DNA methylation markers have universal prognostic value for anal cancer risk in HIV‐negative and HIV‐positive individuals

Anal cancer has increasing incidence and is preceded by high‐grade anal intraepithelial neoplasia (HGAIN; AIN2–3). Previously, we identified and validated several methylation markers for accurate detection of anal cancer and HGAIN with cancer risk in HIV‐positive (HIV+) men who have sex with men (MSM). This study aimed to evaluate these markers in HIV‐negative risk groups. A cross‐sectional series of 176 tissue samples of anal cancer, AIN3, AIN2, AIN1 and control biopsies obtained in HIV‐negative women and men was tested for six methylation markers (ASCL1, LHX8, SST, WDR17, ZIC1 and ZNF582). Accuracy for detection of AIN3 and cancer (AIN3+) was determined by univariable and multivariable mixed‐effect ordinal logistic regression. Methylation levels of all markers increased with increasing severity of disease (P < 0.0001) and were comparable to results in HIV+ MSM. All markers showed high accuracy for AIN3+ detection [area under the curve (AUC): 0.83–0.86]. The optimal marker panel (ASCL1 and ZIC1; AUC = 0.85 for AIN3+) detected 98% of cancers at 79% specificity. In conclusion, DNA methylation markers show a high diagnostic performance for AIN3+ detection in HIV+ and HIV‐negative risk groups, justifying broad application of methylation analysis for anal cancer prevention programmes.

Anal cancer has increasing incidence and is preceded by high-grade anal intraepithelial neoplasia (HGAIN; AIN2-3). Previously, we identified and validated several methylation markers for accurate detection of anal cancer and HGAIN with cancer risk in HIV-positive (HIV+) men who have sex with men (MSM). This study aimed to evaluate these markers in HIV-negative risk groups. A cross-sectional series of 176 tissue samples of anal cancer, AIN3, AIN2, AIN1 and control biopsies obtained in HIV-negative women and men was tested for six methylation markers (ASCL1, LHX8, SST, WDR17, ZIC1 and ZNF582). Accuracy for detection of AIN3 and cancer (AIN3+) was determined by univariable and multivariable mixed-effect ordinal logistic regression. Methylation levels of all markers increased with increasing severity of disease (P < 0.0001) and were comparable to results in HIV+ MSM. All markers showed high accuracy for AIN3+ detection [area under the curve (AUC): 0.83-0.86]. The optimal marker panel (ASCL1 and ZIC1; AUC = 0.85 for AIN3+) detected 98% of cancers at 79% specificity. In conclusion, DNA methylation markers show a high diagnostic performance for AIN3+ detection in HIV+ and HIV-negative risk groups, justifying broad application of methylation analysis for anal cancer prevention programmes.

Introduction
Anal cancer incidence is increasing worldwide. Although still relatively rare in the general population (age-adjusted incidence rates of 1-2 per 100 000 person-years), incidence rates are slightly higher (1.5-2-fold) in women compared with men, increase with increasing age and are disproportionally high in several high-risk groups [1,2]. HIV-positive (HIV+) men who have sex with men (MSM) have the highest incidence rates (85 per 100 000), followed by HIV+ men who have sex with women (32 per 100 000), and HIV+ women (22 per 100 000) [3]. Besides the role of HIV, iatrogenic systemic immunosuppression, for example, for solid organ transplantation or autoimmune diseases (including inflammatory bowel diseases (IBDs), in particular Crohn's disease, or haematological malignancies) also increases the risk for anal cancer (up to 12 per 100 000). HIV-negative MSM (19 per 100 000) and women with prior human papillomavirus (HPV)-induced (pre)cancer also have substantially higher incidence rates (cervical: 6-9 per 100 000 and vulvar: 42-48 per 100 000) [3].
In analogy with cervical cancer screening, screening and treatment of HGAIN to prevent anal cancer in high risk groups is under debate. High-resolution anoscopy (HRA)-guided biopsies, the gold standard for HGAIN diagnostics, is a burdensome, expensive and complicated procedure [7]. Alternative screening techniques, such as anal cytology and HPV testing, have suboptimal diagnostic performance [4,7,8].
Host cell DNA methylation, that is addition of methyl groups (hypermethylation) to cytosines in cytosine located 5 0 of a guanine (CpG) sites, is an epigenetic hallmark in HPV-induced carcinogenesis, which can lead to inactivation of tumour suppressor genes [9]. Recently, we established that host cell DNA methylation is associated with anal carcinogenesis in HIV+ men and identified several methylation markers for accurate detection [area under the curve (AUC = 0.90)] of HGAIN and anal cancer in HIV+ MSM [10]. Subsequently, we validated the most potent markers (ASCL1, LHX8, SST, WDR17, ZIC1 and ZNF582) in a large independent series, confirmed the accuracy (AUC = 0.89) and showed that high methylation levels are associated with progression towards cancer. We therefore consider methylation analysis using these markers to be a promising tool to detect HGAIN and anal cancer in HIV+ MSM and for cancer risk stratification, identifying HGAIN in need of treatment, thereby reducing overtreatment [11].
In present study, we aimed to investigate whether host cell DNA methylation also plays a role in anal carcinogenesis in other risk groups, namely HIV-negative women and men referred for anal complaints or in screening for being at risk for anal cancer, and evaluated the diagnostic potential of the earlier identified six methylation markers in the detection of HGAIN and anal cancer in these risk groups.

