Higher levels of host-cell DNA methylation markers ZNF582 and ASCL1 on anal smears are predictive for progression to anal cancer in patients with previous high-grade lesions

Summary

Background

French and International anal cancer screening recommendations for at-risk populations, published in 2024, are based on cytology and/or high-risk human papillomavirus (HPV) detection on anal smears. Biological markers to triage the patients most at-risk for anal cancer are crucial in prioritising patients needing high-resolution anoscopy consultations, which are frequently overwhelmed.

Methods

The AIN3 cohort is a French national multicenter study including patients with a history of high-grade anal lesions (AIN3). Patients were followed-up for at least 3 years, with anal smears and clinical examinations performed yearly. Levels of ZNF582 and ASCL1 gene methylation were quantified using real-time PCR on anal smears collected at the time of inclusion.

Findings

Overall, 514 anal smears were contributive for host-cell DNA methylation analysis. Patients' mean age was 50.8 years and 40% were women. Among the 41% who were living with HIV, 91% were men. Median follow-up duration was 48 months, and 22 patients (4%) developed anal cancer during follow-up. Higher methylation levels of ZNF582 and ASCL1 were significantly associated with high-grade squamous cell intraepithelial lesion (HSIL) cytology, p16-Ki67 dual-staining positivity, and high-risk HPV and HPV16 positivity on the same anal smear. Both methylation markers showed an AUC of 0.72 for discrimination between HSIL and non-HSIL cytology on the same anal smear. Higher methylation levels of both markers were significantly associated with evolution to anal cancer in univariate and multivariable analyses adjusted for age and HIV status (p < 0.001). When assessing the AUC over 1 and 3 years of follow-up, methylation markers demonstrated superior predictive value for anal cancer compared to other markers.

Interpretation

We have demonstrated the predictive value of host-cell DNA methylation marker levels in anal smears with regard to evolution to anal cancer in a very high-risk population.

Funding

This study was funded by the Agence Nationale de Recherche sur le Sida et les hépatites virales (ANRS) I Maladies Infectieuses Emergentes.

Research in context.

Evidence before this study

As anal cancer incidence has risen over the past decades, particularly in high-income countries, new cancer screening guidelines based on HPV detection and/or cytology have been proposed internationally, especially for at-risk populations. This will lead to an increase in high-resolution anoscopy (HRA) and high-grade anal lesion (HSIL) treatment. There is a critical need for triage markers to better identify the anal lesions most likely to progress to cancer in order to prevent overtreatment and HRA consultation overload. As methylation markers seemed very promising in cervical cancer screening, on December 6, 2019 we searched PubMed using the terms “anal cancer” and “methylation” and found that ZNF582 and ASCL1 methylation levels were correlated to anal lesion severity on biopsies on a transversal study. Since then, this correlation has been confirmed in different populations, men having sex with men living with HIV but also HIV-negative men and women, always in cross-sectional studies and on biopsy samples. Finally, the analytical performances of these markers have been validated on anal smears recently (June 2024).

Added value of this study

We quantified methylation level of ZNF582 and ASCL1 markers on anal smears collected at the inclusion of patients from the AIN3 cohort, a French national multicenter study including patients with a history of high-grade anal lesions (AIN3). During the median follow-up period of 48 months, 22 patients out of the 514 included in this ancillary study evolved to cancer. We demonstrated the association of methylation markers to cytologic abnormalities, HPV positivity and p16/Ki67 dual staining positivity on the same anal smear. More importantly, we showed with a longitudinal study, the predictive value of ZNF582 and ASCL1 markers for anal cancer on non-invasive samples.

Implications of all the available evidence

This study retrospectively demonstrates the predictive value of host-cell DNA markers for anal cancer on anal smears. It's the first step toward implementing methylation quantification as triage markers in a screening algorithm which would help to reduce the number of HRA exams needed by prioritising patients most at-risk and preventing overtreatment of anal lesions. As the guidelines published in 2024 recommend the use of anal smears for anal cancer screening in asymptomatic populations at-risk of anal cancer, this study will enable further studies on wider cohorts and patients less at-risk. Given the rise in incidence of anal cancers, this work is of increasing relevance.

Introduction

Although anal cancer is relatively rare in the general population,¹ increased incidence of anal cancer has been noticed worldwide in the past few years.^1,2 Some groups are more at risk of anal cancer, including (i) people living with HIV (PLWHIV); (ii) men having sex with men (MSM), in particular those living with HIV; (iii) women with a previous history of human papillomavirus (HPV)-induced high-grade lesions or cancer; and (iv) persons undergoing immunosuppression for solid-organ transplantation.³ Nearly all anal cancers are squamous cell carcinomas, and more than 90% are associated with high-risk HPV (hrHPV) infection, with HPV16 being most frequently implicated.⁴ Precancerous anal lesions precede anal cancers in a way similar to cervical cancer development. Anal lesions are classified as anal intraepithelial lesions (AINs), and are graded from 1 to 3.⁵ In contrast to cervical cancer screening, no worldwide consensus exists regarding anal cancer screening. The ANCHOR study published in 2022 has encouraged the performance of high-resolution anoscopy (HRA) to detect high-grade anal intraepithelial neoplasia (HGAIN) (i.e., AIN 2–3) in high-risk populations, especially PLWHIV, and to treat all HGAIN lesions to prevent their evolution to anal cancer.⁶ New anal cancer screening guidelines have been published in France and by the International Anal Neoplasia Society (IANS), which are based on HPV detection and/or cytology, but with no consensus.^7,8 Their implementation in real-world settings is difficult due to scarce access to HRA, even in high-income countries, as this technique requires expensive equipment and expertise acquired by medical staff over a long period of time.⁹ Moreover, even knowing that treating all HGAIN detected by HRA is decreasing anal cancer incidence,⁶ the benefit-to-harm of treating all these lesions is not yet known, as only a subset of these lesions will progress to anal cancer. These two points highlight the need for molecular markers to stratify anal cancer risk, enabling both a reduction in the number of patients recommended for HRA and a decrease in the overtreatment of detected HGAIN lesions.¹⁰

