ctDNA methylation profiling reveals NBL1 as a promising biomarker for early ovarian cancer screening

Abstract

Background

Early detection of epithelial ovarian cancer (EOC) is crucial for improving patient survival. Current screening methods have limitations, highlighting the need for novel biomarkers. Circulating tumor DNA (ctDNA) methylation analysis offers a promising approach.

Methods

This study included 10 patients with EOC and 10 patients with benign pelvic masses. We collected plasma samples from these patients and isolated ctDNA. We then conducted whole-genome methylation sequencing using the TAPS (TET-assisted pyridine borane sequencing) method, which allows for single-base resolution detection of 5-methylcytosine and 5-hydroxymethylcytosine. Bioinformatics analysis was performed to identify differentially methylated genes and regions. We further validated candidate biomarkers using bisulfite sequencing, qRT-PCR, and IHC. TCGA methylation data were analyzed for external validation.

Results

We identified 35 differentially methylated genes, with NBL1 and CASZ1 as potential candidates. NBL1 gene hypermethylation in EOC patients was significantly associated with reduced mRNA expression, suggesting its role as a tumor suppressor gene. CASZ1 methylation patterns were inconsistent between blood and tissue, indicating limited utility as a diagnostic biomarker. We also observed widespread hypo-methylation across the genome and hyper-methylation in specific regions of differential methylation. GO and KEGG pathway enrichment analyses revealed that the differentially methylated genes were involved in various biological processes and pathways relevant to cancer pathogenesis. There is a significant negative correlation between the methylation level and the mRNA level of the NBL1 gene, suggesting that hypermethylation of the NBL1 gene may be associated with a reduction in its expression. Furthermore, immunohistochemical analysis indicates a downregulation of NBL1 expression in ovarian cancer tissues, which contrasts with the strong positive expression observed in benign tissues.

Conclusion

Our study demonstrates the potential of ctDNA methylation analysis for early EOC detection. we propose that NBL1 gene hold potential as screening biomarkers for ovarian cancer.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12957-025-03957-1.

Introduction

Ovarian cancer ranks third in incidence and second in mortality among female reproductive system tumors, posing a severe threat to women’s health [1, 2]. Epithelial ovarian cancer (EOC) is the most common subtype, and there has been no significant improvement in the survival rate for patients with FIGO stage III/IV disease in the past decade. In contrast, the survival rate for early-stage FIGO I/II patients reaches 90%. Plasma tumor markers and imaging examinations remain the primary screening methods, but they have limitations. The early diagnosis rate of EOC is merely 25% [3]. The existing biomarkers CA125 and HE4 are insufficient for early screening. Liquid biopsy technology, especially circulating tumor DNA (ctDNA) methylation analysis, offers a new direction for the early diagnosis of EOC.

ctDNA is fragmented DNA released during the apoptosis, necrosis, and metabolic processes of tumor cells. Research both domestically and internationally has explored the role of ctDNA methylation in the early diagnosis of EOC [4–8]. Although specific methylation biomarkers have not yet been identified [9–12], multiple studies have recognized potential biomarkers. The Early Detection Research Network (EDRN) has not yet proposed performance threshold criteria as a gold standard, but it has provided guidelines for identifying a hypothetical biomarker for ovarian cancer screening. This biomarker should be able to detect 80% of ovarian cancer patients, and the associated specificity of this biomarker should not be less than 92.5%. For a biomarker with a commonly accepted sensitivity of 100%, the specificity requirement is higher than 90.6% [13].

In this study, we utilized ctDNA whole-genome methylation sequencing technology to compare the blood of EOC patients with that of patients with benign pelvic masses. By integrating bioinformatics database analysis, we screened for potential biomarkers. Subsequent validation studies confirmed that NBL1 is a potential biomarker for the early screening and progression of EOC.

Methods

Study layout and clinicopathological samples

From October 2022 to October 2023, we collected a total of 92 surgical cases of ovarian cancer and benign masses, excluding 72 patients, and finally included 20 patients who were aged 35 years or older. Informed consent was obtained from patients at the Department of Obstetrics and Gynecology, the second Xiangya Hospital, Central South University. Plasma sampling was conducted before tumor surgery and stored at -80 °C for future use. The samples included in this study were of Chinese descent.

