Translate this page into:
Non-invasive detection of endometrial cancer through exfoliated cervical cell gene methylation signatures: Evaluation of CELF4/GALR1/ZNF486 methylation biomarkers
-
Received: ,
Accepted: ,
How to cite this article: Fan X, Mao S, Ou Y, Yu F, Wang X, Wang R. Non-invasive detection of endometrial cancer through exfoliated cervical cell gene methylation signatures: Evaluation of CELF4/GALR1/ZNF486 methylation biomarkers. 2026;23:34. CytoJournal. doi: 10.25259/Cytojournal_135_2025
Abstract
Objective:
The current non-invasive screening methods for endometrial cancer (EC), such as transvaginal ultrasound (TVUS), lack sufficient specificity, leading to unnecessary invasive diagnostic procedures. The detection of cancer-specific deoxyribonucleic acid (DNA) methylation alterations in cervical cytology samples presents a promising, minimally invasive alternative. This study aimed to evaluate the diagnostic potential of novel DNA methylation biomarkers (CUGBP Elav-like family member 4 methylation [CELF4m], Galanin receptor 1 methylation [GALR1m], Zinc finger protein 486 methylation [ZNF486m]) in cervical exfoliated cells for EC screening.
Material and Methods:
This case–control study enrolled patients scheduled for diagnostic or surgical curettage at the hospital from September 2024 to June 2025. A total of 104 women underwent endometrial evaluation (32 type I EC, 3 type II EC, 61 benign lesions [BL], 8 atypical hyperplasia). Quantitative methylation-specific polymerase chain reaction assessed biomarker performance against histopathological diagnosis.
Results:
All three markers (CELF4m, GALR1m, ZNF486m) showed significant differential ΔCp values between BL and EC groups (all P < 0.01), with moderate diagnostic accuracy (area under the ROC curve [AUCs]: 0.654-0.704). The single-marker GALR1m outperformed endometrial thickness (ET) measured by TVUS in specificity (60.9% vs. 27.5%, P < 0.001) while maintaining comparable sensitivity (80.0% vs. 74.3%, P = 0.527). A two-marker panel (GALR1m/ZNF486m, Model 1) achieved sensitivity (97.1% [95% confidence interval [CI]: 91.6–100%]) with modest specificity (49.3% [95% CI: 37.5-61.1%]). Among three-gene models, Model 3 demonstrated the highest overall performance (AUC: 0.733 [95% CI: 0.657-0.808]; sensitivity: 91.4% [95% CI: 82.1-100%]), whereas Model 4 balanced sensitivity (80.0% [95% CI: 66.7-93.3%]) and improved specificity (63.8% [95% CI: 52.4-75.1%]).
Conclusion:
DNA methylation signatures in cervical cells show superior diagnostic accuracy to conventional ET measured by TVUS, particularly through multi-marker panels. This non-invasive approach represents a promising EC screening strategy.
Keywords
Deoxyribonucleic acid methylation
Endometrial cancer
Endometrial thickness
Screening
Transvaginal ultrasound
INTRODUCTION
Endometrial cancer (EC) has become the most common malignancy of the female reproductive tract in developed countries, with continuously rising incidence and mortality rates.[1] In high-income countries such as the United States, EC is also the most common gynecologic cancer, with an estimated 69,120 new cases and 13,860 deaths projected in 2025.[2] Similarly, in Europe, EC accounts for approximately 130,000 new diagnoses annually, making it the fourth most common cancer among women.[3] In 2022, China reported 77,700 new EC cases and 13,500 deaths.[4] The increasing incidence of EC is associated with factors such as elevated body mass index (BMI), advanced age, early menarche, late menopause, metabolic syndrome, family history, and endogenous or exogenous estrogen exposure.[5,6] With China’s rapid socioeconomic development, the risk of metabolic syndrome due to high-fat, high-sugar diets has risen. In addition, the widespread lack of oral contraceptive use among Chinese women, combined with multiple risk factors and insufficient protective measures, has led to a year-by-year increase in EC incidence and a trend toward younger onset ages.[4]
EC is classified into Type I and Type II. Type I, also known as “estrogen-dependent” EC, accounts for over 80% of cases and is thought to arise from prolonged unopposed estrogen exposure. Patients are typically younger, often presenting with a history of irregular vaginal bleeding, and frequently develop from atypical hyperplasia (AH).[7] Transabdominal or transvaginal ultrasonography may reveal abnormal endometrial echoes or intrauterine masses. Definitive diagnosis requires histopathological examination of endometrial tissue obtained through diagnostic curettage. However, relying solely on endometrial thickness (ET) measurement has limitations due to physiological hormonal variations. Furthermore, while Type I EC is often linked to endometrial thickening and hyperplasia, Type II (estrogen-independent EC) usually occurs independently of hyperplasia.[8] Accurately diagnosing EC in high-risk populations, such as those with chronic menstrual irregularities, recurrent abnormal uterine bleeding (AUB), or suspected endometrial lesions, and determining the need for invasive pathological examinations remains challenging.