In cervical cancer screening, methylation markers demonstrate strong performance¹¹ and important predictive value,¹² while not yet implemented in screening guidelines due to lack of prospective studies comparing all the existing panels of methylation markers and also lack of cost-effectiveness studies. For instance, the association of FAM19A4 and mir124-2 methylation markers detected 100% of cervical cancers at the time of screening for 1040 women, and the 14-year cumulative cervical cancer incidence was decreased by half with a negative methylation test compared to a normal cytological test.¹¹ However, not all methylation markers developed for cervical cancer screening can be transposed to anal cancer screening, because a sample specificity in gene expression has been noted.^13,14 For instance, in the SPANC study conducted among men with anal high-grade lesions, two host gene methylations markers developed for cervical cancer, CADM1, and MAL, did not show greater predictive value for anal cancer than high-risk HPV typing. Notably, MAL was only limitedly predictive in men living with HIV and not predictive at all for HIV-negative men.¹³ In the same way, Lahiri et al. examined the methylation of FAM19A4 and microRNA124-2 promoters in cervical and anal smears from over 70 HIV-positive women. Only hypermethylation of these markers in cervical smears was associated with anal high-grade lesions, whereas nearly all anal smears showed hypermethylation of these markers, which had no predictive value.¹⁴ This suggests that these markers are mucosa specific. Recently in 2021, Van der Zee and colleagues identified a panel of host-cell DNA methylation markers with increased levels in high-grade and anal cancer biopsies¹⁵ and validated those markers for the detection of AIN3 and anal cancer in large cross-sectional studies including both patients living with HIV and not.^16,17 However, the prognostic value of these markers must be confirmed in larger longitudinal studies including all types of patients at high-risk. Furthermore, anal biopsies are relatively invasive samples that must be acquired under HRA to be as accurate as possible. A recent study in 2024 assessed the feasibility of analysis of host-cell DNA methylation markers on anal swabs, including ZNF582 and ASCL1, compared to the concurrent diagnosis on biopsy.¹⁸ Analogous to methods for cervical cancer screening, demonstrating the predictive value of methylation markers assessed by analysis of anal swabs would be valuable inputs as triage markers in anal cancer screening, as such samples are straightforward to obtain, minimally invasive, and could be generalised to large screening strategies in different settings. They could also help in the triage of patients needing to be referred to HRA when the infrastructure and expertise are available.

The current study aimed to evaluate two host-cell DNA methylation markers (ASCL1 and ZNF582), which had previously been validated using anal biopsies,15, 16, 17 by examining their relevance on anal swabs in a French multicenter, prospective, real-world cohort that included patients with a history of AIN3 lesions. In this study, we compared methylation markers quantification to cytology, hrHPV, and p16/Ki-67 dual-staining results on the same anal smear and evaluated the prognostic value of these methylation markers regarding clinical evolution.

Methods

Patients and clinical specimens

This ancillary study represents part of the AIN3 cohort, a French national multicentric cohort whose primary objective is to estimate the incidence of anal squamous cell carcinoma in patients with AIN3 lesions. Patients have been prospectively included in the cohort since August 2013.

The inclusion criterion was a previous or current AIN3 anal lesion diagnosed by histology, regardless of the mode of diagnosis (screening of at-risk population, post-operative discovery, etc.) and irrespective of sex (self-reported by participants), sexual orientation, or HIV serological status. Participants had an inclusion visit and were then followed annually for at least 3 years. In line with French guidelines, a proctological examination involving standard anoscopy was conducted at each visit, but no systematic HRA was performed. Therapeutic management was left to the discretion of the investigators. An anal smear was performed at inclusion and at each follow-up visit. The inclusion sample was systematically stored and centralised in an independent centre to perform cytological analysis, hrHPV detection, and p16/Ki-67 dual staining. All inclusion anal smear samples between August 1, 2013 and June 17, 2020, were sent to the Bichat-Claude Bernard Hospital Virology Department for ancillary study purposes. Patients with available inclusion anal smears, with enough volume and sufficient host-cell DNA left were included in this ancillary study (Fig. 1). AIN3 cohort database was extracted in June 2023.

Fig. 1.

Anal smears selection flowchart. All anal smears were collected at participant inclusion, independent of current anal lesions. ACTB Ct = β-actin PCR cycle threshold.

