Gene methylation of CADM1 and MAL identified as a biomarker of high grade anal intraepithelial neoplasia

Human papillomavirus (HPV) is detected in up to 96% of anal squamous cell cancers, where screening programs needed. However, the best methodology is still undetermined. Host DNA methylation markers CADM1, MAL and miR124 have been identified in cervical disease, but not anal disease. Anal swabs varying by disease grade were assessed for DNA methylation of CADM1, MAL and miR124-2. Each marker was compared across disease grades, stratified by HPV and HIV status. Receiver operating characteristic curves identified the predictive value of significant gene candidates. CADM1 methylation was significantly higher in high-grade squamous intraepithelial lesions (HSIL) compared with low-grade (LSIL) (p = 0.005) or normal (p < 0.001) samples with 67.2% correctly identified as HSIL. MAL methylation was significantly (p = 0.002) increased in HSIL compared with LSIL in HIV positive participants with 79.8% correctly indicated as HSIL. Gene miR124-2, showed no difference between disease grades. Biomarkers with established diagnostic value in cervical disease have limited utility in the prediction of anal disease, with CADM1 identified as a marker with screening potential in a gay and bisexual men (GBM) population and MAL in HIV positive GBM population. New markers specific to the anal mucosa are required to improve triage of high-risk individuals.

Methylation of the gene CADM1 differentiates between HSIL and LSIL/normal samples. CADM1 methylation was significantly different across all samples when stratified by disease (p < 0.001) ( Fig. 1, panel A1). A further group specific analysis (Wilcoxon), also revealed that CADM1 methylation from anal swab samples from men with HSIL was significantly different compared to those with both LSIL (p = 0.005) and normal samples (p < 0.001) (Fig. 1, panel A1). There was also a significant difference between HSIL and LSIL (p = 0.024) and HSIL and normal (p = 0.002) when the analysis was limited to HRHPV positive samples only (Fig. 1, panel A2).
The AUROC for the methylation of the gene CADM1 was 67.2% for HSIL over LSIL and 60.4% for HSIL over normal ( Fig. 2A). Also, a comparison of HRHPV positive samples indicated an AUROC of 67.3% for HSIL over LSIL and 61.8% for HSIL over normal (Fig. 2B). Furthermore, similar AUROC (between 60 and 62%) were determined for comparisons between HSIL and normal for HIV negative and positive participants including all samples ( Fig. 2A) and HRHPV positive samples only (Fig. 2B). Most significantly, in HIV positive men, CADM1 methylation distinguished HSIL from LSIL with an AUROC of 73.3% ( Fig. 2A) and in HIV positive men with HRHPV infections, the AUC was 78.2% (Fig. 2B).  (Fig. 3, panel C1). Within the HIV positive group, methylation in HSIL was significantly higher compared with LSIL in all samples (p = 0.002) (Fig. 3, panel C1) and HRHPV samples (p = 0.017) (Fig. 3, panel C2), however there was no difference when compared with normal samples (Fig. 3, panels C1 and C2); (as observed in the overall group). The AUROC for the methylation of the gene MAL was 67.6% for HSIL over LSIL (Fig. 4). Furthermore, HSIL compared with LSIL in HRHPV, and HIV positive anal swab samples indicated an AUROC of 78.9%. www.nature.com/scientificreports/ The methylation of gene miR124-2 is not predictive of anal HSIL in GBM. The methylation of the promoter region for miR124-2 compared with composite histologically and cytologically defined anal disease identified no significant differences by anal disease grade between any of the subgroups analysed, including HRHPV positive samples or HIV status (Fig. 5).
The value of different screening algorithms to predict histologically confirmed HSIL. The utilisation of percentage methylation for gene CADM1 and MAL in potential screening algorithms was assessed in combination with and without HRHPV positivity and/or cytological HSIL. Overall, the addition of cytology results had limited effect on the AUROC value of gene methylation (Fig. 6). Cytology alone had the highest AUROC within histologically-confirmed HSIL (68.8%) (Fig. 6A). Using multiple screening markers, CADM1 in combination with HRHPV positivity had an AUROC value of 62.3% for histologically-confirmed HSIL (Fig. 6A). Within the HIV positive population, the highest AUROC was cytology alone (70.1%) (Fig. 6B). Using a combination of different markers lowered the AUROC value with all markers only achieving 56.8% in histology confirmed HSIL (Fig. 6B). Cut-off values for each methylation marker were also assessed, however no combination tested improved the AUROC percentage above cytology alone ( Supplementary Fig. 1).

