DNA methylation for cervical cancer screening: a training set in China

Background Despite rapid improvements in DNA methylation tools for cervical cancer screening, few robust, exploratory studies have been performed using the combination of two host genes, EPB41L3 and JAM3, newly developed assays. Methods A review of abnormal liquid-based cytology and/or high-risk human papillomavirus (hrHPV) data from outpatient clinics in the study center from March 2018 to March 2019 was performed. Eligible patients with definitive histological pathology results were included, and their residual cytology samples were assessed for EPB41L3 and JAM3 methylation. The diagnostic accuracies of various screening strategies for definitive pathology and for cervical intraepithelial neoplasia (CIN) 2 or more severe lesions (CIN2+) were compared. Results In total, 306 patients were successfully tested; 301 cases with cervical histological pathology were included in the final analysis, including 118 (39.2%) and 183 (60.8%) cases of inflammation/CIN1 and CIN2+, respectively. Regarding CIN2+ detection, methylation status and hrHPV plus methylation had similar positive predictive values (0.930 and 0.954, respectively, p = 0.395). Additionally, hrHPV, methylation, and hrHPV plus methylation had similar negative predictive values (0.612, 0.679, and 0.655, p = 0.677) that were significantly higher than that of cytology alone (0.250, p values 0.012, 0.001, and 0.001, respectively). For 49 cases with negative hrHPV results, positive methylation alone was able to differentiate CIN2+ from inflammation/CIN1. Conclusions Methylation of both EPB41L3 and JAM3 is an accurate and feasible screening method for CIN2+.


Introduction
Uterine cervical cancer is one of most common causes of cancer-related death among women worldwide [1]. According to a conservative estimate, the total incidence and mortality of cervical cancer in China were 98,900 and 30,500 cases, respectively, in 2015 [2], accounting for one-fifth of the total number of new cases of cervical cancer worldwide [3]. However, great discrepancies exist in the popularity and quality of screening methods [4]. Although a combination of cervical cytology and/or high-risk human papillomavirus (hrHPV) in women of relevant age is the mainstream screening method [5,6], cytology has several limitations, including a lack of high-throughput characteristics, requirement for high-level skills by pathologists, and low sensitivity [7,8]. The limitations of hrHPV testing pertain to its high cost, low specificity, and possible reproducibility difficulties given the large Chinese population and territory. Inevitably, the cost versus benefit is key for the decision to undergo cervical cancer screening, even though defining an acceptable risk will likely differ between settings [9]. Thus, a new screening strategy with high accuracy and feasibility is urgently needed.
Increasing evidence has shown that epigenetic silencing of tumor-suppressor genes is essential to carcinogenesis and metastasis [10]. The actions of one epigenetic mechanism, DNA methylation, result in the heritable silencing of genes without changes in their coding sequences [11], thereby affecting virtually every step in tumor progression [12]. DNA methylation is also essential for the progression and pathogenesis of cervical cancer, as reflected in its sensitivity for prognosis and therapy in clinical practice. As genotyping and methylation markers are objective and applicable to self-collected samples, these approaches offer logistical advantages, including accessibility in low-and middle-income settings [9]. More than 100 human (host) genes have been reported to be possible methylation biomarkers of cervical cancer [13]. Furthermore, numerous studies have shown that methylation has high screening sensitivity for lesions of cervical intraepithelial neoplasia (CIN) 2 or more severe lesions (CIN2+ or high-grade intraepithelial lesions [HSIL]) and can be used as a triage method in women with positive hrHPV status. Multiple panels consisting of dozens of candidate host genes, virus genes or both, as well as various combinations, have been utilized as classifiers [14]. A number of studies have explored the role of a panel including EPB41L3 [15][16][17][18][19][20][21][22][23][24], JAM3 [25,26], or both [7,[27][28][29][30][31] in the screening or triage of HSIL and/or cervical cancer.
In this exploratory study, we selected a cohort with definitive cervical biopsy pathology findings after abnormal cytology and/or hrHPV testing so as to develop an assay of EPB41L3 and JAM3 methylation. Residual liquid-based cytology samples were assessed for EPB41L3 and JAM3 methylation, and EPB41L3 and JAM3 methylation status was determined for discriminating CIN2+ from normal cervical findings/CIN1 (or low-grade intraepithelial lesions [LSIL]) according to known cervical histological pathology. The diagnostic accuracy of DNA methylation was compared with that of hrHPV-based strategies.

