PAX1 and ZNF582 Methylation as a Triage Strategy in Cytologic and hrHPV+ Scrapings in Both Chinese and Dutch Cohorts

Abstract

Background:

The methylation of paired boxed gene 1 (PAX1) and zinc finger protein 582 (ZNF582) has been proved to have good diagnostic efficacy to detect cervical precancerous lesions in Asian populations, but their performance in Dutch populations is still unclear. We assessed PAX1 and ZNF582 methylation in both Dutch and Chinese cohorts.

Methods:

Cervical scrapings were evaluated from 2 independent groups. The Dutch cohort (population-based screening) provided 190 abnormal cytology scrapings, and 138 cases tested positive for high-risk human papillomavirus (hrHPV). The hospital-based Chinese cohort included 246 specimens (205 abnormal cytology and 227 hrHPV+). Quantitative methylation-specific PCR was employed to analyze methylation.

Results:

PAX1 and ZNF582 methylation showed high specificity of 91% to 97% with moderate sensitivity of 43% to 61% for cervical intraepithelial neoplasia (CIN) 2 or worse (CIN2+) in abnormal cytology and hrHPV+ samples in both cohorts. Sensitivity for CIN3+ increased to 55% to 86% but with lower specificity (85%–93%). For detecting CIN2+, PAX1/ZNF582 methylation performed similarly in Dutch and Chinese cohorts. However, for CIN3+, the Chinese cohort showed significantly higher sensitivity, with comparable specificity. We also evaluated combining these genes with our established markers (C13ORF18, JAM3, and ZSCAN1). The best diagnostic performance was achieved by combining PAX1 or ZNF582 with JAM3 in the Dutch cohort.

Conclusions:

In conclusion, regardless of race or geography, methylation of PAX1 or ZNF582 is highly specific for detecting CIN2+/CIN3+ lesions in hrHPV+ or cytologically abnormal scrapings. Triage testing using these markers might significantly reduce immediate colposcopies.

Impact:

PAX1/ZNF582 methylation triage for cytology+/hrHPV+ scrapings may reduce immediate colposcopies.

Introduction

Marked decreases have been observed in both the incidence and death rates of cervical cancer; cervical cancer continues to be the fourth most common cancer in women with an estimated 660,000 new cases and 350,000 deaths worldwide in 2022 (1). Due to the lack of healthcare infrastructure, inadequate organized screening, and human papillomavirus (HPV) vaccination programs, the burden is more pronounced in low- and middle-income countries (1–4). In 2020, the World Health Organization (WHO) launched the global Cervical Cancer Elimination Initiative, aiming to reduce incidence rates to less than 4 per 100,000 women-years in this century. Continuous and improved surveillance and screening for cervical cancer have been emphasized as a fundamental step for the WHO elimination strategy (5).

Many countries utilize cytology for cervical cancer screening. Although cytology-based programs have reduced cancer incidence in high-income regions, their moderate sensitivity (∼60%) for precancerous lesions leaves many cases undetected (6, 7). In addition, cytology requires costly infrastructure, including high-quality laboratories and trained cytologists, to ensure proper diagnosis and efficient screening flow (8).

High-risk HPV (hrHPV) infection plays an essential role in the pathogenesis of cervical cancer (9). Screening with the HPV test was more effective in preventing cervical cancers. According to the global strategy aimed at eliminating cervical cancer, the WHO recommends primary HPV testing as the preferred screening method (10, 11). The pivotal evidence supporting this change came from randomized controlled trials showing that HPV-based screening leads to a greater reduction in the incidence and mortality of cervical cancer compared with cytology-based screening (12). The high frequency and self-limiting nature of HPV infection contribute to its low specificity and low positive predictive value (PPV), requiring triage testing to effectively detect women with actual disease. Various triage tests, including cytology, p16/Ki-67 dual staining, and HPV16/18 genotyping, have been suggested, but these approaches remain limited by issues such as subjective interpretation, low specificity, and low PPV (13–15). Of concern is the reported infliction of avoidable medical and psychologic burdens when relying only on hrHPV testing (16, 17). Consequently, the strategy of primary hrHPV testing followed by cytology triage has led to an observed increase in colposcopy referral rates (∼9%, as observed in the Dutch screening program) compared with initial cytologic screening (∼5%; ref. 10). This trend is corroborated by findings from other European programs, such as pilots in the United Kingdom and Italy (12, 18). Therefore, there is a demand for additional triage methods and/or ancillary biomarkers to complement primary hrHPV screening. In addition, maintaining standardization and reproducibility calls for more robust techniques, especially in resource-limited regions.