The inclusion criteria for the ovarian cancer group are as follows: (1) Confirmed epithelial ovarian cancer, such as serous carcinoma, clear cell carcinoma, endometrioid carcinoma, etc. (2) Age over 35 years. (3) No history of treatment, surgery, or medication.

The inclusion criteria for the benign pelvic mass group are: (1) Have a complete uterus (no history of hysterectomy, tubal ligation, or bilateral salpingectomy). (2) Age over 35 years. (3) Ultrasonic detection of a pelvic mass. (4) If the mass’s maximum diameter is 3–5 cm, there is no reduction in the mass after two consecutive ultrasound comparisons; if the mass’s maximum diameter is greater than 5 cm.

Exclusion Criteria: (1) A history of any cancer other than ovarian cancer; (2) Severe diseases outside the ovary, especially those with a life expectancy of less than 3 years; (3) Factors that may affect the diagnosis of the disease (such as pacemakers or ferromagnetic implants related to MRI); (4) A history of allogeneic organ transplantation and allogeneic hematopoietic stem cell transplantation; (5) A history of primary immunodeficiency disease; (6) A history of blood transfusion within 1 month; (7) Risks associated with blood collection (e.g., hemophilia, severe anemia with hemoglobin levels below 8.0 g/dL); (8) Pregnant or planning to become pregnant; (9) Any other condition that the investigator deems inappropriate for inclusion in the study or that may interfere with the completion of the study; (10) Inability to provide informed consent or refusal to undergo blood collection.

Basic patient information was collected, and clinical pathological characteristics were evaluated to determine tumor stage and the presence of metastasis. In the discovery phase (plasma group), patients were divided into EOC and control groups (benign pelvic mass group). Similarly, in the validation phase (tissue group), patients were divided into EOC and control groups (benign pelvic mass group).

Whole genome methylation sequencing of cfDNA

We conducted GM-seq sequencing and bioinformatics analysis on the isolated cfDNA from EOC and benign pelvic masses. To avoid any sample bias during the GM-seq process, we included one early-stage and four late-stage samples in the cancer group. Additionally, the samples in the cancer group were histologically confirmed as EOC.

DNA extraction, library construction, sequencing

Whole-genome methylation sequencing of cfDNA was conducted using the GM-seq platform, which is based on the TET enzyme. The methodologies for DNA extraction, library construction, and GM-seq have been detailed in prior literature [14]. Blood samples were subjected to centrifugation at 1608×g for a duration of 10 min. The supernatant was subsequently transferred to microcentrifuge tubes and centrifuged at 16,000×g for an additional 10 min to eliminate any remaining cellular debris. Plasma cfDNA was extracted from plasma samples using TANBead Maelstrom 2400 extraction instrument (TANBEAD, Taoyuan, China) and MagMAX Cell-Free DNA Isolation Kit (ThermoFisher, Waltham, MA, USA). DNA concentration was measured by Qubit™ dsDNA HS Assay Kit (ThermoFisher, Waltham, MA, USA). The size of cfDNA fragments was assessed using the Qsep100 automated Bio-Fragment Analyzer (Bioptic, New Taipei, China). Before library construction, sequences with CpG totally methylated (positive references) and CpG totally unmethylated (negative references) were mixed into the samples as controls. DNA methylation sequencing libraries were constructed using Hieff NGS^® Ultima Pro DNA Library Prep Kit for Illumina (Yeason, Shanghai, China), including end repair, dA tailing, adaptor ligation. Subsequently, 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC) were oxidized to 5-carboxycytosine (5caC) using the TET2 oxidase enzyme, followed by conversion to dihydrouracil (DHU) in the presence of a reducing agent (pyridine borane). DHU serves as a PCR template and is recognized by a DNA polymerase that identifies uracil. Through PCR amplification, 5mC is converted to thymine, facilitating whole-genome sequencing, which was carried out using the Gene + seq2000 sequencer (Geneplus, Suzhou, China).

Read mapping and methylation calling

Raw sequencing data were refined by removing adapter sequences and low-quality reads using the fastp software (v0.19.5) [15]. Clean reads were aligned to the human reference genome (hg19) with Sentieon software(version 202,010). The mapping data for samples, positive controls, and negative controls were extracted from BAM files using Samtools(v1.9) [16]. The average sequencing depth in our study was 10 × and the quality control data for cfDNA-based whole genome methylation sequencing was detailed in Additional file 1: Table S1.