The development and progression of EC involve multiple biological processes and numerous genes. Deoxyribonucleic acid (DNA) methylation plays a crucial role in regulating gene expression and controlling other genes.[9] Thus, DNA methylation testing holds promise as a novel method for early EC screening. Previous studies have confirmed that methylation status in multiple genes, including BHLHE22, CDO1, CELF4, ZNF454, GHSR, SST, ZIC1, PCDHGB7, and ZSCAN12, can serve as effective diagnostic biomarkers for EC.[9-17] In pre-menopausal women with AUB, CDO1m/ CUGBP Elav-like family member 4 methylation (CELF4m) methylation testing in exfoliated cervical cells demonstrated superior diagnostic performance compared to BMI and ET, with sensitivity and specificity of 85.7% and 87.6%, respectively.[11] Meanwhile, the ZSCAN12m/GYPCm panel showed a sensitivity of 90.0% and a positive predictive value ranging from 25.6% to 50.0% in post-menopausal women with AUB.[15] These findings provide proof-of-concept for DNA methylation-based EC screening.
In this case-control study, we evaluated the potential of novel DNA methylation biomarkers, CELF4m, Galanin receptor 1 methylation (GALR1m), and Zinc finger protein 486 methylation (ZNF486m), and their combined panels for EC diagnosis using exfoliated cervical cells from women undergoing endometrial assessment.
MATERIAL AND METHODS
Study design and sample collection
Female participants (104, age range: 24-86 years) were enrolled in our case-control study from September 2024 to June 2025. Inclusion criteria: Women underwent endometrial curettage or surgery at the hospital due to abnormal or dysfunctional uterine bleeding, abnormal transvaginal ultrasound (TVUS) findings, or abnormal findings on digital examination. Exclusion criteria: Unwillingness to sign the informed consent form or transfer to another hospital for further diagnosis and treatment.
This study was approved by the Research and Clinical Trial Ethics Committee of the Changsha Hospital for Maternal and Child Health Care (Approval No.: EC-20240913-01) and was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants.
Before endometrial curettage or surgery, exfoliated cervical cells were collected, preserved in cell storage solution (Hologic, Bedford, MA, USA), and subsequently processed for DNA extraction before storage at -80°C. Following the procedure, clinical data, including age, histologic type of the lesion or tumor, and International Federation of Gynaecology and Obstetrics (FIGO) stage, were retrieved from hospital records.
DNA methylation assay
DNA methylation analysis was performed without prior knowledge of pathological and clinical information. Exfoliated cervical cell DNA 500 ng (volume ≤50 μL) was added to a 0.2 mL polymerase chain reaction (PCR) tube, followed by 100 μL of bisulfite conversion reagent. Following vortex mixing and centrifugation, incubate under the following conditions: 98°C for 10 min, then 64°C for 150 min, store the product at 4°C.
After purification, 2 μL of the converted DNA was used for quantitative real-time PCR (Roche Applied Science, CA, USA) with the GAPDH gene as the internal control. Primer and probe sequences are listed in Supplementary Table 1.
The inter- and intra-batch coefficients of variation for the reagents used were both <5%. The PCR protocol was as follows: Initial denaturation at 95°C for 10 min, followed by 50 cycles of denaturation at 95°C for 10 s, and combined annealing/extension at 58°C with fluorescence acquisition. Samples were then held at 20°C. A Cp(GAPDH) value ≤35 indicated quality control compliance. Gene methylation levels were expressed as ΔCp, calculated as ΔCp(gene) = Cp(gene) - Cp(GAPDH). When Cp(gene) was undetectable, a value of 50 was assigned. The lower the Δ Cp value, the higher the methylation level; The higher the Δ Cp value, the lower the methylation level. Here, Cp refers to the crossing point, i.e., the number of cycles experienced when the fluorescence signal intensity of the sample reaches the preset threshold during the PCR process.
Statistical analysis
Sample size estimation was performed using PASS 15 Power Analysis and Sample Size Software (NCSS, LLC., Kaysville, Utah, USA). A sample of 35 from the EC achieves 85% power to detect a difference of 0.2 between the area under the receiver operating characteristic curve (ROC) area under the ROC curve (AUC) under the null hypothesis of 0.5 and an AUC under the alternative hypothesis of 0.7 using a two-sided z-test at a significance level of 0.050.