DNA methylation represents an epigenetic feature that serves as a mechanism for regulating gene expression (19, 20). The CpG island hypermethylation in oncogene promoters occurs early during the development of many tumors, such as cervical cancer (21–23). Various genome-wide methods have been employed to uncover new genes with selective methylation associated with cervical cancer and its high-grade premalignant precursors (24–30). In published systematic reviews, the paired boxed gene 1(PAX1) and zinc finger protein 582 (ZNF582) were found to be the most researched methylation markers for identifying early cervical cancer in Asian women (31, 32). Studies conducted in Taiwan have shown that in cervical scrapings, PAX1 and ZNF582 hypermethylation correlates with cervical cancer and were proposed as candidate indicators for cervical intraepithelial neoplasia (CIN) 3+ detection (33–42). However, the majority of these investigations were conducted within Asian populations, with limited information available on the use of PAX1/ZNF582 methylation for triaging hrHPV+ cases. The observed heterogeneity in biomarker accuracy may be attributed to differences in clinical referral pathways, screening history, or underlying genetic/epigenetic variations between populations.

The diagnostic efficacy of PAX1 and ZNF582 for cervical precancerous lesions in Dutch populations is still unclear.

This study was designed to evaluate and compare, in Dutch and Chinese cohorts, the possible value of PAX1 and ZNF582 methylation in identifying women at risk for cervical cancer, with the purpose of determining their clinical efficacy via quantitative methylation-specific PCR (QMSP) by determining sensitivity and specificity for detecting CIN2+ and CIN3+ lesions.

Materials and Methods

Patient selection

Two distinct cohorts, one from the Netherlands and one from China, were utilized in this study. With regard to the Dutch cohort (n = 190), cervical scrapings were gathered prospectively from referrals of women to the University Medical Center Groningen (UMCG) colposcopy outpatient clinic, pursuant to an abnormal smear identified through the general population screening program, as documented earlier (24, 25). Pretreatment cervical scrapings were collected, whereas clinicopathologic data were extracted from patient records and stored anonymously in a password-secured institutional database dedicated to gynecologic oncology. Exclusion criteria comprised pregnancy, absence of a complete uterus, and a history of CIN or cervical cancer treatment.

The Chinese cohort (n = 246) consisted of gynecology clinic attendees at Tianjin Central Hospital of Gynecology Obstetrics between March 2017 and February 2018. Inclusion required either an abnormal smear or a positive hrHPV test. The general features are analogous to those observed in the Dutch cohort, as detailed previously (42).

Cytology results were categorized using the 2014 Bethesda System (TBS 2014). For histologic analysis/definite treatment, either colposcopy-guided biopsies or loop excision of the transformation zone were undertaken in line with national standard procedures. The definitive diagnosis relied on histologically confirmed tissue findings, with all patients managed according to national guidelines. Participant recruitment and clinical data gathering procedures were subject to periodic monitoring.

Prior authorization for the ‘methylation studies’ was obtained from the ethics committees at UMCG and Tianjin Central Hospital of Gynecology Obstetrics. All patients contributing biological material provided explicit written informed consent.

HrHPV detection

For the Dutch cohort, hrHPV testing was performed using PCR (GP5+/6+) followed by cobas 4800 HPV PCR (Roche Molecular Systems), as previously described (43), whereas the Chinese cohort was tested solely with the cobas 4800 hrHPV test (Roche Molecular Systems). A positive hrHPV result was defined as detection of HPV16 or HPV18 or any of the remaining 12 hrHPV genotypes (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68; ref. 42).

DNA preparation and bisulfite conversion

The procedures for cervical scraping processing, DNA extraction, and DNA integrity assessment adhered to protocols reported earlier (44). The DNA concentration was assessed using the NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific). Inclusion in this study required that samples demonstrate the ability to produce amplicons of 300 base pairs (bp) or longer, based on DNA structural integrity. The EZ DNA Methylation Kit (Zymo Research) and the EZ DNA Methylation-Gold Kit (Zymo Research) were used for sodium bisulfite conversion of denatured genomic DNA in the Dutch and Chinese cohorts, respectively, following a previously described procedure (45, 46). Following the manufacturer’s protocol, 1 μg DNA per sample underwent sodium bisulfite conversion and was then eluted to achieve a final concentration of 10 ng/μL. Samples were initially randomized across DNA purification batches and then rerandomized across subsequent bisulfite reactions. Leukocyte DNA from healthy women served as the negative control, whereas leukocyte DNA methylated, produced in vitro with M.SssI methyltransferase (New England Biolabs), was used as a positive control for methylation according to a prior report (45).