To analyze methylation data from the BWA meth and BWA mem results, we employed MethylDackel v. 0.5.1 (https://github.com/dpryan79/MethylDackel). We specifically filtered out SNPs, compiled per-base methylation metrics, and generated an output file formatted for compatibility with methylKit in R (-- methylKit) (Akalin et al., 2012). As methylation call accuracy scales logistically with read depth and becomes asymptotic at 10x (Seiler Vellame et al., 2021), we used --minDepth 10 to only retain reads that were sequenced at least 10 times. Three different SNP filtering cut offs were tested (--maxVaraintFrac 0.25, 0.5, and 0.8) with --minOppositeDepth 5. To enhance the detection of potential methylation variations, we selected the most permissive SNP filter (-- maxVariantFrac 0.8) for subsequent analyses.

Analysis of differentially methylated regions (DMRs)

Subsequently, asTair (v3.3.2) was employed for downstream analysis to detect differentially methylated regions (DMRs). This tool is specifically designed for the analysis of bisulfite-free, base-resolution sequencing data generated through modified cytosine-to-thymine conversion methods. It was used to identify methylated sites and assess the mean methylation levels of CpG sites.

CpG sites with a coverage depth of ≥ 5× were selected for further analysis of differentially methylated regions (DMRs) and principal component analysis (PCA), which were conducted using Metilene(v0.2–8) [17] and prcomp function in R package stats (version 4.2.2), respectively.

In the DMRs analysis, the minimum unit was defined as a genomic region containing at least 10 CpG sites with no more than 300 bp between adjacent CpGs. DMRs were identified based on the following criteria: (a) the methylation difference level between groups was greater than 0.1 or smaller than − 0.1, and (b) the q-value was smaller than 0.1. DMRs were then annotated to different functional regions, including promoters, introns, exons and intergenic regions, using R package genomation (version 3.17). Venn diagram was plotted using an online tool named Venny (2.1). DMRs were used to perform gene ontology (GO) pathway enrichment analyses using R package clusterProfiler (version 4.7.1.003) [17]. Unsupervised clustering was performed using R package pheatmap (version 1.0.12).

TCGA methylation data analysis

We downloaded the TCGA ovarian cancer methylation dataset from the TCGA database, which included 582 primary ovarian cancer samples, 19 recurrent ovarian cancer samples, and 12 normal samples. To identify methylation markers associated with early-stage ovarian cancer, differential methylation analysis was performed using 12 normal samples and 43 samples of ovarian cancer stages I-II. The R package TCGAbiolink was utilized for differential methylation analysis, with filtering parameters set at P < 0.05 and diff_mean_abs > 0.1 to identify differentially methylated sites. The differentially methylated sites were then intersected with the differentially methylated sites from the TCGA database to further screen for potential candidate biomarkers.

Bisulfite target gene methylation sequencing

Ten tissue samples collected in the validation phase (5 EOC samples and 5 benign pelvic mass samples) were selected for bisulfite target gene sequencing. Specific validation regions were identified, and bisulfite sequencing PCR primers were designed using the online MethPrimer software (www.urogene.org/methprimer/index.html). These primers were designed to target regions without CpG sites to avoid amplification bias between methylated and unmethylated sequences.

The BSP validation experiment was conducted as follows: 500 ng of genomic DNA was converted using the Zymo EZ DNA Methylation Gold Kit™ according to the manufacturer’s instructions. The thermal cycling program consisted of 94 °C for 1 min, 94 °C for 10s, 58 °C for 30s, and 72 °C for 30 cycles, followed by a final extension at 72 °C for 5 min, with the products then stored at 12 °C. After amplification, the PCR products were selected and purified using the QIAquick Gel Extraction Kit (Qiagen), and the purified PCR products were subcloned. The colonies from each region were sequenced on a 3730 Genetic Analyzer (Applied Biosystems) to analyze the levels of methylated cytosines.