All analyses were performed using R (version 4.5.1). A two-tailed P < 0.05 was considered statistically significant. The normality of continuous variables was assessed using the Shapiro–Wilk test. The methylation values (ΔCp) of CELF4, GALR1, and ZNF486 all had P < 0.05, indicating deviation from a normal distribution. These data are reported as median [interquartile range (IQR)]. Categorical variables are reported as frequency (%) (tableone::CreateTableOne). Differences in medians between groups were assessed using the Mann-Whitney U test (stats::wilcox.test). The distribution of DNA methylation was displayed using boxplots (ggpubr::ggviolin). Four diagnostic models for detecting EC were constructed using logistic regression. Details regarding variable selection for model construction are provided in Supplementary Tables 2-5. Model 1, a logistic regression model was constructed using two predictor variables: GALR1m + ZNF486m; Model 2, a logistic regression model was constructed using two predictor variables: ΔCp(CELF4m) + ΔCp(GALR1m); Model 3, a logistic regression model was constructed using three predictor variables: CELF4m + GALR1m + ZNF486m; Model 4, a logistic regression model was constructed using three predictor variables: ΔCp(CELF4m) + ΔCp(GALR1m) + ΔCp(ZNF486m). Missing data on ET were filled in using multiple imputations (mice::mice [method = “pmm”]). The ability of each indicator to detect EC or AH/EC was demonstrated using ROC curves (pROC::ggroc). The optimal cut-off values for DNA methylation and the models were determined by maximizing the Youden index (pROC::coords [best.method = “youden”]). The cut-offs for CELF4m, GALR1m, and ZNF486m were 9.02, 12.07, and 6.50, respectively, ΔCp values less than or equal to the cutoff were considered positive, while values greater than the cut-off were considered negative. The cut-offs for Model 1-4 were 0.17, 0.39, 0.19, and 0.28, respectively. The probability greater than the cut-off was considered positive, while the probability less than or equal to the cut-off was considered negative. “A/B” denotes a negative result only if both A and B are negative; if either A or B is positive, the final result is considered positive. Taking the pathological results as “gold standard,” the AUC, sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio for each indicator were calculated (DTComPair::acc.paired) and compared with those of ET (DTComPair::sesp.mcnemar).
RESULTS
Participant characteristics
A total of 104 participants were included in the study, including 61 with benign lesions (BL), 8 with AH, and 35 with EC. Among the EC patients, 2 had clear cell carcinoma, and 1 had sarcoma; 16 were classified as Stage I according to the FIGO staging system, 3 as Stage II, 6 as Stage III, and 10 had unknown staging. The median age was 50.5 (IQR::41.0, 58.0), with 50% (52) being over 50 years old. The median ET across all participants was 11.0 (IQR: 8.4, 15.4) mm; the median ΔCp values for CELF4m, GALR1m, and ZNF486m were 7.9 (IQR: 5.0, 16.4), 10.8 (IQR: 6.9, 17.1), and 7.1 (IQR: 2.6, 17.1), respectively. Other characteristics are detailed in Table 1.
| Variable | Benign lesion | Atypical hyperplasia | Endometrial cancer | Total |
|---|---|---|---|---|
| Number of cases | 61 | 8 | 35 | 104 |
| Age | ||||
| Years, median (IQR) | 45.0 (37.0, 51.0) | 50.5 (49.0, 56.0) | 59.0 (53.5, 65.0) | 50.5 (41.0, 58.0) |
| ≤50, n(%) | 43 (70.5) | 4 (50.0) | 5 (14.3) | 52 (50.0) |
| >50, n(%) | 18 (29.5) | 4 (50.0) | 30 (85.7) | 52 (50.0) |
| Menopause | ||||
| Pre-menopause, n(%) | 49 (80.3) | 2 (25.0) | 6 (17.1) | 57 (54.8) |
| Post-menopause, n(%) | 12 (19.7) | 6 (75.0) | 29 (82.9) | 47 (45.2) |
| Endometrial thickness | ||||
| mm, median (IQR) | 11.8 (8.6, 15.0) | 9.0 (7.6, 9.9) | 12.2 (9.0, 18.9) | 11.0 (8.4, 15.4) |
| Un-thickened, n(%) | 16 (26.2) | 2 (25.0) | 4 (11.4) | 22 (21.2) |