Clinical specimens
This study involved a molecular analysis of a cross-sectional series of formalin-fixed paraffin-embedded (FFPE) anal tissue samples across the full spectrum of anal (pre)cancer lesions in HIV-negative women and men (Table 1). These patients were referred to the Amsterdam University Medical Centers (Amsterdam UMC), Location Academic Medical Center (AMC), Amsterdam, The Netherlands, for anal complaints or were in screening for having an increased risk (see risk factors below) of developing anal cancer. A total of 176 samples, including 40 anal SCC specimens (obtained between 1985 and 2011 in 40 patients: 22 women and 18 men) and 136 biopsies [normal control samples n = 30; AIN1 (including anogenital condylomata) n = 57; AIN2 n = 21; AIN3 n = 28; obtained between 2006 and 2020 in 81 patients: 35 women and 46 men], were retrieved from the Pathology Archive of the Amsterdam UMC. For 39 patients, multiple biopsies from different lesions were included. Normal control samples included biopsies taken from lesions suspected for AIN but histopathologically graded as normal or reactive anal epithelium. Median age of patients at time of biopsy per histological group and per gender is reported in Table 1. Of patients of whom biopsies were included, electronic medical records were thoroughly reviewed for HIV status and other risk factors for anal cancer (Table S1). HIV status (positive/negative) was tested or patient reported. For 18 anal SCC samples of older women, the HIV status was not provided in the medical record (no mentioning or test result) and was considered HIV-negative (presumed HIV-negative).
Other reported risk factors we looked for were being MSM; earlier or concurrent cervical (pre)cancer or vulvar (pre)cancer; solid organ transplantation recipient (SOTR); receiving other systemic immunosuppression therapies; and IBD (Crohn's disease or ulcerative colitis).
For comparison, we used previously published methylation data from a large cross-sectional series of anal (pre)cancer tissue samples (n = 345; 30 anal SCC, 74 AIN3, 98 AIN2, 37 AIN1 and 106 normal) of HIV+ MSM tested with the same methylation markers and assays. This concerned a population in screening for anal (pre)cancer at the Amsterdam UMC. Further details have been reported previously [11].

Ethics
We

Histopathological review
Whole tissue sections (WTS) of FFPE tissue blocks were sectioned according to the sandwich method, as described previously [11]. The first and last sections were haematoxylin-and eosin-stained and histopathologically reviewed by a board-certified pathologist (CJMvN), experienced in AIN histopathology, for confirmation of the lesion and to render a revision diagnosis for analysis. In-between WTS were collected in sterile PCR tubes for DNA isolation. The highest grade of AIN represented in the sandwich of the study specimen was considered the diagnosis for this study. Whenever this review was inconclusive (cases difficult to classify), p16 INK4A immunohistochemistry was used to determine the final study diagnosis [12].

DNA isolation and HPV testing
DNA was isolated using the QIAamp DNA FFPE tissue kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions and eluted with the easy-MAG 3 elution buffer (bioM erieux, Boxtel, The Netherlands), as described before [11]. HPV testing and genotyping was performed on isolated DNA by conducting a general primer GP5+/6+-mediated PCR with hybridisation of PCR products in an enzyme immunoassay (EIA). First, an oligonucleotide probe cocktail that detects 14 hrHPV types (HPV16, 18 , without further genotyping. Samples negative in both hrHPV EIA and lrHPV EIA were considered HPV-negative after verification of human DNA using b-globin amplification.