Screening of ovarian cancer-related methylation sites

Given the limited scale of the discovery cohort, only 10 samples were included, we incorporated methylation data from benign samples and stage I-II tumor samples in the public TCGA-OV dataset for parallel analysis to address potential randomness of the results. By intersecting the findings from both the discovery cohort and the TCGA-OV dataset, we identified 9 differentially methylated sites from two candidate genes. These candidate genes were further validated in 5 additional pairs of newly collected samples, ultimately yielding NBL1 gene. Based on this rigorous multi-step validation, we propose that NBL1 gene hold potential as screening biomarkers for ovarian cancer (Fig. 1).

Fig. 1.

Research design overview. Research Design Overview illustrating the overall experimental design. In the discovery phase, blood samples were collected preoperatively from 10 patients (5 with early ovarian cancer and 5 with benign pelvic masses). Genome-wide DNA methylation sequencing (GM-seq) was performed on these samples, identifying 322,252 differentially methylated sites (DMSs) from 91 genes. Concurrently, bioinformatics analysis of the TCGA-OV dataset revealed 1,805 DMSs (from 14 genes). The intersection of DMSs from both datasets was obtained, and stringent filtering criteria (P < 0.05,|diff_mean_abs| >0.1) were applied, resulting in the selection of 9 DMSs derived from the CASZ1 and NBL1 genes. In the subsequent validation phase, tumor tissues were collected intraoperatively from a new cohort of 10 patients (5 EOC and 5 benign pelvic masses). Targeted gene methylation sequencing was conducted on these tissues, revealing 5 DMSs with statistical significance, also originating from the CASZ1 and NBL1 genes. Finally, these selected DMSs were further validated using quantitative real-time PCR (qRT-PCR) and immunohistochemistry (IHC)

Quantitative real-time polymerase chain reaction (qRT-PCR)

The total RNA was extracted using anRNA Extraction Kit. We reverse transcribed 1 µg of RNA into cDNA with Evo M-MLV RT Kit and Oligo(dT)18 primer (Accurate biotechnology, ). The qRT-PCR was performed with SYBR Green Pro Taq HS Pre-Mixed qPCR Kit 11,701(Accurate biotechnology, ). The sequences of the primers are shown below. We normalized the mRNA expression level with that of β-actin and calculated the relative expression levels with the -ΔCT method.

The primer sequences were as follows:

• NBL1Forward 5’-TGTTCCCAGATAAGAGTGC-3’.

• Reverse 5’-GCAGGAGTCACAGTGAACCAG-3’.

• CASZ1aForward 5’-GGATGGAGACAGATGAGTGC-3’.

• Reverse 5’-GTCGGCATAGAGATGGTGTT-3’.

• CASZ1bForward 5’-TCCCCGAGCCCGTAT-3’.

• Reverse 5’-GGGTCCTCCACCCAAGA-3’.

• β-actin Forward 5’-GCCAACCGCGAGAAGATGA-3’.

• Reverse 5’-CATCACGATGCCAGTGGTA-3’.

Immunohistochemistry

In the standard immunohistochemical procedure, slides were incubated with the NBL1 antibody (1:50; HPA007394) overnight at 4°C. The immunohistochemical staining was visualized using 3,3’-diaminobenzidine (DAB). Subsequent to hematoxylin counterstaining, the sections were dehydrated and mounted. The diagnoses of all pathological sections were provided by experts from the Department of Pathology at the Second Xiangya Hospital.The staining scores were evaluated by assessing the stained glandular regions of the ovarian cancer group versus the control group. The assessment involved quantifying the percentage of positive cells and the staining intensity, with the final score being the product of these two values. The immunohistochemical scoring criteria were primarily based on staining intensity (ranging from 0 to 3 points) and the percentage of positive cells (ranging from 1 to 4 points). The overall score was computed as follows. Staining intensity was scored as: no dyeing, 0 points; light yellow, 1 point; yellow or brown, 2 points; brown or tan, 3 points. The percentage of positive cells was scored as: ≤25%, 1 point; 26–50%, 1 point; 51–75%, 2 points; ≥75%, 3 points.

Statistical analysis

All experiments were independently replicated three times, and the data were statistically analyzed using SPSS^® (Version 26, IBM Corp., Armonk, NY, USA). All statistical tests were conducted as two-tailed tests. A P-value of ≤ 0.05 was considered to indicate a statistically significant difference, with a confidence interval set at 95%. Count data for each group were expressed as means ± standard deviation (SD) for normally distributed continuous variables, and as median (interquartile range, IQR, Q1-Q3) for skewed continuous variables.