| Thickened#, n(%) | 38 (62.3) | 5 (62.5) | 13 (37.1) | 56 (53.8) |
| Unknown | 7 (11.5) | 1 (12.5) | 18 (51.4) | 26 (25.0) |
| Specimen of pathology | ||||
| Curettage | 58 (95.1) | 2 (25.0) | 6 (17.1) | 66 (63.5) |
| Surgery | 3 (4.9) | 6 (75.0) | 29 (82.9) | 38 (36.5) |
| Histotypes of cancer | ||||
| Adenocarcinoma, n(%) | - | - | 32 (91.4) | 32 (91.4) |
| Clear cell carcinoma, n(%) | - | - | 2 (5.7) | 2 (5.7) |
| Sarcoma, n(%) | - | - | 1 (2.8) | 1 (2.8) |
| FIGO stage of cancer | ||||
| I, n(%) | - | - | 16 (45.7) | 16 (45.7) |
| II, n(%) | - | - | 3 (8.6) | 3 (8.6) |
| III, n(%) | - | - | 6 (17.1) | 6 (17.1) |
| IV, n(%) | - | - | 0 (0.0) | 0 (0.0) |
| Unknown, n (%) | - | - | 10 (28.6) | 10 (28.6) |
| CELF4m | ||||
| ΔCp, median (IQR) | 9.2 (5.7, 17.8) | 8.5 (6.3, 15.0) | 6.2 (3.2, 8.6) | 7.9 (5.0, 16.4) |
| CELF4m(+), n(%) | 30 (49.2) | 4 (50.0) | 28 (80.0) | 62 (59.6) |
| GALR1m | ||||
| ΔCp, median (IQR) | 14.2 (7.9, 17.8) | 9.2 (6.8, 19.9) | 7.3 (3.0, 10.9) | 10.8 (6.9, 17.1) |
| GALR1m(+), n(%) | 22 (36.1) | 5 (62.5) | 28 (80.0) | 55 (52.9) |
| ZNF486m | ||||
| ΔCp, median (IQR) | 13.5 (4.9, 18.1) | 5.4 (3.7, 12.0) | 3.9 (1.3, 9.5) | 7.1 (2.6, 17.1) |
| ZNF486m(+), n(%) | 20 (32.8) | 5 (62.5) | 25 (71.4) | 50 (48.1) |
Distribution of DNA methylation
The ΔCp median of CELF4m, GALR1m, and ZNF486m showed statistically significant differences between BL and EC (all P < 0.01), whereas no significant differences were observed between BL and AH or between AH and EC (all P > 0.05; [Figure 1a-c]). Figure 1d-f shows the ROC curves of CELF4m, GALR1m, and ZNF486m for detecting EC, displaying their AUC values, optimal cut-off points, sensitivity, and specificity.
For GALR1m and ZNF486m, statistically significant ΔCp median differences were found when stratifying by age (≤50 vs. >50 years), menopausal status (pre- vs. post-menopause), and sample type (curettage vs. surgical specimens) (all P < 0.05; [Supplementary Figures 1-3]). However, CELF4 methylation differed significantly between pre- and post-menopausal groups (P = 0.041; [Supplementary Figure 2a]). CELF4m and GALR1m exhibited no significant associations with ET (all P > 0.05; [Supplementary Figure 4a and 4b]). A difference in ZNF486 methylation was found between the un-thickened and unknown ET groups (P = 0.025; [Supplementary Figure 4c]). CELF4m, GALR1m, and ZNF486m exhibited no significant associations with FIGO stage (all P > 0.05; [Supplementary Figure 5]).
In one endometrial sarcoma case, all three markers (CELF4m, GALR1m, and ZNF486m) tested positive, with ΔCp values of 5.39, 7.26, and 6.37, respectively [Figure 1]. Two endometrial clear cell carcinoma cases, ZNF486m was negative in both cases [Figure 1c], while CELF4m and GALR1m showed positive results [Figure 1a and 1b].
Clinical performance of indicators
For detecting EC, AUCs of the single-gene methylation markers were greater than ET assessment by TVUS. GALR1m demonstrated the highest AUC of 0.704 (95% CI: 0.616-0.793), with a sensitivity of 80.0% (95% CI: 66.7-93.3%) and specificity of 60.9% (95% CI: 49.3-72.4%) [Table 2]. Figure 2 shows the ROC curves for Models 1-4. The AUC for detecting EC ranges from 0.753 to 0.771, while for detecting AH/EC, it ranges from 0.722 to 0.768. Among the two-gene methylation panels, GALR1m/ZNF486m and Model 1 exhibited the highest AUC, sensitivity, and specificity, with values of 0.732 (95% CI: 0.666-0.798), 97.1% (95% CI: 91.6-100.0%), and 49.3% (95% CI: 37.5-61.1%), respectively. For the three-gene methylation panels, Model 3 showed superior performance with an AUC of 0.733 (95% CI: 0.657-0.808) and sensitivity of 91.4% (95% CI: 82.1-100.0%). In contrast, Model 4 maintained a relatively high sensitivity (80.0%, 95% CI: 66.7-93.3%) while achieving improved specificity (63.8%, 95% CI: 52.4-75.1%).