Sample collection at inclusion and precancerous marker analysis

Anal specimens were collected with Anex® brush (Rovers, The Netherlands) discharged by swirling in 20 mL of a ThinPrep PreservCyt® cytology collection device (Hologic, Villepinte, France). Detection of hrHPV was performed on an Abbott Alinity system (Abbott Molecular) following manufacturer instructions. Results were reported as positive for types 16 and/or 18 and/or 12 hrHPV types (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68). Cytology results were classified according to the Bethesda System terminology.¹⁹ Smears were prepared for CINtec PLUS on a ThinPrep processor (Hologic) following manufacturer instructions. Slides were stained using the CINtec PLUS assay kit and processed on a BenchMark ULTRA system (Roche Diagnostics) following manufacturer instructions. Smears were determined to be positive if at least one anal epithelial cell showed both brownish cytoplasmic immunostaining for p16 and red nuclear immunostaining for Ki-67, regardless of cellular morphology.

DNA methylation analysis

After centrifugation of 2 mL of anal samples, DNA was isolated using an EZ1 Advanced XL instrument (Qiagen, Les Ulis, France) and eluted in 60 μL of elution buffer. DNA was bisulfite-converted using the EZ DNA Methylation Kit (Zymo Research, Orange, CA, USA) following manufacturer instructions. Samples were analysed for two methylation markers, ASCL1 and ZNF582 (Self-screen BV, Amsterdam, The Netherlands) using multiplex quantitative, methylation-specific PCR (qMSP) assays targeting the two genes as well as β-actin (ACTB) as a reference gene.¹⁶ The multiplexed PCR was performed on 2.5 μL of bisulfite-treated DNA using a RotorGene Real-Time PCR System (Qiagen) with a double-stranded-DNA-based calibrator that contained the amplicon sequences of the targets and ACTB as an internal quality control.²⁰ Cycle threshold (Ct) values were measured at fixed thresholds for fluorescence. A Ct less than 31 for ACTB indicated sufficient DNA quantity and quality.¹⁶ ΔΔCt ratios were computed using the comparative Ct method by comparing the target Ct values with the Ct values of ACTB and the internal quality control calibrator (2^−ΔΔCt × 100).^21,22 All samples with a valid ACTB Ct (Ct < 31) and no signal for either ZNF582 or ASCL1 were considered negative for methylation and were assigned a Ct of 40.

Statistics

Methylation data were expressed in log₂(ΔΔCt ratio).15, 16, 17, 18^,21 Cox model proportionality and log-linearity assumptions were verified. Methylation levels were compared across anal swabs according to cytological status, hrHPV status, and p16/Ki67 detection by Wilcoxon rank-sum comparison tests. Anal cancer-free survival was defined as the time passed from patient's inclusion date, at which the anal swab was carried out, to anal cancer diagnosis or last available date of follow-up, whichever occurred first. Anal cancer-free survival rate was estimated by the Kaplan–Meier method. To test the prognostic role of methylation biomarkers, univariate and multivariable Cox models were generated. Diagnostic performance of methylation tools measured on baseline anal swabs was then assessed; the question to be addressed in this case was how well those markers measured at baseline could distinguish between patients who had developed cancer and patients who had not in a follow-up interval. Cumulative/dynamic time-dependent AUC (C/D AUC) (nearest neighbour estimator) values were calculated: given a time-point of interest t, C/D AUC $\left(t\right)=P(X_i>X_j/T_i\le t,T_j>t)$, where X is a marker value (methylation levels at inclusion in this case) and T is the time of disease onset (anal cancer diagnosis in this case) for individuals i and j. The 95% confidence intervals at 12 and 36 months were estimated by non-parametric bootstrap basic method with replacement (2000 replications).²³ Different thresholds were defined using methods based on time-dependent ROC curves. Given a timepoint of interest t and a threshold c, sensitivity and specificity can be defined as $Se(c,t)=P(Xi>c/Ti<t)$ and $Sp(c,t)=P(Xi\le c/Ti>t)$, respectively. This methodology estimates optimal thresholds by constructing the ROC curve at time t.²⁴ We determined one threshold optimising sensitivity and specificity (equivalent to the Youden index method).

We then generated univariate and multivariable Cox models to evaluate the hazard ratio (HR) of methylation status above thresholds. Univariate Cox models were also fitted with different marker candidates: hrHPV detection, HPV16 detection, HSIL result on cytology, and methylation. IPCW (Inverse probability of censoring weighting) estimates of the C-index were then calculated to evaluate discrimination by these models. To avoid bias, the C-index was not used to evaluate prediction of a t-year risk of cancer but instead to evaluate the correlation between the continuous event time and a prediction of (the order of) the event times. Therefore, the C-index was estimated at the maximal end of follow-up time (here τ = 98.5 months): $C_{\tau }=P(M_i<M_j/T_i>T_j,T_j<\tau )$, where i and j represent two random subjects, T is the time of disease onset (anal cancer diagnosis in this case), and M is the model prediction. The C/D AUC (IPCW estimator) and corresponding 95% CI at 12 and 36 months were also calculated.

