval No: S2025-054-01), conducted in accordance with the Declaration of Helsinki, and written informed consent was obtained from all participants.

Tissue processing and histopathological validation

Fresh surgical specimens were collected during curative resection and immediately fixed in 4% neutral-buffered paraformaldehyde for 24 h at room temperature. Tissues were processed via standard paraffin embedding, sectioned into 4-µm-thick slices, and stained with H&E. Slides were reviewed by a senior pathologist to confirm OSCC subtype (conventional squamous cell carcinoma) and ensure tumor blocks contained ≥ 30% neoplastic cells, excluding samples with excessive necrosis or inflammation.

DNA extraction and bisulfite conversion

FFPE blocks with confirmed tumor cellularity (≥ 30% neoplastic cells) were selected for DNA extraction. Genomic DNA was isolated using the AmoyDx FFPE DNA Extraction Kit (Cat. No. AM-2020; Xiamen, China) following the manufacturer’s protocol, with elution in 50 µL nuclease-free water. DNA quantity and purity were assessed via NanoDrop 2000c Spectrophotometer (Thermo Fisher Scientific, Waltham, USA), requiring A260/A280 ratios between 1.8 and 2.0.

Bisulfite conversion was performed on 800 ng DNA using the Hybribio Methylation Conversion Kit (Guangdong, China), which converts unmethylated cytosine to uracil while preserving methylated cytosines. Briefly, DNA was denatured with 0.3 M NaOH, treated with bisulfite solution (3.0 M sodium metabisulfite, 0.1 M hydroquinone, pH 5.0), and purified via desalting columns. Converted DNA was eluted in 30 µL nuclease-free water and stored at − 20 °C.

Quantitative methylation-specific PCR (Q-MSP)

Methylation status of SOX1, PAX1, and the endogenous control β-actin was analyzed using TaqMan-based Q-MSP on the SLAN-96 S Real-Time PCR System (Hongshi, Shanghai, China). Primers and probes were designed to target CpG-rich regions in the promoter regions of SOX1 (chr3:181,033,682–181,033,801) and PAX1 (chr20:62,003,312–62,003,420), avoiding single-nucleotide polymorphisms, with sequences validated via bisulfite genomic sequencing in prior studies.

Reaction mixtures (20 µL) contained 5 µL bisulfite-converted DNA, 10 µL 2× TaqMan Universal PCR Master Mix (Applied Biosystems), 0.4 µM forward/reverse primers, and 0.2 µM FAM-labeled probe. Thermal cycling conditions were: 95 °C for 10 min (initial denaturation), followed by 45 cycles of 95 °C for 20 s (denaturation) and 60 °C for 30 s (annealing/extension). Non-template controls and bisulfite-converted methylated human DNA (Zymo Research, Cat. No. D5014) were included as positive/negative controls.

Based on pre-optimization experiments to minimize false positives and improve assay sensitivity/specificity, we defined invalid reactions as SOX1^m threshold cycle (Ct) value > 38.5 and PAX1^m Ct value > 38.0. Methylation status of SOX1/PAX1 was assessed via ΔCt (ΔCt = Ct _{reference gene} − Ct _{target gene}), with respective cutoff ΔCt values of 10.62 (SOX1^m) and 8.95 (PAX1^m).

The methylation index (M-index), reflecting relative methylation levels, was calculated as: M-index = 10,000 × 2^{- |Ct value of target gene - Ct value of internal control gene|} [19].This formula normalizes target gene Ct values to the internal control, enabling inter-sample comparison of methylation intensity.

Statistical analysis

Data were analyzed using SPSS v26.0 (IBM) and GraphPad Prism v9.0 (San Diego, USA). Continuous variables such as M-index are presented as mean ± standard deviation (SD). Between-group comparisons were performed using independent sample T-tests. Survival outcomes were evaluated via log-rank tests. Disease-free survival (DFS) was defined as the interval from diagnosis to recurrence or death, while overall survival (OS) was defined as the time from diagnosis to death due to any cause. Forest plots were generated and analyzed using R software. A p-value < 0.05 was considered statistically significant.