Statistical analysis and graph plotting of the experimental data were performed using GraphPad Prism 9.5 software. The sample size for each group was as indicated in the figure legends. The appropriate statistical analysis method was selected based on the nature of the experimental data, such as independent samples t-test, one-way analysis of variance, or the Mann-Whitney U test. A P-value less than 0.05 was deemed statistically significant.

Human ethics and consent to participate declaration

In this study, we confirm that all human participants involved in the study provided their informed consent to participate. The consent process was conducted in accordance with the guidelines set forth by the Ethics Committee of the Second Xiangya Hospital, which approved the study under the reference number LYF2022125.

Result

Clinical and pathological information of patients

The study cohort encompassed 9 patients with high-grade serous carcinoma, one patient with low-grade serous carcinoma, and 10 patients with benign pelvic masses as controls (Fig. 2). The majority of ovarian cancers were classified as Stage III-IV (90%). The age range of all participants spanned from 35 to 67 years. The characteristics of patients like serum CA125, serum HE4, and tumor diameter are listed in Table 1.

Fig. 2.

Study layout and clinicopathological samples

Table 1.

Clinical and pathological information of patients

Variable	Serum Group (n = 10)		Tissue Group (n = 10)
	Control (n = 5)	Ovarian Cancer (n = 5)	Control (n = 5)	Ovarian Cancer (n = 5)
Mean Age (Range)	47 (35–63)	53 (50–56)	58 (45–66)	55 (44–67)
Histological Type
High-grade Serous Ovarian Cancer	-	4	-	5
Low-grade Serous Ovarian Cancer	-	1	-	0
FIGO Stage
I-II Stage	-	1	-	0
III-IV Stage	-	4	-	5
Postmenopausal Women
Yes	2	5	3	4
No	3	0	2	1
Serum CA125 (U/ml)	41.6 ± 54	421.6 ± 162.33	14.6 ± 6.5	415.1 ± 226.1
Serum HE4 (pmol/L)	32.5 ± 7.8	483.8 ± 506.4	32.5 ± 8	1749.2 ± 1138.1
Average Tumor Diameter (mm)	70 ± 14.6	83.6 ± 51	97.2 ± 33.1	81.8 ± 55.3

Landscapes for whole-genome methylation analysis in blood samples of ovarian cancer patients and benign patients

The GM-Seq was performed for all ten plasma samples that passed quality control. To guarantee the integrity of the input data for subsequent analyses, we performed an initial quality control assessment of the raw sequencing data using the FASTQC tool (refer to Table S2). As depicted in Table S2, the Clean Rate for each dataset exceeded 99%, signifying the high validity of the data. Additionally, the Q30 (%) scores ranged from 91 to 95%, which indicates a high degree of sequencing accuracy. All detected sequencing base qualities were above the threshold of 30, confirming the robustness of the base sequencing quality (refer to Figure S1).

We conducted statistics on the methylation levels across the whole genome for each sample (Table S3). The calculated average methylation level in the control group was 66.2%, while in the OC (ovarian cancer) group, it was 66%. This indicates that there was no significant difference in the overall methylation distribution among the samples.

The whole-genome methylation analysis identified a total of 322,252 differentially methylated sites (DMS), with 99,290 sites (30.8%) being hyper-methylated and 222,962 sites (69.2%) being hypo-methylated. Of these differential methylation sites, 40% were located in introns, 55% in intergenic regions, 2% in promoters, and 3% in exons (Fig. 3-b). Cluster analysis based on the differentially methylated sites was performed for the two groups of samples (Fig. 4-a) indicating that the differential methylation sites could distinguish between the epithelial ovarian cancer group and the benign pelvic mass group.

Fig. 3.

(a) Distribution of differentially methylated sites in genomic functional regions. (b) Distribution of differentially methylated regions in genomic functional regions

Fig. 4.

Integrative analysis of differentially methylated sites: Clustering, GO, and KEGG Pathway Enrichment. (a) Heatmap of differentially methylated site clustering. (b) GO analysis; (c) KEGG analysis. The x-axis represents the number of enriched genes, the color indicates the enrichment significance of the p.adjust value. The color transitions from green to red, indicating a decrease in the p-value and an increase in significance

We identified a total of 91 genes with differential methylation regions, including 56 hyper-methylated regions (61.5%) and 35 hypo-methylated regions (38.5%). Of these differential methylation regions, 36% were located in introns, 42% in intergenic regions, 9% in promoters, and 13% in exons (Fig. 3-a). The data from the differentially methylated sites suggest that there is widespread hypo-methylation across the genome and hyper-methylation in specific regions of differential methylation, which is consistent with previous research findings.