Statistical analyses were performed using R statistical software (version 4.2.0; R Foundation for Statistical Computing, Vienna, Austria) using the ggplot2 v 3.4.0, survivalROC v 1.0.3.1, ThresholdROCsurvival v 1.0.3, boot v1.3-28, and pec v 2022.05.04 packages. The reported p-values are two-sided, and a p-value ≤0.05 was considered to indicate statistical significance.

Ethics

This study was approved by an institutional review board (Comité de Protection des Personnes du Sud-Ouest et Outre-Mer 4 n° IRB IORG0009855), and all patients provided written informed consent for the use of medical data and specimens. AIN3 cohort was registered under the number NCT01877135.

Role of funders

The academic funder, the Agence nationale de recherches sur le sida et les hépatites virales I Maladies Infectieuses Emergentes (ANRS⎥ MIE) French national agency, had no part in study design, data collection, data analyses, interpretation, or writing of report.

Results

Patient characteristics and clinical evolution

Between August 1, 2013 and June 17, 2020, 678 patients were prospectively included, and 574 anal samples collected at inclusion were available for this ancillary study. Due to insufficient DNA quantity (ACTB Ct < 31), 60 samples (10%) were not suitable for methylation analysis (Fig. 1). Among the remaining 514 participants, 60% were men. Overall, the mean age of participants was 50.8 years (SD = 13.4), and 41% were living with HIV, with 91% of those being men. The mean age at first AIN3 diagnosis was 48.8 years (SD = 13.5), and lesions had been diagnosed with standard anoscopy in 56% of cases. The patient characteristics are displayed in Table 1. In this ancillary study, the median duration of follow-up was 47.9 months (IQR = 44.4–51.9). During this period, 22 patients evolved to anal cancer including 16 men, 10 of them living with HIV, and 6 women, none of whom were HIV-infected (Table 1, Figure S1). The median delay between anal smear performed at inclusion and anal cancer diagnosis was 25.8 months (IQR = 9.4–53.8).

Table 1.

Participant characteristics.

	No anal cancer (n = 492)	Anal cancer (n = 22)	All participants (N = 514)
Sex
Female	201 (41%)	6 (27%)	207 (40%)
Male	291 (59%)	16 (73%)	307 (60%)
Age at inclusion in years (mean, SD)	50 (13.4)	60 (11.5)	51 (13.4)
Ethnicity
White	439 (89%)	22 (100%)	461 (90%)
Other	53 (11%)	0 (0%)	53 (10%)
HIV serological status at inclusion
Negative	286 (58%)	12 (55%)	298 (58%)
Positive	203 (41%)	10 (45%)	213 (41%)
Unknown	3 (1%)	0	3 (1%)
Antiretroviral treatment
No	6 (3%)	0 (0%)	6 (3%)
Yes	193 (95%)	10 (100%)	203 (95%)
Unknown	4 (2%)	0 (0%)	4 (2%)
Tobacco consumption
Current	166 (34%)	9 (41%)	175 (34%)
Past	76 (15%)	4 (18%)	80 (16%)
Never	211 (43%)	9 (41%)	220 (43%)
Unknown	39 (8%)	0 (0%)	39 (7%)
Anal sex
Often	118 (24%)	5 (23%)	123 (24%)
Occasional	126 (26%)	6 (27%)	132 (26%)
Never	147 (30%)	9 (41%)	156 (30%)
Unknown	101 (20%)	2 (9%)	103 (20%)

IQR = interquartile range.

Cyto-virological characterisation of anal smears at inclusion

Among the anal smears collected at inclusion from 514 participants in the AIN3 cohort, 6 anal smears were unsatisfactory, 27% (n = 139) were negative for intraepithelial lesions or malignancy (NILM), 39% (n = 198) exhibited atypical squamous cells of undetermined significance (ASC-US), 18% (n = 94) exhibited low-grade squamous intraepithelial lesion (LSIL), 1% (n = 4) exhibited “atypical squamous cells—cannot exclude high-grade squamous intraepithelial lesion” (ASC-H), and 14% (n = 73) exhibited high-grade squamous intraepithelial lesion (HSIL).

Among the 455 anal samples with available results from p16/Ki-67 dual-staining, 53% (n = 239) were positive. Results for hrHPV detection were available for 502 out of the 514 samples collected at inclusion, with 52% (n = 260) of them presenting with HPV16 and 81% (n = 405) with at least one hrHPV (Table S1).

Cross-sectional comparison of methylation levels with other variables on anal smears

Methylation quantification was successful for 514 out of 574 anal smears, resulting in suitable host-cell DNA quantity for 90% of the samples. There was no significant difference in methylation markers levels depending on sex (Table S2).

A significantly higher level of methylation of each marker studied (ZNF582 and ASCL1) was observed in samples with an HSIL cytology result (n = 73) than those with a non-HSIL result (n = 431) (log₂[ΔΔCt ratio] = 1.8 vs −2.2, p < 0.0001 (Wilcoxon rank-sum test) for ZNF582, and 2.6 vs −0.2, p < 0.0001 (Wilcoxon rank-sum test) for ASCL1) (Fig. 2a). Samples with missing cytology (n = 6) or ASC-H diagnosis (n = 4) were excluded from this comparison.

Fig. 2.