QMSP

Methylation of PAX1 and ZNF582 was quantified relative to ACTB (a methylation-independent reference) using previously reported methods (24–26, 45). The primer and probe sequences for PAX1 and ZNF582 primers were adopted from a previous description (39, 47). The ZNF582 probe, designed using Clone Manager Professional 9 to target the region from −300 bp to −100 bp upstream of the transcription start site, is accessible upon request. Methylation levels for each sample were assessed randomly and blinded to clinical information. Sodium bisulfite conversion was duplicated for any sample with a mean ACTB quantity below 1 ng per reaction. QMSP was carried out using an ABI PRISM 7900HT Sequence Detection System (Thermo Fisher Scientific). Each QMSP reaction of 10 μL contained 300 nmol/L of each primer, 200 nmol/L probe, QuantiTect Probe PCR Master Mix (Qiagen), and 2.5 μL bisulfite-converted DNA (approximately 25 ng). All samples were analyzed in triplicate. Absolute quantification of methylated template was achieved using a dilution series generated from in vitro methylated leukocyte DNA. The relative methylation level of the target region was calculated as average quantity of the methylated region of interest divided by average quantity of the reference ACTB gene and multiplied by 10,000 (45). Cutoff values for PAX1 and ZNF582 was established by maximizing Youden index (sensitivity + specificity −1) for identifying CIN2+ lesions. Criteria for methylation positivity were as follows: for PAX1, a signal with a Ct value below 50 in no fewer than 2 of 3 reaction wells; for ZNF582, an average methylation level exceeding 45.3.

To evaluate the impact on sensitivity/specificity using alternative methylation markers, we utilized data from 6 markers (ANKRD18CP, C13ORF18, EPB41L3, JAM3, SOX1, and ZSCAN1) determined in our prior study (42). The panel examined in the current investigation was selected based on our own research line.

Sample acquisition and handling, DNA isolation, bisulfite conversion, and QMSP analysis were all carried out by the same scientist (LN) at both study sites: UMCG in the Netherlands (for the Dutch cohort) and Tianjin Central Hospital of Gynecology Obstetrics in China (for the Chinese cohort), using identical protocols for DNA extraction and QMSP. All analyses and interpretation of QMSP data obtained at UMCG and Tianjin Central Hospital of Gynecology Obstetrics were performed by the same scientist (LN) and same supervisors (GBAW/ES).

Statistical analysis

Statistical analyses were conducted using IBM SPSS Statistics 22 (IBM Corporation). Overall differences in methylation levels among histologic categories were evaluated using the Kruskal–Wallis test. Two endpoints were evaluated: (i) CIN2 or worse (CIN2+) and (ii) CIN3 or worse (CIN3+), from which sensitivity and specificity were subsequently derived. Differences in sensitivity or specificity between populations were analyzed using the Fisher exact test. The highest Youden index was used to identify the optimal marker panels. For gene combinations, a sample was classified as positive when any single marker tested positive. Statistical significance was defined as a P value of less than 0.05.

Results

Study population features

Table 1 presents the histologic data along with available information from hrHPV and cytology results for both study cohorts. In the Dutch cohort, 190 cervical scrapings were obtained from women with the following diagnoses: no CIN (n = 28), CIN1 (n = 30), CIN2 (n = 40), CIN3 (n = 44), and cervical cancer (n = 48). Among the cancer cases, histotypes included squamous cell carcinoma (SCC, n = 31), adenocarcinoma (ADC, n = 13), adenosquamous cancer (n = 3), and undifferentiated cancer (n = 1). HrHPV DNA was found in 138 of these women.

Table 1.

Clinicopathologic data of patients with scrapings and the relative available hrHPV testing and cytology testing results.

Characteristics	Dutch cohort (n = 190)					Chinese cohort (n = 246)
	Normal	CIN1	CIN2	CIN3	Cancer	Normal	CIN1	CIN2	CIN3	Cancer
Number of subjects	​	​	​	​	​	​	​	​	​	​
N	28	30	40	44	48	76	35	48	51	36
Age, years	​	​	​	​	​	​	​	​	​	​
Mean age	42	38	36	36	45	40	39	39	43	49
Median age	44	39	35	35	44	37	37	38	42	50
Range	26–60	23–50	25–60	27–50	29–80	25–70	22–66	22–59	18–63	24–67
Cytology	​	​	​	​	​	​	​	​	​	​
Normal	0	0	0	0	0	36	1	2	2	0
ASC-US	6	8	1	0	0	23	19	28	21	10
LSIL	13	20	32	13	5	15	12	14	9	1
ASC-H	0	0	0	0	0	0	0	0	0	2
HSIL	0	1	7	31	31	2	3	4	17	10
Cancer	0	0	0	0	3	0	0	0	2	13
Not known	9	1	0	0	9	0	0	0	0	0
hrHPV	​	​	​	​	​	​	​	​	​	​
HPV16/18+	8	8	21	30	32	21	7	31	37	31
Other positive	4	13	13	5	4	45	25	15	12	3
Negative	16	9	6	9	12	10	3	2	2	2

Abbreviations: ASC-H, atypical squamous cells, cannot exclude HSIL; ASC-US, atypical squamous cells of undetermined significance; HSIL, high-grade squamous intraepithelial lesion; LSIL, low-grade squamous intraepithelial lesion.