We selected 35 significantly differentially methylated regions (from 35 genes), with 15 genes showing hyper-methylation (42.8%) and 20 genes showing hypo-methylation (57.2%) (Table S4). The 35 genes then were imported into STRING for biological evaluation, including GO term enrichment analysis and KEGG pathway analysis (Figs. 3-b, c). This analysis suggests that the differentially methylated genes identified in the study are implicated in a variety of biological processes and pathways that could be relevant to the pathogenesis of cancer and other diseases, highlighting the importance of epigenetic regulation in these processes.

External validation and screening for candidate biomarkers utilizing the TCGA database

Due to the limitation of sample size, we concurrently introduced a cohort of patient samples from the TCGA database to enhance the credibility of the discovery phase.Utilizing the R package TCGAbiolink (version 1.36.0), we performed differential methylation analysis on the TCGA public database, comparing 43 stage I-II ovarian cancer samples with 12 normal samples. After applying a filter threshold of P < 0.05 and|diff_mean_abs| >0.1, we identified a total of 1,805 differentially methylated sites, originating from 14 genes.These sites include 1,511 (83.7%) hypermethylated sites and 294 (16.3%) hypomethylated sites (Table S5).

To strengthen the reliability of potential biomarkers, we adopted a systematic strategy to refine the list of promising candidates. We intersected the differentially methylated genes from our initial study with those obtained from the TCGA database. Our filtering criterion required a q.value < 0.05 and|mean.meth.diff| >0.1.This process led to the identification of the most promising 9 differentially methylated sites: chr1:10717784, chr1:10676361–10676362, chr1:10717738, chr1:10717724, chr1:10676927–10676928 and chr1:19645819, chr1:19646010, chr1:19842791–19842792, chr1:19841472–19841473 (Table S6).

Experimentally validating the relationship between gene methylation and tissue expression of NBL1 and CASZ1

We utilized bisulfite target gene sequencing to analyze tissue samples from 5 patients with epithelial ovarian malignant tumors and 5 patients with benign pelvic masses. The analysis uncovered significant differences in 5 of the candidate biomarkers, with the following sites exhibiting hypermethylation in the cancer group: chr1:10717784, chr1:10676361–10676362 (within the CASZ1 gene); chr1:19645819, chr1:19646010, chr1:19842791–19842792 (within the NBL1 gene) (Fig. 5).

Fig. 5.

Methylation levels of differentially methylated sites. (a) chr1: 19,645,819. (b) chr1: 19,646,010. (c) chr1: 19,842,791–19,842,792. (d) chr1: 10,717,784. (f) chr1: 10,676,361–10,676,362

qRT-PCR technology was employed to evaluate the mRNA levels of CASZ1a, CASZ1b, and NBL1 in 4 EOC specimens (from the initial 5, with one omitted due to insufficient sample size) and 5 benign pelvic masses, corresponding to the tissue methylation profiles (Figure S2). In comparison to normal tissue, the expression level of NBL1 was more than twofold reduced in EOC specimens (p < 0.05) (Fig. 6a). The expression levels of CASZ1a and CASZ1b in EOC specimens were increased by 10-fold relative to the benign pelvic mass group, although this elevation did not reach statistical significance (Figs. 6b, c).

Fig. 6.

Gene expression and methylation analysis: amplification curves and mRNA-methylation correlations. (a) Relative expression level of NBL1; (b) Relative expression level of CASZ1a; (c) Relative expression level of CASZ1b; (d) Relationship between mRNA expression and methylation level at chr1:19842791; (e) Relationship between mRNA expression and methylation level at chr1:19646010; (f) Relationship between mRNA expression and methylation level at chr1:19645819

There is a significant negative correlation between the methylation level and the mRNA level of NBL1 (Figs. 6d-f, p < 0.05), suggesting that hypermethylation in NBL1 may contribute to the silencing of NBL1 gene expression.