Boxplots of DNA methylation levels of different markers (a) HSIL, (b) p16/Ki-67, (c) hrHPV and (d) HPV16, on the same anal swab. The upper and lower whiskers extend from the hinges (Q3 and Q1, respectively) to the largest and smallest data points within 1.5 times the interquartile range (IQR) from the respective hinges. HSIL = High Grade Intraepithelial Lesion, hrHPV = High risk Human Papillomavirus, ASC-H = Atypical Squamous Cells that cannot exclude high-grade squamous intraepithelial lesion, log₂[ΔΔCt ratio] = Methylation levels (log₂ transformed ΔΔCt ratio). ∗∗p-value (Wilcoxon rank-sum test) < 0.0001.

On anal smears analysed by p16/Ki-67 double-staining (n = 453), a significantly higher level of methylation of each studied marker was observed in samples positive for p16/Ki-67 (n = 239) than in negative smears (n = 214) (log₂[ΔΔCt ratio] = 0.1 vs −3.3, p < 0.0001 (Wilcoxon rank-sum test) for ZNF582, and 1.4 vs −1.7, p < 0.0001 (Wilcoxon rank-sum test) for ASCL1) (Fig. 2b).

Regarding hrHPV, detection and typing were available for 502 anal smears. Methylation levels quantified in samples positive for at least one hrHPV (n = 405) were higher than in samples negative for all hrHPV (n = 97) (log₂[ΔΔCt ratio] = −1.0 vs −3.8, p < 0.0001 (Wilcoxon rank-sum test) for ZNF582, and 0.6 vs −2.7, p < 0.0001 (Wilcoxon rank-sum test) for ASCL1) (Fig. 2c). Focussing on HPV16, the hrHPV mostly implicated in anal cancer, methylation levels quantified in HPV16-positive samples (n = 260) were higher than in samples negative for HPV16 (n = 242) (log₂[ΔΔCt ratio] = −0.1 vs −3.1, p < 0.0001 (Wilcoxon rank-sum test) for ZNF582, and 1.3 vs −1.0, p < 0.0001 (Wilcoxon rank-sum test) for ASCL1) (Fig. 2d).

After stratifying these analyses based on HIV serological status, all of the associations of methylation markers on anal smears with the other variables described above (HSIL cytology, p16/Ki-67 dual-staining, HPV16 and hrHPV detection) were statistically significant among HIV-negative participants, and all but one (hrHPV detection) were statistically significant among participants living with HIV (Table S3).

The ROC curves comparing methylation levels in the non-HSIL (NILM, ASC-US, LSIL) and HSIL (ASC-H excluded) anal smears exhibited an AUC of 0.72 (95% CI = 0.66–0.79) for ZNF582 and 0.72 (95% CI = 0.65–0.79) for ASCL1, p = 0.93 (DeLong's test) (Fig. 3).

Fig. 3.

Receiver operator characteristics (ROC), associated area under the curve (AUC), and 95% CI of methylation quantification for two host-cell genes, ASCL1 (blue) and ZNF582 (red), for HSIL versus non-HSIL cytological results on the same anal smear.

Prognostic value of methylation markers for progression to anal cancer

In univariate analysis, higher methylation levels were quantified on anal samples from patients who had evolved to cancer during the median study period of 47.9 months (HR = 1.45, 95% CI = 1.25–1.67, p < 0.0001 (Wald Test) for ZNF582 and HR = 1.47, 95% CI = 1.23–1.75, p < 0.0001 (Wald Test) for ASCL1). In multivariable analysis, when adjusted for age and HIV serological status (variable selected based on literature), methylation levels remained significantly associated with a higher risk of anal cancer (HR = 1.43, 95% CI = 1.21–1.69, p < 0.0001 (Wald Test) for ZNF582 and HR = 1.40, 95% CI = 1.15–1.71, p = 0.0009 (Wald Test) for ASCL1) (Table 2).

Table 2.

Univariate and multivariable analyses of methylation levels of ZNF582 and ASCL1 quantified from anal smears at inclusion with regard to evolution to anal cancer.

log2[ΔΔCt ratio]	Univariate analysis			Multivariable analysis (adjusted on age and HIV status)
	HR	95% CI	p-value	HR	95% CI	p-value
ZNF582	1.45	1.25–1.67	<0.0001	1.43	1.21–1.69	<0.0001
ASCL1	1.47	1.23–1.75	<0.0001	1.40	1.15–1.71	0.0009

p-values were obtained using Wald tests.

HR = hazard ratio, CI = confidence interval, log₂[ΔΔCt ratio] = Methylation levels (log₂ transformed ΔΔCt ratio).

Using the C/D AUC approach, we assessed the predictive performance of the two methylation markers for the detection of anal cancer. After a 1-year follow-up, 8 cancers were detected, and the AUC values were 0.82 (95% CI = 0.68–0.96) and 0.79 (95% CI = 0.63–0.96) for ZNF582 and ASCL1, respectively. From these results, there was a probability of 82% for a patient who had evolved to cancer in less than 1 year after sampling to exhibit a higher level of ZNF582 methylation than a patient who would evolve to cancer over a longer period. After a 3-year follow-up, the C/D AUC values were 0.82 (95% CI = 0.69–0.95) and 0.80 (95% CI = 0.67–0.92) for ZNF582 and ASCL1, respectively (Figure S2).