| Indicator | AUC (95%CI) | P AUC | Sensitivity (95%CI) | P Sensitivity | Specificity (95%CI) | PSpecificity | LR+ (95%CI) | LR- (95%CI) |
|---|---|---|---|---|---|---|---|---|
| For detecting EC (n=35) in all cases (n=104) | ||||||||
| Thickened endometrium | 0.509 (0.418-0.600) | Ref. | 74.3 (59.8-88.8) | Ref. | 27.5 (17.0-38.1) | Ref. | 1.03 (0.80-1.31) | 0.93 (0.47-1.84) |
| CELF4m | 0.654 (0.564-0.743) | 0.017 | 80.0 (66.7-93.3) | 0.527 | 50.7 (38.9-62.5) | 0.006 | 1.62 (1.21-2.17) | 0.39 (0.19-0.80) |
| GALR1m | 0.704 (0.616-0.793) | 0.002 | 80.0 (66.7-93.3) | 0.527 | 60.9 (49.3-72.4) | <0.001 | 2.04 (1.46-2.87) | 0.33 (0.16-0.65) |
| ZNF486m | 0.676 (0.581-0.771) | 0.004 | 71.4 (56.5-86.4) | 0.739 | 63.8 (52.4-75.1) | <0.001 | 1.97 (1.35-2.87) | 0.45 (0.26-0.78) |
| CELF4m/GALR1m | 0.646 (0.571-0.720) | 0.018 | 91.4 (82.1-100.0) | 0.034 | 37.7 (26.2-49.1) | 0.237 | 1.47 (1.19-1.81) | 0.23 (0.07-0.70) |
| CELF4m/ZNF486m | 0.638 (0.564-0.712) | 0.026 | 91.4 (82.1-100.0) | 0.058 | 36.2 (24.9-47.6) | 0.257 | 1.43 (1.17-1.76) | 0.24 (0.08-0.73) |
| GALR1m/ZNF486m | 0.732 (0.666-0.798) | <0.001 | 97.1 (91.6-100.0) | 0.005 | 49.3 (37.5-61.1) | 0.014 | 1.92 (1.51-2.43) | 0.06 (0.01-0.41) |
| Model 1 (Prob>0.17) | 0.732 (0.666-0.798) | <0.001 | 97.1 (91.6-100.0) | 0.005 | 49.3 (37.5-61.1) | 0.014 | 1.92 (1.51-2.43) | 0.06 (0.01-0.41) |
| Model 2 (Prob>0.39) | 0.720 (0.627-0.812) | 0.001 | 71.4 (56.5-86.4) | 0.782 | 72.5 (61.9-83.0) | <0.001 | 2.59 (1.68-4.01) | 0.39 (0.23-0.68) |
| Model 3 (Prob>0.19) | 0.733 (0.657-0.808) | <0.001 | 91.4 (82.1-100.0) | 0.034 | 55.1 (43.3-66.8) | 0.003 | 2.04 (1.54-2.69) | 0.16 (0.05-0.57) |
| Model 4 (Prob>0.28) | 0.719 (0.631-0.807) | <0.001 | 80.0 (66.7-93.3) | 0.527 | 63.8 (52.4-75.1) | <0.001 | 2.21 (1.55-3.15) | 0.31 (0.16-0.62) |
| For detecting AH/EC (n=43) in all cases (n=104) | ||||||||
| Thickened endometrium | 0.492 (0.404-0.579) | Ref. | 72.1 (58.7-85.5) | Ref. | 26.2 (15.2-37.3) | Ref. | 0.98 (0.77-1.24) | 1.06 (0.56-2.02) |
| CELF4m | 0.626 (0.535-0.718) | 0.026 | 74.4 (61.4-87.5) | 0.782 | 50.8 (38.3-63.4) | 0.007 | 1.51 (1.11-2.06) | 0.50 (0.28-0.89) |
| GALR1m | 0.703 (0.615-0.792) | <0.001 | 76.7 (64.1-89.4) | 0.617 | 63.9 (51.9-76.0) | <0.001 | 2.13 (1.47-3.09) | 0.36 (0.20-0.65) |
| ZNF486m | 0.685 (0.594-0.776) | <0.001 | 69.8 (56.0-83.5) | 0.796 | 67.2 (55.4-79.0) | <0.001 | 2.13 (1.41-3.21) | 0.45 (0.28-0.73) |
| CELF4m/GALR1m | 0.639 (0.560-0.717) | 0.015 | 88.4 (78.8-98.0) | 0.052 | 39.3 (27.1-51.6) | 0.144 | 1.46 (1.16-1.83) | 0.30 (0.12-0.71) |
| CELF4m/ZNF486m | 0.650 (0.574-0.726) | 0.006 | 90.7 (82.0-99.4) | 0.033 | 39.3 (27.1-51.6) | 0.102 | 1.50 (1.20-1.87) | 0.24 (0.09-0.63) |
| GALR1m/ZNF486m | 0.727 (0.653-0.801) | <0.001 | 93.0 (85.4-100.0) | 0.013 | 52.5 (39.9-65.0) | 0.005 | 1.96 (1.48-2.58) | 0.13 (0.04-0.41) |
| Model 1 (Prob>0.17) | 0.727 (0.653-0.801) | <0.001 | 93.0 (85.4-100.0) | 0.013 | 52.5 (39.9-65.0) | 0.005 | 1.96 (1.48-2.58) | 0.13 (0.04-0.41) |
| Model 2 (Prob>0.39) | 0.675 (0.582-0.767) | 0.004 | 62.7 (48.3-77.2) | 0.346 | 72.1 (60.9-83.4) | <0.001 | 2.25 (1.41-3.59) | 0.52 (0.34-0.78) |
| Model 3 (Prob>0.19) | 0.717 (0.635-0.799) | <0.001 | 86.0 (75.7-96.4) | 0.109 | 57.4 (45.0-69.8) | 0.001 | 2.02 (1.47-2.77) | 0.24 (0.11-0.53) |
| Model 4 (Prob 0.28) | 0.700 (0.611-0.789) | <0.001 | 74.4 (61.4-87.5) | 0.796 | 65.6 (53.6-77.5) | <0.001 | 2.16 (1.47-3.19) | 0.39 (0.23-0.67) |
AH: Atypical hyperplasia, EC: Endometrial cancer, CI: Confidence interval, LR+: Positive likelihood ratio, LR: Negative likelihood ratio, AUC: Area under the ROC curve, Prob: Probability
A similar trend was observed in the detection of AH/EC [Table 2].