DNA methylation analysis using multiplex quantitative methylation-specific PCR
For DNA methylation analysis, isolated DNA was bisulfite-converted using the EZ DNA Methylation Kit (Zymo Research, Orange, CA, USA) according to the manufacturer's instructions, converting unmethylated cytosines into uracils while conserving methylated cytosines, as described before [11]. EpiTect MethyLight Master Mix (Qiagen) was used, together with fluorescent dye-labelled probes, 50 ng input of bisulfite-converted DNA and 100-300 nM of each primer, specifically amplifying methylated bisulfite-converted DNA. H 2 O was used as negative control. Bisulfite-converted sample DNA was analysed for six methylation markers (ASCL1, LHX8, SST, WDR17, ZIC1 and ZNF582) using two multiplex quantitative methylation-specific PCR (qMSP) assays, each targeting three genes and the reference gene, b-actin (ACTB). One multiplex qMSP targeted ASCL1, LHX8 and ZNF582, while the other assay targeted SST, WDR17 and ZIC1 [11]. A cycle threshold of < 32 for ACTB indicated sufficient DNA, DNA quality and adequate bisulfite conversion [11]. The multiplexes were performed on the ViiA 7 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) using a doublestranded DNA-based calibrator as internal quality control, which contains the amplicon sequences of the targets and ACTB [13]. Cycle threshold values were measured at fixed thresholds for fluorescence. DDCq ratios were computed using the comparative quantification cycle (Cq) method by comparing the target Cq values with the Cq values of ACTB and of the internal quality control calibrator (2 ÀDDCq 9 100) [14,15].

Statistical analysis
Differences in methylation levels across the different histological categories (normal, AIN1-3, SCC) within and across risk groups (HIV-negative women, HIV-negative men, HIV+ MSM) and HPV status classes were visualised using boxplots and tested for statistical significance using the Kruskal-Wallis test, followed by pairwise Mann-Whitney U-tests, with the Bonferroni correction. For evaluation of the diagnostic performance of the markers, we performed univariable mixed-effect ordinal logistic regression for the individual markers, while multivariable mixed-effect ordinal logistic regression with stepwise backward selection based on the likelihood ratio criteria was used to obtain and evaluate an optimal marker panel [16][17][18]. We used three ordered classes of outcomes: [≤ AIN1] (normal control samples and AIN1), [AIN2] and [AIN3+] (AIN3 and anal SCC). To account for a correlation between multiple samples obtained from the same patient, we incorporated a random subject effect in the model. From the regression analyses, we obtained predicted probabilities (PP; values ranging from 0 to 1), representing the risk for each class outcome given the methylation marker level of a sample. In interpreting the PP, care should be taken that these probabilities are based on the distribution of the class labels in the study and not in the general population. For instance, a class that has fewer samples compared with the other might perform poor in predicting the true class.
Compared with the use of logistic regression analysis in our previous papers on HIV+ MSM, ordinal regression better accounts for the ordered nature of the data (normal to AIN1-3 to SCC). Moreover, the ordinal regression analysis enables the inclusion of the intermediate class of AIN2 samples in the prediction model [17,18]. For practical reasons and clinical applicability, we converted the ordinal outcome in a binary outcome by collapsing the PP of the three classes into cases [AIN3+] (AIN3 and anal SCC) or controls [≤ AIN1] (normal control samples and AIN1) and dividing the PP of the middle class [AIN2] into half to contribute to the PP of cases and controls.
Diagnostic performance of the models was visualised using receiver operating characteristic (ROC) curves and assessed through the AUC, as well as through the sensitivity and specificity at the Youden's Index (J) threshold (threshold that maximises the sum of sensitivity and specificity). For samples with PP above the J-threshold, methylation results were considered methylation-positive. To evaluate the performance of our models on samples outside the set, we performed leave-one-out cross-validation (LOOCV). All analyses were performed on log 2 -transformed DDCq ratios.
For comparison, we applied the optimal marker panel developed for AIN3+ detection in HIV+ MSM in previous study [11]. Therefore, the multivariable logistic regression model, combining markers ZNF582, ASCL1 and SST, was used to compute PPs and determine methylation-positive detection rates based on the noncross-validated (non-CV) J-threshold (≥ 0.434) [11]. Due to differences in populations and model development, a further comparison in AUC, sensitivity and specificity was not possible.

Methylation levels across histological categories of anal disease
Methylation levels of all six markers differed significantly (P < 0.0001) over all histological categories for samples of HIV-negative women and men combined. Between consecutive histological categories, a trend towards increasing methylation levels with increasing severity of disease was observed. All markers demonstrated a significant increase in methylation levels from AIN1 to AIN2 and marker ASCL1 also for AIN3 to SCC (Fig. 1). For the other consecutive grades, significance was not reached.