Results

Clinical and pathological characteristics of patients

A total of 164 patients with histopathologically confirmed OSCC were included, comprising 107 males (65.2%) and 57 females (34.8%), with a mean age of 63.3 ± 13.0 years (range: 27–88 years). Tumor locations included the tongue (28.0%), oropharynx (15.2%), gingiva (20.7%), buccal mucosa (14.6%), lip (12.2%), and floor of mouth (9.2%). Histological differentiation was categorized as high (49.4%), moderate (45.1%), or low (5.5%). Clinical staging according to the 8th AJCC/UICC TNM classification revealed: Stage I (19.9%), Stage II (17.3%), Stage III (18.6%), and Stage IV (21.8%). Smoking and alcohol consumption were reported in 39.6% and 40.2% of patients, respectively. The flowchart was shown in Fig. 1. Detailed demographics and clinicopathological features are summarized in Table 1. Representative hematoxylin and eosin (H&E)-stained sections confirming OSCC histology are shown in Fig. 2.

Table 1.

Clinical and Pathological Characteristics of Study Cohort

Variable	Case number	%
Gender
Male	107	65.24
Female	57	34.76
Age
<50	20	12.19
50—69	90	54.88
≥70	54	32.93
Primary site
oropharynx	25	15.24
Tongue	46	28.05
lip	20	12.20
buccal	24	14.63
gingiva	34	20.73
floor of mouth	15	9.15
History of smoking
YES	65	39.63
NO	99	60.37
History of alcohol
Yes	66	40.24
No	98	59.76
differentiation
High	81	49.39
Moderate	74	45.12
Low	9	5.49
Pathological T category
T1	61	39.10
T2	58	37.18
T3	27	17.31
T4	10	6.41
Pathological N category
N0	80	66.12
N1	15	12.40
N2	26	21.49
Clinical stage
I	31	19.87
II	27	17.31
III	29	18.59
IV	34	21.79

Fig. 2.

Representative Hematoxylin and Eosin (H&E)-Stained Sections of OSCC Tissues. A Well-differentiated OSCC (gingiva) showing organized squamous cell nests with keratin pearl formation. B Moderately differentiated OSCC (buccal) with irregular cell borders and nuclear atypia. C Poorly differentiated OSCC (floor of mouth) characterized by epithelioid cells arranged in cord-like structures, mitotic figures, and prominent stromal fibrous tissue hyperplasia. Sections were reviewed by a senior pathologist to confirm tumor cellularity (≥ 30%) and exclude inflammation/necrosis

Methylation status of SOX1 and PAX1 in tumor tissues vs. adjacent normal mucosa

ROC curve analysis revealed that methylated SOX1 and PAX1 conferred substantial diagnostic utility in the OSCC cohort relative to the normal control group, with respective area under the curve (AUC) values of 0.6541 and 0.8006(Figure 3A). Scatter plots of ΔCt values showed clear separation between OSCC and control groups for both genes. The predefined cutoff ΔCt values (10.62 for SOX1^m, 8.95 for PAX1^m) effectively stratified the two cohorts while ensuring favorable specificity and sensitivity(Figure 3B/C).

Fig. 3.

Diagnostic performance of methylated SOX1 and PAX1 in OSCC. A ROC curves for methylated SOX1 (red) and PAX1 (blue) in distinguishing OSCC from control samples. The corresponding AUC values are 0.6541 (SOX1^m) and 0.8006 (PAX1^m). B, C Scatter plots of ΔCt values (Ct _{reference gene} − Ct _{target gene}) for SOX1^m(B) and PAX1^m(C). Red points represent OSCC samples; blue points represent control samples

Q-MSP analysis demonstrated significantly higher mean methylation indices (M-index) for SOX1 and PAX1 in OSCC tissues compared to adjacent normal mucosa (P = 0.039 and P < 0.0001, respectively; Fig. 4). The mean M-index values were 273.3 ± 790.2 (tumor) vs. 93.6 ± 263.2 (normal) for SOX1, and 720.9 ± 1389.6 (tumor) vs. 108.5 ± 372.2 (normal) for PAX1. Methylation positivity rates (defined by Ct threshold) were 56.7% (93/164) for SOX1 and 73.8% (121/164) for PAX1 in tumors, versus 26.1% (23/88) in normal tissues for both genes (Table 2).