In the pre-menopausal cohort (n = 57, including 51 non-EC cases), all three markers (CELF4m, GALR1m, and ZNF486m) demonstrated higher specificity (>55%) compared to ET assessment (31.3%, 95% CI: 18.6-44.1%) by TVUS (all P <0.01; [Supplementary Table 6]).
DISCUSSION
Our study identified three DNA methylation biomarkers (CELF4m, GALR1m, ZNF486m) in exfoliated cervical cells that outperform ET for EC detection. All markers showed significantly higher AUCs (0.654–0.704) than ET (0.509, all P < 0.05). The GALR1m/ZNF486m panel achieved 97.1% sensitivity and 49.3% specificity (vs. ET’s 74.3%/27.5%), while three-gene combinations further improved specificity (55.1-63.8%) with maintained sensitivity (80.0-91.4%), demonstrating DNA methylation’s diagnostic potential for EC.
Previous studies have explored the use of exfoliated cells for cytomorphological analysis,[18] NGS detection of EC genes,[19] urine-derived exosomes for miRNAs,[20] and multi-omics testing.[21] However, these methods have limitations such as insufficient tumor cell abundance, stringent sample preservation requirements, subjective morphological interpretation, and high costs with complex procedures. We propose that DNA methylation, with its sample stability, simple detection process, and high throughput, represents a more ideal biomarker for EC.
In our study, TVUS of ET demonstrated relatively low sensitivity and specificity as an indicator. Using cutoff values of ≥11 mm for premenopausal and ≥5 mm for postmenopausal women, we obtained sensitivity and specificity of 74.3% and 27.5%, respectively. Previous studies have shown considerable variation in ET’s diagnostic performance, with one study reporting 96.2% sensitivity and 51.5% specificity for post-menopausal women using a >5 mm cut-off.[22] A recent study showed ET’s sensitivity and specificity at 72.5% and 43.9%, respectively.[17] These discrepancies may stem from different study populations, our cohort included both pre- and post-menopausal patients undergoing endometrial pathological diagnosis, where premenopausal hormonal factors and AUB could affect ET’s diagnostic efficacy.
The selected target genes CELF4 and GALR1 have established epigenetic associations with endometrial malignancies. Analysis of >27,000 CpG sites in normal endometrium (n = 23) and EC (n = 64) revealed GALR1 methylation as one of the most frequent epigenetic alterations in EC.[23] Exfoliated cervical cell GALR1m showed AUC = 0.704 for EC diagnosis, consistent with previous reports (AUC = 0.63).[13] CELF4m demonstrated significant differences between normal/hyperplastic and EC/AH tissues, with a dual-gene CELF4m/CDO1m test showing 84.9% sensitivity and 86.6% specificity for AH/EC detection.[24] Our findings corroborate these patterns, showing differential methylation between normal and EC tissues with 80% sensitivity but relatively lower specificity (50.7% and 60.9% for individual genes). The gene combination (CELF4m/GALR1m) improved sensitivity at the expense of some specificity. These differences might be caused by the different pathological compositions of the cases.
Notably, we found no methylation differences between early- and late-stage EC for any gene, mirroring previous observations with CDO1m and CELF4m,[24] potentially due to small sample sizes (<40 EC cases). However, all three genes showed >50% positivity in AH cases. Screening for AH in high-risk women is crucial, given its ~23% progression rate to EC.[25]
ZNF486 encodes a predicted DNA-binding transcription factor involved in transcriptional regulation. While implicated in multiple myeloma prognosis,[26] docetaxel-related prostate cancer immune infiltration,[27] and breast cancer outcomes,[28] ZNF486 has not been featured in recent EC methylation meta-analyses.[9,16,29,30] Our study identified significant ZNF486m differences between BL and EC [Figure 1c], demonstrating balanced sensitivity (71.4%) and specificity (63.8%) [Table 2]. Its combination with GALR1m achieved optimal performance, mirroring successful two-gene panels in other cancers.[11,31,32] Among our models, Model 1/3 (binary methylation status) showed identical performance to GALR1m/ZNF486m, while Model 2/4 (continuous ΔCp values) produced smoother ROC curves, similar to approaches in EC[15] and bladder cancer,[33] offering greater clinical flexibility for sensitivity/specificity optimization [Figure 2].