Ethical approval
The Institutional Review Board from the study center approved the study (No. JS-1954). All patients provided their consent before enrollment. The registration number is NCT03961191 (clinicaltrials.gov, registered on May 23, 2019). All procedures performed in the study involving human participants were in accordance with the ethical standards of the institutional and National Research Committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

Study design
This study is to develop an assay of EPB41L3 and JAM3 methylation for detection of CIN2+. A review of data regarding abnormal cytology and/or hrHPV testing results for patients who visited outpatient clinics from March 2018 to March 2019 was performed. We attempted to find 200 cases of inflammation or CIN1, 200 cases of CIN2/3, and 200 cases of cervical carcinomas diagnosed by histological pathology. After registration of the study in May 2019, the eligible patients were asked to return to the outpatient clinics to sign consent forms to participate in the study, and their residual cytology samples were sent for DNA methylation analysis and hrHPV genotyping. The primary endpoints were the cutoff values of EPB41L3 and JAM3 methylation and their diagnostic accuracies regardless of hrHPV status in liquid-based cytology specimens for CIN2+. The secondary endpoint was comparison of the diagnostic accuracies of various screening methods. However, this comparison would be confirmed in future validation set due to the great bias in this training set caused by enrollment of patients.

Patient enrollment and sample size
We planned to perform DNA methylation analysis of 300 cases, comprising three groups of 100 cases of inflammation/CIN1, CIN2/3 (HSIL), or cervical carcinoma. After considering factors including patient unwillingness to participate in the study and insufficient residual cytology samples for analysis, we planned to collect information on abnormal cytology and/or hrHPV for 600 cases from patient records. Data for epidemiological characteristics and medical history were obtained from medical records and supplemented by patient interviews. The inclusion criteria consisted of the following: aged 18 or older; abnormal cytology and/or hrHPV according to the criteria of the American Cancer Society, American Society for Colposcopy and Cervical Pathology, and American Society for Clinical Pathology screening guidelines [5] and its updated version [6] for further intervention; available residual cytology samples for methylation analysis in the study center; definitive cervical histological pathology findings in the study center; no history of precancerous cervical lesions or other cancers; no history of medical radiotherapy; negative HIV results and no history of organ transplantation or immunosuppressive therapy; and willingness to participate in the study. Cases not meeting all the criteria were excluded. In addition, to avoid bias, less common histological subtypes of cervical adenocarcinoma (ADC) were also excluded except for endocervical ADC or usual-type ADC, which is the most common ADC type and is regarded as HPV-associated tumor [32]. Patients with endometrial and ovarian cancer underwent methylation testing but were not included in the final analysis for cutoff values. All histological materials were re-evaluated by two pathologists (YY and HW).

Collection of study materials
The residual cytology samples were obtained from outpatient clinics of the Department of Obstetrics and Gynecology of the study center. These samples had previously been tested for cytology and/or hrHPV. All the samples were stored in PreservCyt Solution (Thinprep Pap Test; Hologic, USA) at room temperature. The cytology evaluation was performed using a Thinprep 2000 (Hologic, USA), and the results are reported according to the Bethesda 2014 system [33]. Primary hrHPV analysis utilized the Cobas 4800 System (Roche Molecular Systems, Inc., USA, only for HPV 16, 18, and others subtypes not specified). However, because primary hrHPV testing results were not available for all patients, the samples were subjected to hrHPV genotyping analysis as well as methylation analysis.
A 2-ml residual cytology sample was collected. Genomic DNA was extracted with the TIANamp Genomic DNA kit (Tiangen Biotech Co., Ltd, China), and the concentration was measured with a Nanodrop-300 microspectrophotometer (Thermo Fisher Scientific Inc., USA). Modification of the isolated DNA was performed using EZ-96 DNA Methylation-Lighting TM MagPrep (Zymo Research CO., USA) according to the manufacturer's instructions. The standard conversion amount of genomic DNA was 1 μg. The DNA yield was less than 1 μg (23% of the total), and a minimal input of 70 ng was used. The bisulfite-converted DNA was eluted with 40 μL elution buffer and used as the template for polymerase chain reaction (PCR).