Fig. 4.

Boxplots comparing mean M-index values between OSCC tumor tissues (n = 164) and adjacent normal mucosa (n = 88). A SOX1 M-index: tumor = 273.3 ± 790.2 vs. normal = 93.6 ± 263.2, P = 0.039 (independent samples t-test). B PAX1 M-index: tumor = 720.9 ± 1389.6 vs. normal = 108.5 ± 372.2, P < 0.0001 (independent samples t-test). Boxes represent interquartile range (IQR), with median lines; whiskers denote 1.5×IQR. Outliers shown as individual dots

Table 2.

Methylation Status of SOX1 and PAX1 in OSCC Tumor Tissues vs. Adjacent Normal Mucosa

Gene		Oral tissue site		P-value
		Adjacent Normal	Tumor
SOX1	M-index	93.57±263.21	273.32±790.16	0.039
	Positive rate	26.14%(23/88)	56.71%(93/164)
PAX1	M-index	108.52±372.16	720.92±1389.62	<0.0001
	Positive rate	26.14%(23/88)	73.78%（121/164）

Diagnostic performance of SOX1 and PAX1 methylation

Single-gene methylation analysis showed SOX1 methylation (SOX1^m) had a sensitivity of 56.7% and specificity of 80.7% for OSCC detection, with an odds ratio (OR) of 5.47 (95% CI: 2.96–10.10, P < 0.0001). PAX1 methylation (PAX1^m) exhibited higher sensitivity (73.8%) but equivalent specificity (80.7%), with an OR of 11.75 (95% CI: 6.24–22.14, P < 0.0001). Combined detection (positivity in either genes) increased sensitivity to 81.7% but reduced specificity to 63.6%, with an OR of 7.81 (95% CI: 4.34–14.07, P < 0.0001; Table 3).

Table 3.

Diagnostic Performance of SOX1 and PAX1 Methylation for OSCC Detection

	Tumor tissues	Adjacent normal tissues	Sensitivity	Specificity	OR	95%CI	P
SOX1 m,n(%)
Positive	93（56.7）	17（19.3）	56.71%	80.68%	5.47	2.96-10.10	＜0.0001
Negative	71（43.3）	71（80.7）
PAX1 m,n(%)
Positive	121(73.8)	17(19.3)	73.78%	80.68%	11.75	6.24-22.14	＜0.0001
Negative	43(26.2)	71(80.7)
PAX1 m + SOX1m,n(%)
Positive	134(81.7)	32(36.4)	81.71%	63.64%	7.81	4.34-14.07	＜0.0001
Negative	30(18.3)	56(63.6)

Association of methylation with clinicopathological features

Univariate analysis revealed SOX1^m was significantly associated with advanced tumor size (T stage, χ² = 8.04, P = 0.045), lymph node metastasis (N stage, χ²= 4.27, P = 0.039), and higher clinical stage (χ²=8.33, P = 0.040). No significant associations were observed between PAX1^m and T stage, N stage, or clinical stage. Neither methylation marker correlated with age, gender, tumor location, smoking, alcohol use, fibrous stroma(for assessment of tumor microenvironment status), vascular invasion including blood vessel and lymphatic vessel invasion (for assessment of potential metastatic risk of tumor cells), or histological differentiation (Table 4).

Table 4.