Comparison of predictive value of methylation markers levels and other variables for anal cancer progression

Discrimination C-index was calculated to assess the predictive value of the different markers analysed on anal smears with regard to anal cancer. Predictive value of p16/Ki67 dual staining was not evaluable as all smears from patients who evolved to anal cancer were positive for this marker. The C-index was 0.51 for an HSIL cytology, 0.59 for at least one hrHPV detection, and 0.72 for HPV16 detection on anal smears. All of those markers exhibited a lower C-index than both methylation markers taken separately, with C-indexes of 0.84 for ZNF582 and 0.80 for ASCL1.

The comparison of C-indexes were concordant with the results from the C/D AUC analyses, both showing lower predictive values for non-methylation markers than those of both methylation markers at 1 and 3 years of follow-up. Indeed, at 1 year of follow-up, the C/D AUC for HSIL anal cytology was 0.62 (95% CI = 0.45–0.79), and it was 0.60 (95% CI = 0.58–0.62) and 0.75 (95% CI = 0.72–0.77) for at least one hrHPV and HPV16 detection, respectively. Similarly, at 3 years of follow-up, the C/D AUC for HSIL anal cytology and the C/D AUC for the detection of at least one hrHPV and HPV16 were still lower than those of the methylation markers (Table 3).

Table 3.

Predictive value for anal cancer progression of different markers performed on anal smears collected at inclusion: host-cell DNA methylation markers levels (ZNF582 and ASCL1), high-grade intraepithelial lesions detected in cytology (HSIL), anal infection by at least on high-risk HPV (hrHPV) and infection by at least HPV16.

	C-index (95% CI)	C/D AUC at 1 year of follow-up (95% CI)	C/D AUC at 3 years of follow-up (95% CI)
ZNF582	0.84 (0.76–0.93)	0.82 (0.68–0.96)	0.82 (0.69–0.95)
ASCL1	0.80 (0.72–0.89)	0.79 (0.63–0.96)	0.80 (0.67–0.92)
HSIL cytology	0.51 (0.39–0.57)	0.62 (0.45–0.79)	0.63 (0.50–0.75)
At least one hrHPV	0.59 (0.55–0.62)	0.60 (0.58–0.62)	0.59 (0.57–0.61)
HPV16	0.72 (0.65–0.83)	0.75 (0.72–0.77)	0.75 (0.72–0.78)

C/D AUC = Cumulative/dynamic time-dependent Area Under the Curve.

Defining thresholds for anal cancer risk based on methylation markers in anal smears

Based on 3-year follow-up information from our dataset, we calculated methylation markers level thresholds of 2.24 and 2.33 for ZNF582 and ASCL1 log₂[ΔΔCt ratio], respectively, with respective corresponding sensitivities of 60% and 69% and specificities of 85% and 78%.

Applying those thresholds to the anal samples from participants in this sub-study, ZNF582 and ASCL1 methylation levels above 2.24 and 2.33 log₂[ΔΔCt ratio], respectively, were associated with an increased risk of anal cancer (HR = 12.4, 95% CI = 4.83–31.8, p < 0.0001 (Wald Test) for ZNF582, and HR = 12.5, 95% CI = 4.23–37.1, p < 0.0001 (Wald Test) for ASCL1) (Table 4). In multivariable analysis, adjusted for age and HIV serological status, methylation markers levels above the defined thresholds for ZNF582 and ASCL1 remained associated with an increased risk of anal cancer (HR = 9.1, 95% CI = 3.3–25.0, p < 0.0001 (Wald Test) for ZNF582 and HR = 9.2, 95% CI = 3.0–28.7, p = 0.00013 (Wald Test) for ASCL1) (Table 4).

Table 4.

Application of methylation quantification thresholds for anal cancer risk at 3 years.

log2[ΔΔCt ratio]	Univariate analysis			Multivariable analysis (adjusted on age and HIV status)
	HR	95% CI	p-value	HR	95% CI	p-value
ZNF582 > 2.24	12.4	4.8–31.8	<0.0001	9.1	3.3–25.0	<0.0001
ASCL1 > 2.33	12.5	4.2–37.1	<0.0001	9.2	3.0–28.7	0.00013

p-values were obtained using Wald tests.

HR = hazard ratio, CI = confidence interval, log₂[ΔΔCt ratio] = Methylation levels (log₂ transformed ΔΔCt ratio).

Discussion

In this ancillary study of the French AIN3 cohort, we first confirmed the feasibility of host-cell DNA methylation quantification on anal smears. In a cross-sectional study, we then compared those methylation markers to other available markers for anal cancer screening on anal smears, specifically cytology, hrHPV detection, HPV16 typing, and p16-Ki67 dual-staining. Most importantly, we showed a significant association between methylation levels of ASCL1 and ZNF582 with anal cancer progression.