Study limitations include (1) a small sample size with limited EC subtypes, particularly in pre-menopausal women (only 6 EC cases), potentially underpowering sensitivity estimates, and (2) a lack of non-EC patient follow-up to track progression by methylation status. Nevertheless, our data confirm that while TVUS remains the primary EC screening method due to its non-invasiveness, affordability, and convenience, its high sensitivity but low specificity often leads to unnecessary invasive procedures. Single-gene methylation testing maintains TVUS-level sensitivity while improving specificity, and multi-gene panels further enhance both metrics, potentially reducing invasive interventions and preserving female fertility.
SUMMARY
In summary, we have identified promising epigenetic biomarkers in exfoliated cervical cells that demonstrate high diagnostic accuracy for detecting EC. This non-invasive approach is objective and highly reproducible. We will expand the sample size to validate the clinical utility of these DNA methylation-based biomarkers (CELF4m, GALR1m, and ZNF486m) for EC detection as well as to investigate the relationship between their methylation patterns and the prognosis of EC patients.
AVAILABILITY OF DATA AND MATERIALS
The data used and analyzed during the current study are available from the corresponding author on reasonable request.
ABBREVIATIONS
AH: Atypical hyperplasia
AUC: Area under the ROC curve
BL: Benign lesion
CELF4m: CUGBP Elav-like family member 4 methylation
CI: Confidence interval
Cp: Crossing point
EC: Endometrial cancer
FIGO: International Federation of Gynaecology and Obstetrics
GALR1m: Galanin receptor 1 methylation
IQR: Interquartile range
LR+: Positive likelihood ratio
LR−: Negative likelihood ratio
TVUS: Transvaginal ultrasound
ZNF486m: Zinc finger protein 486 methylation
ACKNOWLEDGMENT
Not applicable.
AUTHOR CONTRIBUTIONS
XF: Concepts, definition of intellectual content, literature search, clinical studies, experimental studies, data acquisition, data analysis, statistical analysis, manuscript preparation, manuscript editing and review; SM: Data acquisition, experimental studies, data analysis, manuscript editing and review; YO and FY: Data acquisition, data analysis, manuscript editing and review; XW: Concepts, design, clinical studies, experimental studies, manuscript editing and review; RW: Concepts, design, definition of intellectual content, literature search, clinical studies, experimental studies, data acquisition, data analysis, manuscript editing and review. All the authors have approved the release of the final version, and are responsible for all aspects of the work ensuring that any issues related to the accuracy or completeness of any part of the work have been properly investigated and resolved. All the authors are eligible for ICMJE authorship.
ETHICS APPROVAL AND CONSENT TO PARTICIPATE
This study was approved by the Research and Clinical Trial Ethics Committee of the Changsha Hospital for Maternal & Child Health Care (No.: EC-20240913-01), and it was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all the participants before the publication of this study.
CONFLICTS OF INTEREST
There are no conflicts of interest.
EDITORIAL/PEER REVIEW
To ensure the integrity and highest quality of CytoJournal publications, the review process of this manuscript was conducted under a double-blind model (authors are blinded from reviewers and vice versa) through an automatic online system.
FUNDING: Not applicable.
References
- Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA A Cancer J Clin. 2024;74:229-63.
- [CrossRef] [PubMed] [Google Scholar]
- Cancer statistics for the year 2020: An overview. Int J Cancer 2021:1-12.
- [CrossRef] [PubMed] [Google Scholar]
- Cancer incidence and mortality in China, 2022. J Natl Cancer Cent. 2024;4:47-53.
- [CrossRef] [PubMed] [Google Scholar]
- Risk factors for endometrial cancer: An umbrella review of the literature. Int J Cancer. 2019;145:1719-30.
- [CrossRef] [PubMed] [Google Scholar]
- A timely update of immunohistochemistry and molecular classification in the diagnosis and risk assessment of endometrial carcinomas. Arch Pathol Lab Med. 2021;145:1367-78.
- [CrossRef] [PubMed] [Google Scholar]
- Capacity of endometrial thickness measurement to diagnose endometrial carcinoma in asymptomatic postmenopausal women: A systematic review and meta-analysis. Ann Palliat Med. 2021;10:10840-8.
- [CrossRef] [PubMed] [Google Scholar]
- Research progress of DNA methylation markers for endometrial carcinoma diagnosis. J Cancer. 2025;16:812-20.
- [CrossRef] [PubMed] [Google Scholar]
- Endometrial cancer detection using a cervical DNA methylation assay (MPap) in women with abnormal uterine bleeding: A multicenter hospital-based validation study. Cancers (Basel). 2022;14:4343.
- [CrossRef] [PubMed] [Google Scholar]
- DNA methylation detection is a significant biomarker for screening endometrial cancer in premenopausal women with abnormal uterine bleeding. Int J Gynecol Cancer. 2024;34:1165-71.
- [CrossRef] [PubMed] [Google Scholar]
- Hypermethylated CDO1 and ZNF454 in cytological specimens as screening biomarkers for endometrial cancer. Front Oncol. 2022;12:714663.