Association of SOX1 and PAX1 Methylation with Clinicopathological Features

Variable	SOX1 positive		SOX1 negative				PAX1 positive		PAX1 negative
	NO.	%	NO.	%	χ2	𝑃 value	NO.	%	NO.	%	χ2	𝑃 value
Gender					2.05	0.15					3.55	0.06
Male	65	60.7	42	39.3			84	78.5	23	21.5
Female	28	49.1	29	50.9			37	64.9	20	35.1
Age					3.01	0.22					1.43	0.49
<50	8	40.0	12	60.0			14	70.0	6	30.0
50—69	55	61.1	35	38.9			64	71.1	26	28.9
≥70	30	55.6	24	44.4			43	79.6	11	20.4
Primary site					4.47	0.48					3.28	0.66
oropharynx	17	68.0	8	32.0			20	80.0	5	20.0
Tongue	23	50.0	23	50.0			30	65.2	16	34.8
lip	13	65.0	7	35.0			14	70.0	6	30.0
buccal	11	45.8	13	54.2			19	79.2	5	20.8
gingiva	19	55.9	15	44.1			27	79.4	7	20.6
floor of mouth	10	66.7	5	33.3			11	73.3	4	26.7
History of smoking					1.02	0.31					1.22	0.27
YES	40	61.5	25	38.5			51	78.5	14	21.5
NO	53	53.5	46	46.5			70	70.7	29	29.3
History of alcohol					0.26	0.61					0.70	0.40
Yes	39	59.1	27	40.9			51	77.3	15	22.7
No	54	55.1	44	44.9			70	71.4	28	28.6
Fibrous stroma					0.08	0.79					0.85	0.36
Yes	69	56.1	54	43.9			93	75.6	30	24.4
No	24	58.5	17	41.5			28	68.3	13	31.7
Vascular invasion					0.002	0.97					1.03	0.31
Yes	9	56.2	7	43.8			14	87.5	2	12.5
No	84	56.8	64	43.2			107	72.3	41	27.7
Differentiation					0.95	0.62					3.79	0.15
High	43	53.1	38	46.9			62	76.5	19	23.5
Moderate	45	60.8	29	39.2			55	74.3	19	25.7
Low	5	55.6	4	44.4			4	44.4	5	55.6
Pathological T category					8.04	0.045					2.98	0.40
T1	26	42.6	35	57.4			42	68.9	19	31.1
T2	39	67.2	19	32.8			41	70.7	17	29.3
T3	17	63.0	10	37.0			23	85.2	4	14.8
T4	6	60.0	4	40.0			8	80.0	2	20.0
Pathological N category					4.27	0.039					0.005	0.94
N0	43	53.8	37	46.3			60	75.0	20	25.0
N1-N2	30	73.2	11	26.8			31	75.6	10	24.4
Tumour stage					8.33	0.040					1.81	0.61
I	12	38.7	19	61.3			21	67.7	10	32.3
II	18	66.7	9	33.3			22	81.5	5	18.5
III	19	65.5	10	34.5			23	79.3	6	20.7
IV	24	70.6	10	29.4			25	73.5	9	26.5

Survival outcomes

Over a median follow-up of 48 months (range: 6–52 months), SOX1^m-positive patients exhibited significantly shorter DFS compared to SOX1^m-negative patients (median DFS: 28 vs. 40 months; log-rank P = 0.03; Fig. 5A), with a hazard ratio (HR) of 0.53 (95% CI: 0.31–0.92). No significant difference in overall survival (OS) was observed between SOX1^m groups (HR = 0.61, 95% CI: 0.29–1.30, P = 0.22; Fig. 5C). PAX1^m status did not predict DFS (HR = 1.01, P = 0.97) or OS (HR = 0.92, P = 0.86). Multivariate Cox regression for disease-free survival (DFS) showed that age (HR = 3.60, 95% CI: 1.50–8.61, P = 0.004), smoking (HR = 4.26, 95% CI: 1.08–16.73, P = 0.038), and clinical stage (HR = 4.54, 95% CI: 1.15–17.96, P = 0.031) were independent prognostic factors for worse DFS. For overall survival (OS), age (HR = 5.46, 95% CI: 1.47–20.20, P = 0.011), smoking (HR = 5.61, 95% CI: 1.01–31.17, P = 0.049), and drinking (HR = 0.14, 95% CI: 0.03–0.69, P = 0.015) were significant predictors. Drinking was associated with better OS, while older age and smoking correlated with worse OS. (Fig. 6).

Fig. 5.