The feasibility of host-cell DNA methylation markers quantification on anal smears has only been reported once before in a SPANC cohort study which transposed markers developed for cervical cancer screening (CADM1, MAL, and mir124-2) on anal smears.¹³ The markers evaluated in our study have been validated on anal biopsies of different histologic grades,15, 16, 17 but their feasibility on anal smears was only reported in a recent work in PLWHIV in 2024¹⁸ on anal smears collected during HRA with a concurrent anal biopsy in a reference dermatological centre in Amsterdam. These results required confirmation, 1) on samples collected during routine proctological exam without HRA and, 2) across different centres, because the implementation of these markers will require that anal sampling can be carried out by any practitioner with varying levels of expertise, as a hindrance to these tests is that anal smears often exhibit lower cellularity than cervical smears. We have confirmed that ZNF582 and ASCL1 methylation analysis can be achieved from anal smears, with a 90% rate of success for methylation quantification.

ZNF582 and ASCL1 methylation levels were found to be significantly associated with the severity of lesions assessed by cytology, with hrHPV positivity and HPV16 positivity, and with p16-Ki67 dual-staining on the same anal smear. Importantly, these associations were independent of HIV-status in our study, while in the SPANC cohort study, MAL methylation level was relevant only in the HIV-infected MSM population.¹³

In our study, the AUC for discriminating HSIL and non-HSIL cytologies was 0.72 for both host-cell DNA methylation markers. The only other study regarding differentiation of HSIL and LSIL using methylation markers on anal smears¹³ reported similar AUC values of 0.68 and 0.67 for the MAL and CADM1 genes, respectively.

The ZNF582 and ASCL1 methylation markers have been extensively studied by Van der Zee et al., and their association with the severity of anal lesions has been demonstrated, although only in cross-sectional studies of anal tissue biopsies.15, 16, 17 They established the discriminatory value of ZNF582 and ASCL1 markers for AIN3+ versus lower-than-AIN3 anal lesions, with respective AUC values of 0.89 and 0.87 in a cross-sectional study of anal biopsies from MSM living with HIV¹⁶ and respective AUC values of 0.84 and 0.83 for the same genes in HIV-negative men and women.¹⁷ Another study of viral methylation markers in anal biopsies demonstrated an AUC of 0.82 in identifying HSIL and anal cancer from normal anal tissue samples.²⁵ The lower AUC observed in our study may be explained by the fact that we assessed methylation markers on exfoliated cells from anal swabs, containing a high diversity of cellular material and not targeting anal lesions, contrary to biopsies realised under HRA.

To evaluate the predictive value of host-cell DNA markers for anal cancer, longitudinal studies are necessary. Van Der Zee et al. confirmed in 10 anal cancer cases that anal biopsies collected before cancer diagnosis (AIN2 or AIN3, maximum 28 months before anal cancer) all exhibited hypermethylation of the ZNF582 and ASCL1 markers.¹⁵ Our study assessed the predictive value of host-cell DNA methylation markers for anal cancer using real-world clinical prospective follow-up of patients and anal cancer outcomes. Higher levels of ZNF582 and ASCL1 methylation on anal smears were indeed associated with progression to cancer, both in univariate and multivariable analyses adjusted for age and HIV serological status. The predictive performance of the two markers was evaluated, with C/D AUC values of 0.82 and 0.79 for ZNF582 and ASCL1, respectively, for anal cancer occurrence at 1 year after sampling. These predictive values were stable at 3 years of follow-up, with a C/D AUC >0.80 for both methylation markers. Both the C-index and the C/D AUC for HSIL cytology, hrHPV infection and HPV16 anal positivity were lower than those for the two methylation markers, suggesting better performance of host-cell DNA methylation markers in evaluating the risk of progression to anal cancer. These results are highly encouraging regarding the use of methylation markers on anal smears for anal cancer screening algorithm.

To validate these two methylation markers on anal smears and evaluate their predictive value for progression to anal cancer, it was necessary to use continuous variable of methylation quantification. However, defining a cutoff for each of these markers would facilitate their implementation in the anal cancer screening algorithm. Based on our data, thresholds were defined for both markers to warrant satisfactory specificity (85% and 78% for ZNF582 and ASCL1, respectively) without excessively reducing sensitivity (60% and 69%, respectively). Indeed, hrHPV detection is already a highly sensitive marker for precancerous anal lesions, and markers with higher specificity are needed for patient triage. Participants with methylation levels above the defined thresholds in anal smears had a significantly higher risk of progression to anal cancer, with a hazard ratio above 9 for both markers, which was also confirmed when adjusted for age and HIV status in multivariable analysis.

Although this prognostic value for markers on anal swabs is highly promising, it is important to note that applying the thresholds in our study would have failed to detect 2 cancers out of 8 that occurred in the year after anal swab sampling, 2 out of 8 diagnosed between 1 and 3 years after sampling, and 1 out of 6 diagnosed after 3 years of follow-up. However, these thresholds were defined in a very specific population at higher risk of anal cancer than the standard “at-risk” population, as the French AIN3 cohort enrolled patients with a history of high-grade anal lesions. The defined thresholds were very stringent; otherwise, a large proportion of the swabs obtained at inclusion would have been defined as positive for methylation. Defining a threshold for an at-risk anal cancer population without a history of high-grade anal lesions is mandatory to define markers thresholds for triage.