In the control group, the strong positivity of NBL1 in the cells surrounding the glands suggests high expression of NBL1. In contrast, the weak positivity of NBL1 in the cells surrounding the glands of ovarian cancer patients indicates that NBL1 protein expression is suppressed in cancer (Figs. 7A, B).

Fig. 7.

Clinical significance of NBL1 expression in ovarian cancer versus control patients. NBL1 staining was performed on consecutive paraffin-embedded tissue blocks from 5 ovarian cancer tissues and 5 benign ovarian mass tissues. Representative microscopic images of paired NBL1 staining are shown at 20x and 40x magnification, with a scale bar of 50 μm. A shows ovarian cancer tissue with low NBL1 staining in the glandular portion. B depicts high NBL1 expression in the control group patients

Conclusion

Our study demonstrates the potential of ctDNA methylation analysis for early EOC detection. we propose that NBL1 gene hold potential as screening biomarkers for ovarian cancer.

Discussion

In the present study, we initially conducted whole-genome methylation sequencing on blood samples from patients with EOC and a control group (benign pelvic mass), which led to the identification of 35 differentially methylated genes potentially significant for the early diagnosis of EOC. Subsequent intersection analysis with differentially methylated genes from the TCGA database narrowed down the candidates to 9 differentially methylated cites (from the CASZ1 and NBL1 genes). Targeted methylation sequencing in tissue samples, along with mRNA and protein level validations, was performed for these candidate biomarkers. The study found that the methylation level of the NBL1 gene in both blood and tissue samples from ovarian cancer patients was higher than that in the benign pelvic mass group and correlated with mRNA and protein expression levels in tissue samples. These findings suggest that methylation of the NBL1 gene in ctDNA holds promise as a potential biomarker for early ovarian cancer screening and may be involved in the pathogenesis and development of ovarian cancer through gene expression silencing. In contrast, CASZ1 gene displayed inconsistent methylation patterns between blood and tissue samples in ovarian cancer patients, and tissue methylation did not correlate with mRNA expression, indicating that CASZ1 methylation may not be associated with its expression in ovarian cancer.

Many research teams have explored a variety of cancer methylation biomarkers. Although there is currently no single gene-specific methylation biomarker or a set of methylation biomarkers for the early diagnosis of ovarian cancer, numerous studies have investigated the potential of various DNA methylation biomarkers in plasma or serum cfDNA to identify gene-specific methylation biomarkers that can distinguish between ovarian cancer patients and those with benign ovarian diseases or healthy controls [11–15]. Gen Li and colleagues explored the use of Transformer-based AI technology and circulating free DNA (cfDNA) methylation markers to improve early diagnosis of ovarian cancer. The researchers screened 493 ovarian cancer-related methylation markers from 3 million CpG sites across the genome in 3,000 samples and validated these markers in 1,800 independent cfDNA samples. Using a pre-trained methylation Transformer model (MethylBERT), they developed an ovarian cancer diagnostic model that achieved 80% sensitivity and 95% specificity in the early diagnosis of ovarian cancer, outperforming the traditional LASSO-Logistic regression model. As we know, most existing diagnostic biomarkers fail to meet the standards set by the Early Detection Research Network (EDRN). Research has shown that liquid biopsy technology based on ctDNA methylation has a promising application in OC early screening, but there are problems and shortcomings in current OC early screening research [18–21]. Firstly, the sensitivity and specificity of the identified models have not met the assumed biomarker criteria. Additionally, the experimental technique for ctDNA methylation, which involves bisulfite conversion in most studies, is prone to DNA fragmentation and loss, resulting in low ctDNA concentrations and adverse effects on subsequent data analysis. In this study, we employed the latest trace methylation detection technology, GM-seq (Genome Methylome sequence) sequencing, to conduct high-throughput methylation analysis of plasma ctDNA in ovarian cancer and benign pelvic masses. This technology uses an enzyme conversion library construction technique for whole-genome single-base resolution methylation detection and can stably detect as low as 10 ng of plasma ctDNA. NBL1, screened using this technology, could be a promising biomarker for early ovarian cancer detection.