Disease-free survival (DFS) and overall survival (OS) stratified by SOX1 and PAX1 methylation positivity (SOX1^m+ vs. SOX1^m−; PAX1^m+ vs. PAX1^m−;n = 164). A DFS: median survival 28 months (SOX1^m+) vs. 40 months (SOX1^m−), log-rank P = 0.03, hazard ratio (HR) = 0.53 (95% CI: 0.31–0.92). B DFS: PAX1^m+ vs. PAX1^m−, no significant difference (log-rank P = 0.97, HR = 1.01, 95% CI: 0.55–1.87). C OS: SOX1^m+ vs. SOX1^m−, no significant difference (log-rank P = 0.22, HR = 0.61, 95% CI: 0.29–1.30). D OS: PAX1^m+ vs. PAX1^m−, no significant difference (log-rank P = 0.86, HR = 0.92, 95% CI: 0.40–2.14)

Fig. 6.

Forest plots of multivariate Cox regression for DFS(A) and OS(B). Variables include age (<60 vs. ≥60 years, n = 164), Grade status (high/moderate vs. low differentiation, n = 164), T stage (T3/T4 vs. T1/T2, n = 156), N stage (N + vs. N0, n = 121), clinical stage (III/IV vs. I/II, n = 121), and SOX1^m/PAX1^m status (positive vs. negative, n = 164)

Discussion

Aberrant DNA methylation, a cardinal epigenetic hallmark of carcinogenesis, drives tumor progression by dysregulating gene expression through promoter hypermethylation of tumor suppressor genes or hypomethylation of oncogenes [21]. While extensively characterized in colorectal, cervical, and hepatic cancers, the methylomic landscape of OSCC remains poorly understood, limiting the discovery of epigenetic biomarkers for early diagnosis and prognosis. This study investigated the status of SOX1^m and PAX1^m(two transcription factor genes with critical roles in mucosal development) in OSCC tissues, linking their epigenetic alterations to clinicopathological features and survival outcomes.

A recent bioinformatics analysis based on TCGA datasets identified SOX1, SFRP4, IRF4, and PCDH17 among the top 200 differentially methylated regions (DMRs) in HNSCC, with cfDNA-based DMR analysis supporting their potential as circulating biomarkers [22]. Our validation revealed significantly elevated mean M-index values and positive rates of SOX1^m and PAX1^m in OSCC tumor tissues compared to adjacent normal mucosa. These findings align with prior reports showing frequent hypermethylation of PAX1 in both primary and recurrent OSCC specimens [23]. However, studies specifically addressing SOX1^m in oral diseases are sparse. Only one study analyzing exfoliated oral cells from 31 OSCC patients reported significantly higher SOX1 M-index levels compared to controls [24]. Similarly, a separate study observed an upward trend in SOX1^m in patients with oral epithelial dysplasia [25].Nevertheless, the functional implications of SOX1^m in OSCC progression and patient prognosis remain unexplored.

To address this, we performed the first analysis correlating SOX1^m with pathological features and survival outcomes in OSCC. SOX1^m positivity was significantly associated with larger tumor size, nodal metastasis, and advanced clinical stage, suggesting a role for SOX1^m in promoting tumor aggressiveness.In breast cancer, SOX1 has been shown to suppress tumorigenesis by downregulating β-catenin, cyclin D1, and c-myc expression, while ectopic expression of SOX1 in vitro attenuates proliferation, invasion, and enhances apoptosis [13].Our findings indicate that SOX1 functions as a tumor suppressor gene in OSCC, and that its epigenetic silencing via promoter methylation may contribute to enhanced malignancy. The specific mechanism warrants further investigation.

Previous studies on PAX1^m in OSCC yielded mixed results. While Sun et al. reported a correlation between PAX1 M-index and T/clinical stage in 135 OSCC patients [23], Cheng et al. detected oral exfoliated cells from 80 OSCC patients and found that the higher M-index of PAX1 was significantly correlated with a larger tumor size, but had no association with the clinical stage. Additionally, the positive rate of PAX1^m was also not associated with the T stage or the clinical stage [19], Our data show no significant correlation between the positive rate of PAX1^m and T stage, N stage, or clinical stage, underscoring the need for larger cohort studies to clarify its clinical relevance.