Our study exhibits some limitations. First, this project was a sub-study of the AIN3 French cohort, which included patients with a history of AIN3 lesions; therefore, the thresholds for methylation levels established in our study cannot be generalised. Moreover, the participants in our cohort were mostly white (89%). In addition, no systematic anal biopsy for histological analysis was performed at inclusion, and only cytological results were available. Another limitation of our study is the low number of anal cancer cases, as indicated by the wide confidence intervals. Therefore, even in a high-risk population such as that studied here, we were only able to conclude on the probable clinical importance of ZNF582 and ASCL1 methylation biomarkers. We cannot exclude confounding bias in the multivariable model and build-in selection bias.²⁶ Our findings are accompanied with statistical uncertainty, especially with wide 95% CI in thresholds analysis.^27,28

On the other hand, our study has several strengths. First the heterogeneity of the patients included in the French AIN3 cohort, which included both men and women and was not limited to those living with HIV, is more representative of all the population at-risk of anal cancer than studies conducted only in people living with HIV for instance. A further strength is the prospective clinical follow-up, which enabled the evaluation of the real prognostic performance of methylation status.

Following the ANCHOR study's strengthening the need for treatment of all AIN3 lesions to prevent anal cancer and recommending HRA-based anal cancer screening to detect all high-grade lesions,⁶ French and IANS guidelines for screening in at-risk populations have been recently published in 2024.^7,8 Because HRA access is scarce, even in high-income countries, available triage markers for non-invasive anal samples are necessary for the accurate recommendation of only the most patients at high-risk to HRA. Testing of host-cell DNA methylation markers on anal biopsies has seemed to show strong performance in stratifying anal lesions; however, invasive biopsies are not suitable for screening. Our study has demonstrated high predictive value of ZNF582 and ASCL1 methylation markers on anal smears with regard to progression to anal cancer, suggesting that these markers could be positioned in anal cancer screening algorithms as a bottleneck, i.e., triage marker, after a highly sensitive marker such as hrHPV or HPV16 detection. From this perspective, thresholds for these markers should be determined in the high-risk population that has been outlined (PLWHIV, MSM, and women with previous HPV-induced genital cancer) and then generalised and applied in consensual algorithms for anal cancer screening.

Our study establishes an association between host-cell DNA methylation markers quantified on anal smears and progression to anal cancer in a real-world cohort of patients at high-risk. Further studies are needed to confirm the prognostic value of these markers, crucially in a population that is less ‘at-risk’ with longitudinal follow-up of methylation markers levels to define thresholds that can be generalised.

Contributors

All authors contributed substantially to the manuscript as follows:

Conceptualisation (VMF, AD, AC, DD, CC, LA), data curation and verification (VMF, AD, CR, AN), formal analysis (VMF, AD, GC, MB, CR, AN, CB, ATH, RS), funding acquisition (VMF, CC, LA), investigation (VMF, MD, EV, LuS, GS, DB, SR, LaS, LA), methodology (VMF, AD, CC, CR), project administration (AD, CR), resources (MD, EV, GC, AN, MB), software (AD, AN, CR), supervision (VMF, AD, CC, LA), validation (VMF, AD, MB, CR, CC, LA), visualisation (VMF, AD, MB, CR, CC, LA), writing – original draft (VMF, AD, CC, LA), and writing – review and editing (VMF, AD, MD, EV, MB, LS, GS, CR, ATH, DB, SR, LS, CB, AC, DD, RS, CC, LA).

All authors read and approved the final version of the manuscript.

The AIN3 study group is made up of French proctologists who included and followed up patients in the AIN3 cohort. They conducted clinical exams, performed the anal swabs and biopsies, carried out the surgeries and treatments, and filled in the clinical research forms. The GREP study group is made up of French gastroenterologists who discussed and validated the French research project in proctology.

Data sharing statement

Deidentified participant data and analytical code are available upon reasonable request to the corresponding author. Data reuse is permitted after agreement with the research promoter.

Declaration of interests

Valentine M. Ferré received support for attending meetings and travel from Copan, honoraria from AstraZeneca and Moderna and grant support from MSD Avenir outside of this work. Albertus T. Hesselink is employed by Self-screen B.V. Self-screen B.V. develops, manufactures and licences high-risk HPV and methylation marker assays and holds patents on these tests. Diane Descamps received payment or honoraria for lectures or presentations from Gilead Science and MSD and participated on Advisory Board for Gilead Science and ViiV Healthcare. Charlotte Charpentier received support for attending meetings and travel from Copan. Laurent Abramowitz received consulting fees from MSD vaccin, participated on Advisory Board for MSD vaccin and grant support for the AIN3 cohort by MSD vaccin. Renske D.M. Steenbergen is a minority shareholder of Self-screen B.V., a spin-off company of Amsterdam UMC, location VUmc. Renske D.M. Steenbergen also declares consultancy fees from AstraZeneca. All the other authors declare that they have no competing interests.

The AIN3 cohort received fundings from academic agencies (SNFCP, Bourse COMAD of SNFGE, SIDACTION, LA LIGUE, Bourse FARE of SNFGE), and industrial groups (MSD-France, SANOFI-PASTEUR-MSD-France) but not for this ancillary study. The GREP research group is funded by membership fees.