DNA methylation testing
Methylation of the EPB41L3 and JAM3 genes was evaluated using TaqMan-based technologies with the Methylated Human EPB41L3 and JAM3 Gene Detection kit (real-time fluorescent PCR) (Beijing SinoMDgene Technology Co., LTD, China) and an ABI 7300 Real Time Fluorescence Quantitative PCR system (Life Tech, USA). According to the manufacturer's recommendations, PCRs were performed in a total volume of 25 μl, containing 15 μl of methylation-specific PCR mix, 5 μl of bisulfite-converted DNA, and optimized concentrations of primers and probes. Leukocyte DNA and HeLa DNA treated with sodium bisulfite were used as negative and positive controls, respectively. Moreover, a nontemplate control was tested in each run to monitor the PCR specificity. The PCR conditions were as follows: 96°C for 10 min, followed by 45 cycles at 94°C for 15 s, 64°C for 5 s, and 60°C for 30 s.
Three sets of primers and probes were designed using Primer Premier 5: one set for GAPDH, which does not contain CpGs, with both methylated and unmethylated sequences amplified equally, and was used as an internal reference to evaluate total bisulfite conversion; two sets for the candidate genes EPB41L3 and JAM3, specifically amplifying the methylated locus of interest. For each sample, the methylation level of each gene was determined by the ΔCt value (ΔCtE = Ct EPB41L3 − Ct GAPDH ; ΔCtJ = Ct JAM3 − Ct GAPDH ). Raw results were exported from the system, and the ΔCt value was calculated. If no amplification for EPB41L3 or JAM3 occurred, the Ct value was regarded as 45. When the ΔCt value for at least one of the targets was below its cutoff which was determined as described later, the sample was considered "positive"; when both genes were above their cutoffs, the sample was considered "negative." A sample was regarded as "invalid" when the GAPDH Ct value was above its cutoff (≥ 35.75). The expression of DNA methylation in any of both genes is defined "positive." Genotyping of hrHPV hrHPV genotyping was performed with TaqMan-based technology using an ABI 7500 Real Time Fluorescence Quantitative PCR system (Life Tech, USA) or a Stratagene Mx3000p Fluorescence Quantitative PCR system (Stratagene, USA) with an HPV nucleic acid genotyping diagnostic kit (Real time Fluorescent PCR) (Beijing SinoMDgene Technology Co., Ltd, China). The diagnostic kit, a quantitative in vitro assay, detects a pooled result for hrHPV types, including HPV 16,18,31,33,45,52,6,11,35,51,39,59,68,56,58, and 66, with type-specific probes. Briefly, 10 ng DNA was added per well, denatured, and combined with a type-specific probes derived from highrisk HPVs. PCR was performed as follows: (1) reaction with the UDG enzyme at 37°C for 2 min; (2) initial denaturation at 95°C for 3 min; and (3) 11,18,33,51,52,56,59, and 66, respectively) no more than 36 were recorded as positive. A positive hrHPV plus methylation status indicates hrHPV positivity as well as methylation positivity for EPB41L3 or JAM3 (or both).

Statistics
The cutoff values of DNA methylation were calculated with a receiver operating characteristic curve (ROC) and Youden index analysis (specificity + sensitivity − 1) based on comparison between cervical inflammation/ CIN1 and CIN2 or more severe cervical lesions. Nonnormally distributed variables and categorical data were compared between different screening groups by using nonparametric tests. The specificity, sensitivity, negative predictive value (NPV), and positive predictive value (PPV) in various screening groups were also calculated. The odds ratio (OR) and 95% confidence interval (95% CI) of positive ratios of different screening methods for various histological types were calculated with logistic regression models. Unless otherwise stated, all analyses were performed with a two-sided significance level of 0.05 and were conducted with the use of the Statistical Product and Service Solutions (SPSS) Statistics 20.0 software (IBM Corporation, Armonk, USA).