As detailed in our previous study (39), the Chinese cohort comprised 246 scrapings subjected to methylation analysis: no CIN (n = 76), CIN1 (n = 35), CIN2 (n = 48), CIN3 (n = 51), and cervical cancer (n = 36). Cancer cases consisted of SCC (n = 30), ADC (n = 4), adenosquamous cancer (n = 1), and clear-cell carcinoma (n = 1). Of these 246 women, 205 presented with abnormal cytology and 227 were hrHPV+.

Assessment of PAX1 and ZNF582 methylation in cytologically abnormal cervical scrapings

To explore the impact of DNA methylation analysis in cytologically abnormal cervical scrapings, we analyzed PAX1 and ZNF582 methylation in the Dutch (n = 190) and Chinese cohorts (n = 205). Relative levels of methylation were different among different histologic groups (P < 0.001; Fig. 1A–D).

Figure 1.

Methylation ratio of PAX1 (A and B) and ZNF582 (C and D) tested with QMSP in abnormal cytology scrapings from patients with no CIN, CIN1, CIN2, and CIN3 lesions and cancer. Relative levels of methylation were different among different histologic groups (P < 0.001). The horizontal lines represent the optimal threshold (45.3).

In the Dutch cohort, 31 of 48 (65%) cancers tested positive for PAX1 methylation, whereas all samples from women without CIN lesion were PAX1−. PAX1 methylation positivity was substantially greater in the Chinese cohort, affecting 35 of the 36 cervical cancer cases (97%; P < 0.001). PAX1 positivity rates were comparable between the 2 cohorts for CIN2+ (Dutch: 57/132; Chinese: 69/131) and for no CIN/CIN1 women (Dutch: 3/58; Chinese: 4/74). As shown in Table 2, PAX1 methylation demonstrated equally high specificity (both 95%) for detecting CIN2+ lesions, whereas the detection rate remained modest (Dutch: 43%; Chinese: 53%). For CIN3+ detection, sensitivity increased to 55% in the Dutch cohort and reached 75% in the Chinese cohort, whereas high specificity was maintained (Dutch: 91%; Chinese: 93%). With regard to ZNF582 methylation, positivity among the Dutch cancer cases was 38 of 48 (79%), whereas the Chinese cohort showed a significant higher positivity of 100% (36/36 cancers; P = 0.002). In the Dutch cohort, no CIN− samples tested positive for ZNF582, whereas only 1 of 40 such samples was positive in the Chinese cohort. ZNF582 positivity rates for CIN2+ lesions (Dutch: 67/132; Chinese: 80/131) and for no CIN/CIN1 cases (Dutch: 2/58; Chinese: 5/74) were comparable between the 2 cohorts (Fig. 1A–D). For CIN2+ detection, sensitivity and specificity were 51% and 97%, respectively, in the Dutch cohort compared with 61% and 93% in the Chinese cohort. For identifying CIN3+, ZNF582 methylation yielded a sensitivity of 63% and specificity of 89% in the Dutch cohort compared with 86% and 90% in the Chinese cohort (Table 2).

Table 2.

Sensitivity and specificity of methylation markers in scrapings with an abnormal cytology from Dutch and Chinese cohorts.

Methylation markers	Dutch cohort (n = 190)						Chinese cohort (n = 205)
	CIN2+ vs. CIN2−			CIN3+ vs. CIN3−			CIN2+ vs. CIN2−			CIN3+ vs CIN3−
	Sens (%)	Spec (%)	(YI + 1) (%)	Sens (%)	Spec (%)	(YI + 1) (%)	Sens (%)	Spec (%)	(YI + 1) (%)	Sens (%)	Spec (%)	(YI + 1) (%)
Individual markers	​	​	​	​	​	​	​	​	​	​	​	​
Newly tested genes	​	​	​	​	​	​	​	​	​	​	​	​
PAX1	43	95	138	55	91	146	53	95	148	75	93	168
ZNF582	51	97	148	63	89	152	61	93	154	86	90	176
Known genes	​	​	​	​	​	​	​	​	​	​	​	​
C13ORF18	42	95	137	51	89	140	41	99	140	58	96	154
JAM3	64	93	157	75	81	156	67	78	145	88	76	164
ZSCAN1	53	95	148	65	87	152	68	82	150	89	78	167
Methylation marker panels	​	​	​	​	​	​	​	​	​	​	​	​
PAX1/ZNF582	52	95	147	65	88	153	62	92	154	86	88	174
PAX1/JAM3	64	90	154	76	79	155	68	78	146	89	76	165
ZNF582/JAM3	64	91	155	76	80	156	71	78	149	92	74	166
PAX1/JAM3/C13ORF18	66	86	152	77	76	153	69	78	147	89	74	163
ZNF582/JAM3/C13ORF18	66	88	154	77	77	154	73	78	151	92	73	165
PAX1/JAM3/ZSCAN1	64	90	154	76	79	155	76	69	145	92	63	155
ZNF582/JAM3/ZSCAN1	65	90	155	77	79	156	76	69	145	92	63	155
PAX1/ZSCAN1	55	95	150	66	86	152	69	81	150	89	76	165
ZNF582/ZSCAN1	57	95	152	68	85	153	69	81	150	89	76	165
PAX1/C13ORF18/ZSCAN1	58	91	149	71	83	154	70	81	151	89	75	164
ZNF582/C13ORF18/ZSCAN1	60	91	151	72	82	154	70	81	151	89	75	164
PAX1/C13ORF18	52	91	143	65	86	151	56	93	149	78	90	168
ZNF582/C13ORF18	57	93	150	70	85	155	63	93	156	86	88	174
PAX1/C13ORF18/JAM3/ZSCAN1	66	86	152	77	76	153	76	69	145	92	63	155
ZNF582/C13ORF18/JAM3/ZSCAN1	67	86	153	78	76	154	77	69	146	92	62	154