Tumor heterogeneity reflected the inherent diversity of genetic and phenotypic characteristics, both within individual tumors and between distinct tumors [26]. Recent studies have demonstrated that conditioned medium derived from cancer-associated fibroblasts enhanced the methylation levels and migratory capacity of OSCC cells, highlighting the interplay between the tumor microenvironment and epigenetic modifications [27]. Xinyu Li et al. also revealed that how methylation-driven dysregulation varied across distinct cell populations within the tumor microenvironment [28]. Although our data indicated that PAX1^m and SOX1^m exhibited no significant correlation with fibrous stroma or vascular invasion, the association between their methylation status and tumor heterogeneity warranted further investigation and exploration.

Risk factor analysis identified both SOX1^m and PAX1^m as independent risk factors for OSCC, supporting their involvement in tumorigenesis. Identifying functional methylation events is critical for improving early diagnosis, prognostication, and therapeutic strategies in cancer [29, 30].Future mechanistic studies on SOX1^m and PAX1^m may inform novel prevention and treatment approaches for OSCC.

In a four-year longitudinal follow-up of 164 patients, we observed that SOX1^m-positive individuals had significantly shorter DFS, while no significant difference in overall survival (OS) was noted. PAX1 methylation status did not significantly impact either DFS or OS. These findings emphasize the need to closely monitor disease progression in SOX1^m-positive patients. Prognostic factors associated with poor outcomes in our cohort included older age, smoking and higher clinical stage, consistent with previous reports [31, 32].

Furthermore, epigenetic therapies hold considerable promise for the precision management of OSCC. The integration of targeted epigenetic agents with advanced drug delivery systems is poised to expedite clinical translation, whereas the potential of SOX1^m and PAX1^m to support early detection, risk stratification, and personalized therapy warrants further investigation.

However, this study has several limitations. Firstly, this was a single-center retrospective cohort study where patients were recruited from a concentrated source. Factors such as age, gender, tumor location, and treatment regimens may exhibit regional/hospital-specific characteristics, leading to limited sample representativeness. Secondly, multivariate Cox regression only adjusted for a limited number of variables, such as age, smoking status, and clinical stage, but did not incorporate established prognostic factors for OSCC, including surgical margin status and the presence of perineural invasion.

Prospective validation in multi-center cohorts (e.g., TCGA-HNSCC, GECCO) will solidify the clinical relevance of SOX1^m and PAX1^m. Integrative multi-omics analyses—combining DNA methylation, RNA expression, and proteomics—will unravel the functional networks disrupted by their silencing. Mechanistic studies should explore whether SOX1^m drives epithelial-mesenchymal transition (EMT) or immune evasion in OSCC, while preclinical models testing demethylating agents (e.g., decitabine) may inform targeted epigenetic therapies.

Conclusion

This study established SOX1^m and PAX1^m as frequent epigenetic events in OSCC, with SOX1^m uniquely linked to aggressive disease and poorer DFS. These findings highlighted the potential of epigenetic biomarkers to enhance risk stratification and guide personalized treatment strategies. Translating these insights into clinical practice requires large-scale validation and functional characterization, which may unlock new avenues for early intervention in this deadly malignancy.

Abbreviations

OSCC

Oral squamous cell carcinoma

Q-MSP

Quantitative methylation-specific PCR

FFPE

Formalin-fixed paraffin-embedded

DFS

Disease-free survival

OS

Overall survival

SOX

Sex-determining region Y-box

PAX

Paired box

DMRs

Differentially methylated regions

Authors’ contributions

YDL: Writing – review & editing, Writing – original draft, Formal analysis, Data curation. XYL: Formal analysis, Data curation. LWS: Investigation, Conceptualization, Methodology. MZY: Data curation.AJL: Validation, Supervision. ZRL: Conceptualization, Writing – review & editing. All authors read and approved the final manuscript.

Funding

This work was supported by the 2023 Innovation Cultivation Fund of the Seventh Medical Center of the General Hospital of the People’s Liberation Army (NO.qzx-2023-13).

Data availability

The data presented in this study are available on request from the corresponding author.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of the seventh Medical Center of Chinese PLA General Hospital (Approval No: S2025-054-01), conducted in accordance with the Declaration of Helsinki, and written informed consent was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.