Patient characteristics
A flow diagram of the study is presented in Fig. 1

Diagnostic accuracies of various screening methods
The ORs for the positive ratios of different screening methods for various histological types are listed in Table 2.
With inflammation as the reference, significantly higher ratios for a positive methylation status were found for CIN3 ( Table 3. As the entire study cohort had abnormal cytology and/or hrHPV testing results, comparisons of sensitivity and specificity were limited to methylation status and methylation plus hrHPV status. These two screening methods exhibited a similar sensitivity (72.13% and 67.76%, p = 0.362) and specificity (91.53% and 94.92%, p = 0.300). Conversely, cytology alone and hrHPV displayed very poor specificities of 3.39% and 25.42%, respectively.
In 49 cases with negative hrHPV results, positive methylation alone was able to differentiate CIN2+ from inflammation/CIN1 (

Discussion
In this cohort, we defined cutoff values for EPB41L3 and JAM3 to differentiate CIN2+ from other cervical lesions. The selection of specific DNA was based on previous studies, which revealed the good performance of EPB41L3 [20,34] and JAM3 [25]. The targeted CpG sites of methylation from EPB41L3 and JAM3 genes locate in the promoter and exon 1, respectively. DNA methylation of these two genes showed areas under the ROC curve of 0.846 and 0.863, respectively, which were similar to or better than previous reports of EPB41L3 and JAM3 and other genes for the discrimination of HSIL+ from <  [19,35,36]. Moreover, the combined methylation status had a good sensitivity, specificity, PPV, and NPV. Because we selected a population with abnormal cytology and/or hrHPV findings, analysis of the diagnostic accuracy of hrHPV and/or cytology would not reflect the real-world situation. Nonetheless, methylation of EPB41L3 and JAM3 with or without hrHPV testing achieved the best specificities and PPVs and similar NPVs to those of hrHPV status. It has been reported that there is no association between the methylation of any gene and the presence of human papillomavirus [35]. In our study, DNA methylation had good discrimination even in patients with negative hrHPV results (OR 4.727, p = 0.027). The triage capacity of DNA methylation in our study was in accordance with previous reports [28,35]. However, studies on various DNA methylation triage methods in the Chinese population have reached conflicting conclusions [37][38][39]. These discrepancies reflect bias associated with the selected study populations and candidate genes. Based on these findings, we suggest that DNA methylation can serve as an independent screening method for CIN2+ lesions. Despite describing other types of DNA methylation, there are several supporting studies [17,28,30,[40][41][42]. In the study by Boers et al., DNA methylation of a combination of genes (C13ORF18/JAM3/ANKRD18CP) had higher specificity than hrHPV after a positive Pap smear and had comparable diagnostic accuracies to those of Pap smear in hrHPV-positive scrapings [28]. Their findings together with others suggest that DNA methylation testing may constitute a replacement of HPV DNA testing altogether with only modest improvements in test technology [13].
Methylation assays are relatively easy to set up, perform, and automate. Furthermore, DNA methylation assays exhibit competitive performance with other current triage options at the forefront of reflex triage tests for women who are positive for hrHPV [13]. Testing can even be performed directly from self-collected specimens [14]. In our study, methylation testing achieved a success rate of 98.4% (306/311) when using residual cytology samples stored in the previous year. The good quality of the assay also guarantees the independent utilization of DNA methylation for cervical cancer screening. Regardless, as Lorincz et al. [14] noted in 2014, epigenetic biomarkers need to be considered within the context of differential diagnostic situations and different independent sources, with all the strengths and limitations of the compared tests in full view. Because we achieved favorable results in this exploratory study, we initiated a validation trial among patients with various definitive gynecological pathologies (NCT03960879 study registered at clinicaltrials.gov) in the study center to confirm our hypothesis.
We found that the severity of cervical lesions was associated with the proportion of DNA methylation, and these data were significant for CIN3, cervical SCC, endocervical ADC, and even other gynecological carcinomas ( Table 2). These findings are consistent with previous reports [31,43]. The combination of methylation and hrHPV also showed excellent discriminating capacity. However, compared with inflammation or CIN1, a significantly increased proportion of hrHPV positivity was found among CIN3 and SCC cases but with similar ratios among endocervical ADC and other malignancies. EPB41L3 methylation and HPV types in CIN1 suggest that progression from a normal epithelium to CIN1 or CIN3 is usually promoted by the same HPV type but occurs via distinct DNA epigenotypes [44].
In our study, DNA methylation was better than hrHPV for discriminating cervical ADC. Patients with cervical ADC often present a series of characteristic clinical features, such as a high positive lymph node rate, distant metastasis rate, and recurrence rate, corresponding to a poor prognosis [45,46]. A much lower prevalence of hrHPV is found in ADC than in squamous carcinomas, 100% of which are positive for HPV infection [47][48][49][50][51]. HPV-negative endocervical ADCs of the usual type vary in frequency from 4.8 to 40.0% across China [50,52], and early diagnosis of HPV-negative cervical ADCs is still challenging [53]. The favorable discrimination ability of DNA methylation in our study indicates that it is probably an applicable and feasible method for the detection and diagnosis of ADC. Nevertheless, the entire cohort of ADC in our study was too small to achieve conclusions as definite as the situations in CIN2+ and SCC. For other types of ADC, such as ADC in situ or special subtypes (mucinous, endometrioid, etc.), the role of DNA methylation needs to be verified in our validation trial (NCT03960879). SCC and ADC may also have different specific DNA methylation types [54,55], and these differences deserve further exploration.
In the current study, we determined the cutoff values for EPB41L3 and JAM3 methylation for differentiating CIN2+ from other cervical lesions. First, as for all patients with abnormal cytology and/or hrHPV results, the diagnostic accuracies of cytology, hrHPV, and cytology plus hrHPV were limited by selection bias in this cohort, and the accuracies of DNA methylation need to be reliably reproduced. Table 3 just provided an example for this comparison, and a validated result is still expected in an unbiased trial (NCT03960879), which is still going on. Second, DNA methylation testing was not performed for HIV-positive patients or patients with ADC other than the endocervical subtype; regardless, previous studies of immunocompromised populations have shown that DNA methylation has an essential impact on the detection of CIN2+ [56,57]. Third, all cervical histological pathology results in this study were based on cervical biopsies, which may involve small but maybe essential differences compared to the final pathology results, likely leading to accurate analysis and bias. Last, we did not perform a follow-up of the prognosis of the patients, limiting the interpretation of the DNA methylation data with regard to the carcinogenesis and progression of cervical cancer, as demonstrated in other studies [58]. These limitations may be resolved in future validation trials.