NOTE: Using a threshold for positivity at a methylation ratio of 45.3 for ZNF582 or 132 for ZSCAN1.

Abbreviations: Sens, sensitivity; Spec, specificity; Youden index, YI.

No significant difference was observed between the 2 cohorts in the test accuracy of PAX1 and ZNF582 methylation for detecting CIN2+ (P = 0.1392 for PAX1; P = 0.1068 for ZNF582). However, for CIN3+ detection, the Chinese cohort showed significantly higher sensitivity (P = 0.0072 for PAX1; P ≤ 0.001 for ZNF582) than the specificity-matched Dutch cohort (Table 2; Fig. 2A and B).

Figure 2.

Sensitivity (Sens) and specificity (Spec) of methylation markers for CIN2+ (A) and CIN3+ (B) detection in samples with an abnormal cytology and for CIN2+ (C) and CIN3+ (D) detection in hrHPV+ samples. *Significant differences in sensitivity or specificity between the Chinese and Dutch cohorts are indicated (*, P < 0.05).

Methylation status of PAX1 and ZNF582 among hrHPV+ women

As hrHPV testing is increasingly becoming the primary screening test for cervical cancer, we assessed PAX1 and ZNF582 methylation in hrHPV+ specimens. In both cohorts, the relative levels and positivity rates increased significantly with greater histologic severity (P < 0.001; Fig. 3A–D).

Figure 3.

Methylation ratio of PAX1 (A and B) and ZNF582 (C and D) analyzed with QMSP in hrHPV+ scrapings from patient with no CIN, CIN1, CIN2, and CIN3 lesions and cancer. Relative levels of methylation were different among different histologic groups (P < 0.001). The horizontal lines represent the optimal threshold (45.3).

Although the Dutch cohort showed a lower rate of PAX1 positivity among cancers (27/36, 75%) compared with the Chinese cohort (33/34, 97%), no CIN− lesions in the Dutch cohort (0/12) tested positive for PAX1 methylation. Across both cohorts, PAX1 methylation positivity was similar for CIN/CIN1 (Dutch: 3/33; Chinese: 6/98) and CIN2+ patients (Dutch: 52/105; Chinese: 66/129).

ZNF582 methylation was positive in every patient with cancer from the Chinese cohort, whereas 3 cases in the Dutch cohort were negative for this marker. All CIN− samples in the Dutch cohort tested negative for ZNF582 (Fig. 3). Comparable positivity rates were observed between the 2 cohorts for CIN2+ (Dutch: 60/105; Chinese: 77/129) and no CIN/CIN1 scrapings (Dutch: 2/33; Chinese: 6/98).

In both cohorts, PAX1 and ZNF582 methylation performed similarly to abnormal cytology for triaging hrHPV+ samples (Table 3; Fig. 2C and D).

Table 3.

Sensitivity and specificity of methylation markers in hrHPV+ samples in Dutch and Chinese cohorts.