Neuroblastoma suppressor of tumorigenicity 1 (NBL1) is a tumor suppressor gene identified in neuroblastoma, which inhibits tumor formation and progression by regulating the cell cycle and promoting apoptosis [18, 19]. NBL1 is also implicated in the development and progression of various cancers. The miR-1301-3p promoted the invasion, migration, and EMT progression of esophageal cancer by downregulating NBL1 expression [20]. In small cell lung cancer, neuron-specific enolase (NSE) may weaken the inhibitory effect of NBL1 on BMP2 through its downregulation, thereby enhancing the interaction between BMP2 and BMPR1A; conversely, NSE activatedthe BMP2/Smad/ID1 pathway by inhibiting NBL1, promoting the stem cell characteristics of small cell lung cancer, and correlating with poor prognosis in patients [21]. NBL1 was as a highly specific gene in prostate cancer and normal prostate tissues, which was highly expressed in DU145 PCa cell lysates and culture media, and NBL1 was detectable in both normal and cancerous prostate tissues, with significantly reduced expression in the latter [22].

These studies suggest that NBL1 acts as a tumor suppressor gene in various cancers. However, the association between NBL1 and ovarian cancer has not been explored to date. Our analysis revealed that NBL1 was hypermethylated and downregulated at the mRNA level in ovarian cancer tissues compared to normal ovarian tissues. These results imply that NBL1 may function similarly to other cancers as a tumor suppressor gene in ovarian cancer development.

NBL1 is known to possess inhibitory activity against bone morphogenic proteins (BMPs), particularly BMP-4. BMP-4 and its receptors are widely expressed in the ovary, and studies have shown that BMP-4 treatment increases the migration and adhesion of primary OC cells [2, 23]. The above research suggests that NBL1 may be involved in the pathogenesis and development of ovarian cancer as a BMP inhibitor.

Castor zinc finger 1 (CASZ1), also known as zinc finger protein 693 (ZNF693), is an evolutionarily conserved transcription factor involved in a variety of embryonic development and physiological processes [24]. In cancer, CASZ1 can function as either a tumor suppressor or an oncogene, depending on the cellular context. Yi-Ying Wu and colleagues have found that CASZ1 is upregulated in ovarian cancer tissues and cell lines, and its knockdown inhibits metastasis in vivo, suggesting a role in promoting epithelial-mesenchymal transition (EMT) in ovarian cancer cells. The hypermethylation status of CASZ1 in ovarian cancer tissues, in contrast to the elevated mRNA expression, suggests that CASZ1 methylation may not be involved in its expression within the tissue. Moreover, since ctDNA methylation patterns in cancer often correlate with tissue methylation levels, the discrepancy between the hypomethylation observed in blood whole-genome methylation sequencing and the hypermethylation in tissue-targeted methylation sequencing for CASZ1 suggests that CASZ1 methylation may not have significant implications for the early diagnosis of ovarian cancer [25].

The main limitations of this study include: (1) Although we employed a novel ctDNA whole-genome methylation sequencing method (GM-NGS) for sample analysis, the study was constrained by limited resources, resulting in the inclusion of only 20 samples for the screening and validation of candidate biomarkers. This sample size is relatively small, and future studies will require larger prospective cohort samples for validation. (2) Postoperative follow-up of patients was not conducted, precluding the analysis of the impact of NBL1 promoter region methylation levels on the prognosis of ovarian cancer patients. (3) The role of NBL1 methylation and its downstream molecules, such as BMP, in the pathogenesis and progression of ovarian cancer was not thoroughly investigate.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Shunxian Zhou, Chun Fu have been responsible for the idea and the design of the study. All authors have been writing and have had critical discussions of the manuscript. All authors have approved the final version of the manuscript. Additionally, the named authors have no conflict of interest, financial or otherwise.

Funding

This study was supported by grants from the National Natural Science Foundation of China (81771546, 82271674), Natural Science Foundation of Hunan Province (2024JJ9129). Major Scientific Research Project for High level Health Talents in Hunan Province in 2023 (R2023095).

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Ethics approval and consent to participate

In this manuscript, we confirm that all human participants involved in the study provided their informed consent to participate. The consent process was conducted in accordance with the guidelines set forth by the Ethics Committee of the Second Xiangya Hospital, which approved the study under the reference number LYF2022125.

Consent for publication

Consent for publication has been obtained from all individuals included in this study. Written informed consent was obtained using our institutional consent form.

Competing interests

The authors declare no competing interests.