Methylation levels in relation to HPV status
In general, we observed no significant difference between methylation levels and HPV status (HPV-negative, lrHPV-positive, non-HPV16 hrHPV types, or HPV16) within histological categories of samples of HIV-negative women and men combined (Fig. S1). Only in AIN2 samples, methylation levels of ASCL1, SST and ZIC1 were significantly higher in HPV16positive samples compared with samples positive for a non-HPV16 hrHPV type (P < 0.05). However, numbers for several categories were low.

Methylation levels in HIV-negative women and men compared with HIV+ MSM
Next, we compared methylation levels with previously obtained methylation results in HIV+ MSM [11].  [29,30]. Type-specific positivity includes those contributed by multiple infections. One sample can therefore be counted more than once. Only hrHPV (including probable/possible hrHPV) types taken into account. c lrHPV types in hrHPV-neg samples taken into account. Figure 2 shows methylation levels of all six markers per histological category of samples of HIV-negative women, HIV-negative men and HIV+ MSM separately. We noted only a few significant differences (P < 0.05) for ASCL1, SST, WDR17 and ZNF582, but the direction of differences between groups was not consistent. Overall, all markers show highly similar methylation levels across the different patient groups, including a trend towards increasing methylation levels with increasing severity of anal disease. Hence, in general, methylation level patterns were comparable in samples of HIV-negative women, HIV-negative men and HIV+ MSM.

Determination and diagnostic performance of an optimal marker panel for HIV-negative women and men
Multivariable regression analysis was used to determine and evaluate an optimal marker panel for AIN3+ detection in HIV-negative women and men combined. This analysis with backward selection yielded a model consisting of markers ASCL1 and ZIC1 as optimal panel with a non-CV AUC of 0.86 [95% confidence interval (CI): 0.80-0.91; Table 3]. Using the J-threshold (≥ 0.483), this panel provided a sensitivity of 85% and specificity of 78% and classified 98% (39/40) of the cancers, 68% (19/ 28) of AIN3, 62% (13/21) of AIN2, 12% (7/57) of AIN1 and 13% (4/30) of normal control samples as methylation-positive (Table 3).
Age and gender did not significantly contribute to the model (data not shown), supporting the combined analysis of HIV-negative women and men. Marker WDR17 was the only marker that classified the SCC that was missed by the panel as methylation-positive. All hrHPV-negative SCC were classified as methylation-positive using the panel.