​Methylation markers	Dutch cohort (n = 138)						Chinese cohort (n = 227)
	CIN2+ vs. CIN2−			CIN3+ vs. CIN3−			CIN2+ vs. CIN2−			CIN3+ vs. CIN3−
	Sens (%)	Spec (%)	(YI + 1) (%)	Sens (%)	Spec (%)	(YI + 1) (%)	Sens (%)	Spec (%)	(YI + 1) (%)	Sens (%)	Spec (%)	(YI + 1) (%)
Individual markers	​	​	​	​	​	​	​	​	​	​	​	​
Newly tested genes	​	​	​	​	​	​	​	​	​	​	​	​
PAX1	50	91	141	65	87	152	51	94	145	73	92	165
ZNF582	57	94	151	73	85	158	60	94	154	84	91	175
Known genes	​	​	​	​	​	​	​	​	​	​	​	​
C13ORF18	44	94	138	55	87	142	40	100	140	57	97	154
JAM3	70	97	167	83	78	161	67	76	143	87	74	161
ZSCAN1	59	91	150	75	82	157	67	82	149	88	78	166
Methylation marker panels	​	​	​	​	​	​	​	​	​	​	​	​
PAX1/ZNF582	59	91	150	76	84	160	60	92	152	84	89	173
PAX1/JAM3	70	91	161	85	75	160	67	76	143	88	74	162
ZNF582/JAM3	70	94	164	85	76	161	71	76	147	90	72	162
PAX1/JAM3/C13ORF18	71	88	159	85	72	157	69	76	145	88	72	160
ZNF582/JAM3/C13ORF18	71	91	162	85	73	158	72	76	148	90	71	161
PAX1/JAM3/ZSCAN1	70	91	161	85	75	160	75	70	145	90	65	155
ZNF582/JAM3/ZSCAN1	71	91	162	86	75	161	76	70	146	90	64	154
PAX1/ZSCAN1	61	91	152	76	81	157	68	81	149	88	76	164
ZNF582/ZSCAN1	63	91	154	77	79	156	68	80	148	88	76	164
PAX1/C13ORF18/ZSCAN1	63	88	151	77	78	155	69	81	150	88	76	164
ZNF582/C13ORF18/ZSCAN1	65	88	153	79	76	155	69	80	149	88	75	163
PAX1/C13ORF18	56	88	144	72	82	154	54	94	148	76	91	167
ZNF582/C13ORF18	62	91	153	77	81	158	61	94	155	84	90	174
PAX1/C13ORF18/JAM3/ZSCAN1	71	88	159	85	72	157	76	63	139	90	59	149
ZNF582/C13ORF18/JAM3/ZSCAN1	72	88	160	86	72	158	77	62	139	90	58	148
HPV16/18 genotyping	​	​	​	​	​	​	​	​	​	​	​	​
HPV16/18+	79	52	131	87	45	132	77	71	148	82	59	141

NOTE: Using a threshold for positivity at a methylation ratio of 45.3 for ZNF582 or 132 for ZSCAN1.

Abbreviations: Sens, sensitivity; Spec, specificity; Youden index, YI.

Performance of methylation marker panels for diagnosing abnormal cytology scrapings

To determine whether sensitivity could be improved without compromising specificity, we performed combined analysis of PAX1 and ZNF582. The PAX1/ZNF582 combination did not yield superior diagnostic value compared with either marker alone (P = 0.093 for PAX1; P = 0.705 for ZNF582).

We then combined PAX1 or ZNF582 with 6 other previously identified methylation markers (ANKRD18CP, C13ORF18, EPB41L3, JAM3, SOX1, and ZSCAN1; ref. 42). Optimal methylation marker panels for CIN2+ detection were chosen according to the maximum Youden index. In the Dutch cohort, for instance, adding JAM3 to the marker panel increased sensitivity by 13% to 20%, whereas specificity remained virtually unchanged (−4%). For CIN3+ detection, sensitivity increased by 20%, but specificity decreased by 12%. In contrast, none of the combination panels outperformed single PAX1 or ZNF582 marker in the Chinese cohort (Table 2; Fig. 2A and B).

Diagnostic utility of methylation panels for hrHPV+ samples

We further assessed the sensitivity and specificity of various PAX1 and ZNF582 combinations for detecting CIN2+ and CIN3+, restricting the analysis to hrHPV+ samples only (Table 3; Fig. 2C and D). The PAX1/ZNF582 combination did not provide superior diagnostic value compared with either marker alone (P = 0.05 for PAX1; P = 0.438 for ZNF582).

In both cohorts, combining PAX1 or ZNF582 with 1 or 2 of the 6 other markers generally resulted in higher sensitivity but lower specificity for CIN2+ and CIN3+ detection compared with each marker alone. An exception was observed only in the Dutch cohort, in which all combinations significantly improved sensitivity for CIN2+ (best for JAM3), whereas the specificity remained consistently high at ∼91% (Table 3). However, in the Chinese cohort, no combination panel outperformed single PAX1 or ZNF582 marker.

HPV16/18 genotyping has been suggested as a method for triaging women who test positive for hrHPV (48). In the Dutch cohort, certain methylation marker combinations (including PAX1 or ZNF582) achieved significantly higher specificity for detecting both CIN2+ and CIN3+ (P < 0.001) with comparable sensitivity (Table 3). In the Chinese cohort, several methylation panels showed similar sensitivity and specificity to HPV16/18 genotyping to detect CIN2+ (all P > 0.05), whereas for CIN3+, the methylation panels demonstrated significantly better specificity (all P < 0.05) compared with HPV16/18 genotyping.