Conclusions
In this exploratory study, methylation of EPB41L3 plus JAM3 had a similar diagnostic accuracy to that of hrHPV for detecting CIN2+. DNA methylation may be an alternative screening method. A further validation trial is needed to confirm these findings.
Additional file 1:. Supplementary Table 1. Raw data for the study. Acknowledgements None

Statement of submission
The paper is not under consideration by another journal, and the results presented in this work have not been previously presented or published.
Authors' contributions PL and LL conceived of the original idea for the study, interpreted the results, carried out the statistical analysis, edited the paper, and were the overall guarantors. LK, LW, ZW, XX, and LL obtained ethical approval, collected the study samples, contributed to the preparation of the data set, interpreted the results, and contributed to drafts of the paper. LW, ZW, and PL carried out the DNA methylation analysis. YY and HW conducted the pathological evaluation. XX and MW contributed to the study design, interpreted the results, and commented on drafts of the paper. All authors have approved the final version of the manuscript.

Funding
This study is supported by the Chinese Academy of Medical Sciences Initiative for Innovative Medicine (CAMS-2017-I2M-1-002). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Availability of data and materials
All data for this study are contained in the supplement file.

Ethics approval and consent to participate
The Institutional Review Board of Peking Union Medical College Hospital approved this study (No. JS-1954). The registration number is NCT03961191 (clinicaltrials.gov, registered on May 23, 2019).

Consent for publication
Consent for publication was obtained from all patients.