- [CrossRef] [PubMed] [Google Scholar]
- DNA methylation testing for endometrial cancer detection in urine, cervicovaginal self-samples and cervical scrapes. Int J Cancer. 2023;153:341-51.
- [CrossRef] [PubMed] [Google Scholar]
- Hypermethylated PCDHGB7 as a biomarker for early detection of endometrial cancer in endometrial brush samples and cervical scrapings. Front Mol Biosci. 2021;8:774215.
- [CrossRef] [PubMed] [Google Scholar]
- Performance of the WID-qEC test versus sonography to detect uterine cancers in women with abnormal uterine bleeding (EPI-SURE): A prospective, consecutive observational cohort study in the UK. Lancet Oncol. 2023;24:1375-86.
- [CrossRef] [PubMed] [Google Scholar]
- The accuracy of DNA methylation detection in endometrial cancer screening: A systematic review and meta-analysis. Int J Gynecol Obstet. 2024;169:557-66.
- [CrossRef] [PubMed] [Google Scholar]
- The endometrial cancer detection using non-invasive hypermethylation of CDO1 and CELF4 genes in women with postmenopausal bleeding in Northwest China. Cytojournal. 2024;21:15.
- [CrossRef] [PubMed] [Google Scholar]
- Morphological differences between liquid-based cytology and conventional preparation in endometrial endometrioid carcinoma grade 1 and grade 3, and the differentiation of grades in each method. Acta Cytol. 2021;65:227-34.
- [CrossRef] [PubMed] [Google Scholar]
- Next-generation sequencing analysis of endometrial screening liquid-based cytology specimens: A comparative study to tissue specimens. BMC Med Genomics. 2020;13:101.
- [CrossRef] [PubMed] [Google Scholar]
- A non-invasive liquid biopsy screening of urine-derived exosomes for miRNAs as biomarkers in endometrial cancer patients. AAPS J. 2018;20:82.
- [CrossRef] [PubMed] [Google Scholar]
- Targeted proteomics identifies proteomic signatures in liquid biopsies of the endometrium to diagnose endometrial cancer and assist in the prediction of the optimal surgical treatment. Clin Cancer Res. 2017;23:6458-67.
- [CrossRef] [PubMed] [Google Scholar]
- Ultrasound detection of endometrial cancer in women with postmenopausal bleeding: Systematic review and meta-analysis. Gynecol Oncol. 2020;157:624-33.
- [CrossRef] [PubMed] [Google Scholar]
- GALR1 methylation in vaginal swabs is highly accurate in identifying women with endometrial cancer. Int J Gynecol Cancer. 2013;23:1050-5.
- [CrossRef] [PubMed] [Google Scholar]
- Hypermethylated CDO1 and CELF4 in cytological specimens as triage strategy biomarkers in endometrial malignant lesions. Front Oncol. 2023;13:1289366.
- [CrossRef] [PubMed] [Google Scholar]
- Current and future approaches to screening for endometrial cancer. Best Pract Res Clin Obstet Gynaecol. 2020;65:79-97.
- [CrossRef] [PubMed] [Google Scholar]
- Identifying the prognostic significance of mitophagy-associated genes in multiple myeloma: A novel risk model construction. Clin Exp Med. 2024;24:249.
- [CrossRef] [PubMed] [Google Scholar]
- Identification of docetaxel-related biomarkers for prostate cancer. Andrologia. 2021;53:e14079.
- [CrossRef] [Google Scholar]
- Identification and prognostic value exploration of cyclophosphamide (cytoxan)-centered chemotherapy response-associated genes in breast cancer. DNA Cell Biol. 2021;40:1356-68.
- [CrossRef] [PubMed] [Google Scholar]
- Minimally invasive and emerging diagnostic approaches in endometrial cancer: Epigenetic insights and the promise of DNA methylation. Diagnostics (Basel). 2024;14:2575.
- [CrossRef] [PubMed] [Google Scholar]
- Unveiling DNA methylation: Early diagnosis, risk assessment, and therapy for endometrial cancer. Front Oncol. 2024;14:1455255.
- [CrossRef] [PubMed] [Google Scholar]
- PAX1/SOX1 DNA methylation versus cytology and HPV16/18 genotyping for the triage of high-risk hpv-positive women in cervical cancer screening: Retrospective analysis of archival samples. BJOG. 2025;132:197-204.
- [CrossRef] [PubMed] [Google Scholar]
- Combination analysis of PCDHGA12 and CDO1 DNA methylation in bronchial washing fluid for lung cancer diagnosis. J Korean Med Sci. 2024;39:e28.
- [CrossRef] [PubMed] [Google Scholar]
- ITIH5 and ECRG4 DNA methylation biomarker test (EI-BLA) for urine-based non-invasive detection of bladder cancer. Int J Mol Sci. 2020;21:1117.
- [CrossRef] [PubMed] [Google Scholar]