Discussion

As far as we know, the methylation status of PAX1 or ZNF582 in cervical malignancies was mostly analyzed in Asian populations. In this study, we analyzed methylation of PAX1 and ZNF582 in both Chinese and Dutch cohorts. Methylation of PAX1 or ZNF582 demonstrated high specificity for detecting CIN2+/CIN3+ lesions among women with abnormal cytology and/or positive hrHPV. The diagnostic performance of PAX1 and ZNF582 methylation for detecting CIN2+ was very similar between the 2 cohorts. Although for CIN3+ detection, the Chinese cohort showed significantly higher sensitivity than the Dutch cohort with similar specificity. In the Dutch cohort, the optimal diagnostic performance was obtained by pairing PAX1 or ZNF582 with JAM3. The heterogeneity in test performance metrics between cohorts is most reasonably attributed to variations in epidemiologic and clinical contexts, as opposed to inherent differences in disease natural history.

Methylation of PAX1 and ZNF582 in cervical cancer was initially described by Lai and colleagues in a Taiwanese cohort (40, 49). According to existing literature assessing PAX1 methylation in cervical specimens, the weighted average sensitivity and specificity for CIN2+ detection were 61% (range: 36%–87%) and 81% (range: 85%–99%), respectively. For CIN3+ detection, the corresponding values were 73% (range 64%–87%) and 79%, respectively (range: 53%–93%; refs. 30, 35, 36, 39, 47, 49–55). Our meta-analysis found that all published studies on ZNF582 were conducted exclusively in China, yielding a pooled sensitivity of 71% for CIN3+ detection and a specificity of 81% (32). Only 3 studies provided extractable data on CIN2+ detection using ZNF582 methylation, which yielded the weighted mean sensitivity of 65% (range 63%–67%) and specificity of 85% (range 80%–89%; refs. 33, 40, 50). The present study demonstrated comparable sensitivity and specificity for both PAX1 and ZNF582 methylation.

Most published studies on PAX1 and ZNF582 methylation have used CIN3+ as the primary study endpoint. CIN2 and CIN3 lesions follow different natural histories, as the majority of CIN2 cases tend to regress on their own, with only about 16% potentially advancing to severe dysplasia or higher-grade disease (56). Thus, CIN2 probably represents a mixture of newly acquired infections and early-stage CIN3, yet it frequently receives intervention in routine clinical practice. Accordingly, our study assessed the diagnostic utility of methylation markers for detecting both CIN2+ and CIN3+ lesions.

Most of these studies were performed in Asian populations. In this study, PAX1 and ZNF582 methylation was assessed in both a Dutch cohort and a Chinese cohort by the same scientists using the same QMSP protocols enabling a direct comparison between the enrolled 2 cohorts (Dutch and Chinese). The 2 populations showed high consistency, and our findings were also consistent with published reports from other Asian cohorts (32, 33, 35, 36, 39, 47, 49–55). However, several limitations should be acknowledged in this study. The Chinese cohort was recruited from a hospital setting and therefore not generalizable to the general population. By contrast, the Dutch cohort only included women with abnormal cytology who were referred for colposcopy following population-based screening, meaning that hrHPV+ women with normal cytology were not included. Regardless of these differences, the methylation marker panels showed highly similar diagnostic performance across the 2 cohorts.

A major challenge in cervical cancer screening is identifying effective triage strategies after primary hrHPV testing to detect women who actually have disease. In 2017, the Netherlands was among the first nations to adopt hrHPV testing as its national population-based screening strategy for early cervical cancer detection, using cytology as the triage tool. Using cytology as a triage test comes with several drawbacks, including subjectivity, requirement for trained personnel, and poor accuracy on self-collected materials. In addition, cytologists may provide biased assessment when aware of hrHPV status, resulting in a higher proportion of false positives (10). In the Netherlands, an assessment of the initial years of population-based screening showed a high number of referrals per screening round, with only ∼60% of referred women diagnosed with CIN2+ lesion, indicating poor specificity (10, 57, 58). The methylation panels evaluated in our current study achieved significantly higher diagnostic performance (highest specificity of 94% in both cohorts) compared with traditional cytology used for triage, which has a specificity of ∼80% (58). Our results demonstrated similar triage performance in hrHPV+ women to the recently published studies (42, 59–62). In an effort to improve the sensitivity to ensure that most women with disease could be identified, we combined PAX1 or ZNF582 with other methylation markers that we identified previously (24–27, 42). We found that the combination methylation panel with JAM3 showed the best performance with higher sensitivity without losing specificity in the Dutch cohort.