Discussion
The most important outcome of this study is the confirmation that akin HIV+ MSM, host cell DNA methylation also plays a role in anal carcinogenesis in HIV-negative women and men. We observed comparable methylation level patterns in samples of HIV-negative women and men as to our previous studies in HIV+ MSM including a significant increase in methylation levels with increasing severity of anal disease [11,10]. All six methylation markers showed a good cross-validated diagnostic performance for AIN3 and anal cancer detection (AUC = 0.83-0.86) in this series. Of these markers, SST achieved the best diagnostic performance (AUC = 0.86) for the detection of AIN3 and anal cancer with an optimal sensitivity and specificity of 90% and 73%, respectively, and classified 98% of the cancers as methylation-positive. By evaluating combinations of markers, we composed an optimal marker panel for HIV-negative women and men consisting of ASCL1 and ZIC1 with a cross-validated AUC of 0.85. Although the marker panel had a slightly lower AUC than SST alone, it detected a similar proportion of SCC at a higher specificity (79%).
For HIV+ MSM, we recently showed that high methylation levels are associated with progression towards cancer. Accordingly, methylation markers may be used to stratify HGAIN at risk of progression to cancer and therefore requiring treatment, while preventing overtreatment of HGAIN lesions with a low progression risk [11]. Interestingly, the AIN and cancer detection rates obtained by present marker panel developed for HIV-negative patients were comparable to the detection rates obtained by applying the optimal marker panel previously developed for HIV+ MSM [11], which uses the same methylation markers and qMSP assays. Therefore, we believe that current findings in samples of HIV-negative women and men open up a broader application of methylation markers in anal cancer prevention strategies for risk groups regardless of sex or HIV status. Also for cervical (pre)cancer, comparable methylation levels and diagnostic performance of methylation analysis have been reported for HIV+ and HIV-negative women [13,[22][23][24][25]. Further validation studies in different risk populations, having different anal cancer risks and prevalence rates of HGAIN are now warranted.
We observed a comparable HPV distribution in this series of HIV-negative women and men to what was reported in a recent meta-analysis by Lin et al. [26], including an increasing trend of HPV16 positivity towards anal cancer. This trend is known to be less prominent in HIV+ MSM, in whom HGAIN lesions and SCC are more often caused by hrHPV types other than HPV16, as we also observed in our previous studies on HIV+ MSM [11,10]. Compared to HIV+ MSM, we observed less multiple infections in our HIV-negative series, as was also reported by Lin et al. [26]. The The box of the boxplots bounds the IQR divided by the median, Tukey-style whiskers extend to a maximum of 1.5 IQR beyond the box, and • marks an outlier sample. Overall differences between histological categories upon Kruskal-Wallis omnibus test ( # P < 0.0001), followed by post hoc testing using the Mann-Whitney U-test and Bonferroni multiple testing correction: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns: nonsignificant.Normal, normal control samples; AIN: anal intraepithelial neoplasia (grades 1-3); SCC: anal squamous cell carcinoma. is slightly lower to what has been reported for HIVnegative patients (89-90%) in the meta-analysis, although within the range of the included studies [26]. Potential explanations for the varying HPV positivity rates are differences in populations with different anal cancer aetiology, geographical differences and/or use of different HPV detection methods.
To the best of our knowledge, we are the first to have evaluated host cell DNA methylation markers in anal tissue samples from multiple target risk populations of sufficient size. Previous studies on host cell DNA methylation investigated samples of small mixed populations of predominantly HIV+ men, not allowing comparison of methylation levels between risk groups [27,28].
Our study has several limitations. First, for this study we included samples from a HIV-negative population consisting of patients being referred for anal complaints and patients in screening for having an increased risk of developing anal cancer. Therefore, our findings cannot be extrapolated to a screening population, but are merely a proof of principle to show the potential of methylation markers to, upon discovery with HRA, identify HGAIN biopsies in need of treatment to prevent overtreatment in a HIV-negative referral population. Notwithstanding, present findings also indicate that methylation analysis on lessinvasive anal swabs to triage for HRA, especially in risk groups other than HIV+ MSM, who have a lower cancer risk and are not routinely screened with HRA may warrant further investigation. Second, we compared current series with our previous study on a HIV+ MSM screening population with a different study population composition, possibly leading to selection bias. HGAIN lesions found in screening programmes might be detected earlier compared with lesions found in a referral population and could therefore represent lesions earlier during carcinogenesis. Methylation levels in the present study were comparable to those in our previous study on samples of a HIV+ MSM screening population. Nevertheless, the different study populations may mask a potential difference in cancer risk between HIV+ and HIV-negative patients. This difference in populations, together with different multivariable models for marker panel Table 3. Ordinal regression analysis on diagnostic performance for AIN3+ detectionunivariable regression of the six individual markers (ASCL1, LHX8, SST, WDR17, ZIC1 and ZNF582) and multivariable regression for optimal marker panel (ZIC1 and ASCL1) and methylationpositive detection rate at J-threshold. LOOCV (and non-CV for the marker panel) AUCs are reported, together with sensitivity, specificity and detection rate are for the J-threshold. Outcome: AIN3+ (AIN3 and anal SCC) in anal tissue samples of (HIV-negative) women and men.

ASCL1
LHX8  structure within the same patients, we incorporated a random subject effect in our model to minimise sampling bias. Last, although this is the largest study evaluating host cell DNA methylation markers in HIVnegative women and men to date, numbers were still generally too low for several subgroups analyses, for example comparing HIV-negative women and men. Although we found no significant contribution of gender in the multivariable analysis and we could combine HIV-negative women and men, we acknowledge that this was still a heterogeneous group with different risk factors. Moreover, by comparing results with our previous series on HIV+ MSM, which was much larger, comparisons could have been influenced by an imbalance in numbers.

Conclusions
We showed that similar to anal carcinogenesis in HIV+ MSM, host cell DNA methylation also plays a role in anal carcinogenesis in HIV-negative women and men. The high resemblance in methylation levels and good diagnostic performance of the methylation markers in samples of both HIV-negative and HIV+ patients open up a broad application of methylation markers in anal cancer prevention strategies for all risk groups of anal cancer.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Boxplots of DNA methylation levels relative to a reference gene ACTB (log2-transformed DDCq ratios; y-axis) in relation to HPV status in the different histological categories of anal tissue samples of HIV-negative women and men combined (x-axis) for six markers. Table S1. Number of included tissue samples of HIVnegative men (A) en HIV-negative women (B), including age at biopsy, anatomical location, HIV status and risk factors per histological category.