Studies have shown that HPV genotype distribution varies by geographic region (62). In Asia, HPV52 and HPV58 are commonly found in invasive patients with cervical cancers, whereas in China HPV16 and HPV18 are the dominant types (63). Apart from cytology triage, HPV16/18 genotyping can also be used as a triage method for hrHPV+ women, but its efficacy has recently been reported to be discounted in the era of HPV vaccines (48, 64). We compared the triage performance of HPV16/18 genotyping with that of our methylation markers. In the Chinese cohort, HPV16/18 genotyping achieved 77% sensitivity and 71% specificity to detect CIN2+, consistent with findings reported by Bu and colleagues (65) from Guangdong, China (65% and 67%, respectively). Our methylation panels performed non-inferiorly to HPV16/18 genotyping for CIN2+ detection while achieving significant better specificities for the detection of CIN3+. Our findings indicate that specific methylation panels outperform hrHPV16/18 genotyping in hrHPV+ women. Our findings are consistent with the recently published studies which reported that PAX1/ZNF582 methylation had better accuracy than HPV16/18 for detecting CIN3+ in hrHPV+ women (42, 60). Although this study provides comparative performance metrics for PAX1/ZNF582 methylation and HPV16/18 genotyping within the same cohort, a prospective, randomized trial specifically designed as a direct comparison is needed to robustly evaluate their relative clinical effectiveness as triage strategies.

Few studies have analyzed these methylation panels in both European and Asian populations with cervical neoplasia to validate our findings (42). Recently, a commercial assay consisting of the markers PAX1 and JAM3 is launched showing a sensitivity and specificity of 85% and 89%, respectively, in a non-HPV16/18 population (66). Comparison with our data is hard as our non-HPV16/18 population is very small. However, diagnostic performance seems to be very similar. We did not exclude HPV-unknown/cytology+ patients to reflect a real-world clinical scenario in which HPV status is sometimes unavailable or unreported. Our current study was not based on a population-based screening cohort; therefore, we did not estimate prevalence-dependent metrics such as PPV/negative predictive value (NPV). Moreover, for biomarkers mechanistically involved in carcinogenesis such as PAX1/ZNF582 methylation, specificity may not be a stable parameter but rather vary inversely with disease prevalence (67). These considerations warrant caution when extrapolating our findings to screening populations. Due to the limited sample size, caution is warranted in interpreting the observed difference. It should be noted that the performance of the combined panel, in particular, may be subject to overfitting and its superiority over individual markers requires confirmation through large-scale, independent, prospective validation. Based on our current study, incorporating our methylation panels to triage hrHPV+ or abnormal cytology individuals could help reduce immediate referrals.

In conclusion, both Dutch and Chinese cohorts exhibited high consistency in PAX1 and ZNF582 methylation, suggesting that methylation patterns are independent of geographic and racial differences. Our findings demonstrate that PAX1 or ZNF582 (combined with our previously identified methylation markers in the Dutch cohort) may serve as a promising triage tool for women with abnormal cytology or a hrHPV+ scraping, potentially reducing the need for immediate colposcopies.

Data Availability

Data are provided within the article.

Authors’ Disclosures

N. Li reports methylation analyses not only for the markers described in this study but also for additional markers not included here, highlighting its potential for implementation in cervical cancer screening and for the detection of other types of cancer. A.G.J. van der Zee reports a patent for WO 2016/048138/PCT/NL2015/050650 licensed to CC Diagnostics (filed by UMCG). E. Schuuring reports grants from CC Diagnostics, Abbott Molecular, MSD, Northern Netherlands Cooperation Association/European Foundation for Regional Development (SNN/EFRO), and PentaBase, nonfinancial support from CC Diagnostics and committee for clinically essential targets (cieKNT)/Dutch Society of Pathology, and personal fees from QCMD, Roche, and committee for diagnostic assessment (cieBOD)/Dutch Society of Pathology outside the submitted work; a patent for WO 2016/048138/PCT/NL2015/050650 licensed to CC Diagnostics (filled by UMCG) and a patent for WO2009115615 A2 licensed to mdxHealth; and methylation analyses not only for the markers described in this study but also for additional markers not included here, highlighting its potential for implementation in cervical cancer screening and also for the detection of other types of cancer. G.B.A. Wisman reports nonfinancial support from CC Diagnostics, personal fees from Novosanis, and grants from CC Diagnostics, Abbott Molecular, and MSD outside the submitted work; a patent for WO 2016/048138/PCT/NL2015/050650 licensed to CC Diagnostics (filed by UMCG); and methylation analyses not only for the markers described in this study but also for additional markers not included here, highlighting its potential for implementation in cervical cancer screening and also for the detection of other types of cancer. No disclosures were reported by the other author.

Authors’ Contributions

N. Li: Formal analysis, investigation, methodology, writing–original draft. Y. Hu: Resources, writing–review and editing. A.G.J. van der Zee: Conceptualization, writing–review and editing. E. Schuuring: Conceptualization, resources, supervision, writing–review and editing. G.B.A. Wisman: Conceptualization, resources, supervision, methodology, writing–review and editing.