Discussion
In the current study, men with HSIL had an increase in the host DNA methylation of CpG sites within promoter regions for the gene CADM1 and, to a lesser extent, MAL. Among HIV positive individuals only, gene MAL had significantly less gene methylation in men whose highest grade of disease was LSIL compared with both normal histology and HSIL. This limits the utility of MAL as a biomarker of disease detection. The methylation of CpG sites for both CADM1 and MAL have previously been indicated as potential biomarkers in HPV-related cervical intraepithelial neoplasia (CIN) with reports of 78% specificity and 70% sensitivity in women with HRHPV infections for histological CIN3+ 30 . This is the first report of these markers being associated with anal HSIL in men and identifies their potential to be included in combination with other biomarkers for anal cancer screening programs.
The predictive value of CADM1 methylation for the determination of HSIL indicated that approximately 67% of anal samples would be correctly diagnosed as HSIL with this single marker, with no difference between HIV or HRHPV status.
Other studies of gene methylation markers from anal HSIL and cancer in HIV positive and negative men include p16, Ki-67, ASCL1, SST, ZIC1, ZNF582 and the HPV E4 gene where significant differences have been reported from tissue samples [31][32][33] . The methylation of the CpG promoter region of gene MAL was observed to be significantly increased only in HIV positive GBM with HSIL (p, ≤ 0.01) as compared with those with LSIL, but not when compared to normal samples or in HIV negative participants. Changes to the expression of the MAL gene have previously been identified in several different cancers with early reports identifying MAL as a tumour suppressor 34 . Subsequent reports have also identified decreases in MAL expression observed in cervical, colon, breast, stomach, bladder, salivary gland, head and neck and non-small cell lung cancers 35 . In the current www.nature.com/scientificreports/ report, the significant increase in MAL methylation in HSIL compared with LSIL in HIV positive individuals is perplexing, due to the similar methylation patterns between normal and HSIL samples. This unusual progression of changing methylation between normal, low and high grade disease in HIV positive GBM could be due to the MAL protein acting as a raft for other oncogenic proteins, as proposed by Lara-Lemus 35 . However, the limitation of this change to HIV-positive individuals could also be due to HIV-related changes in T-lymphocytes. Individuals infected with HIV experience a sustained hyper immune activation, triggered by changes in T cell dynamics and function 36 . It is possible that during acute infections of HPV, HIV-positive individuals have a varied response to acute HPV infection compared with HIV negative individuals. Although, the significant findings of ZNF582 with a 93% specificity for ≥ AIN3 does show promise for the diagnostic potential of methylation markers in the anal mucosa 32 . In addition to these strictly male gender assigned at birth and currently identifying as male studies, further analysis of the gene EPB41L3 and HPV16 in anal samples from HIV positive and negative men and women identified a combined sensitivity of 90.6%, specificity of 50.7% and AUROC of 0.82 in identifying HSIL and cancer from normal tissue samples 37 . It is important to note that these anal specific studies all utilised tissue samples whereas the current study focuses on swab samples. This difference is an important point of difference. If a methylation marker is to be utilised as a screening tool it needs to be sensitive enough to detect disease without the collection of tissue samples, which require highly specialised equipment and personnel. By only focusing on tissue samples the application of the findings are limited until the relevant sample has been assessed. This is where the current study is unique compared to all other anal methylation studies.
However, it should also be noted that determining small changes for low levels of disease is a difficult task which is compounded by utilising swab samples, which contain a high diversity of cellular material, to measure lesion specific changes. Our result indicating higher methylation levels in normal versus LSIL samples could be a consequence of the sample utilised and not differences between disease presentations.
Methylation-based silencing of precursor micro RNAs of microRNA 124 have previously been indicated in various human neoplasms, including the cervix 38,39 , where a significant inverse correlation was observed in the methylation of the promoter region and the presence of miR124-2 RNA. The same study also demonstrated www.nature.com/scientificreports/ in-vitro, that silencing of miR124-2 is functionally involved in cervical cancer development 38 . However, in the current study, methylation of the promoter region of miR124-2 was not found to be different between men with HSIL, LSIL or normal samples. The absence of this important biomarker could be due to the nature of the samples collected, as anal swab samples contain a mixture of normal and diseased cells. Other explanations could be that this is a phenotype specific to the anal canal. Further analyses utilising specific tissue samples from laser capture microdissection techniques and investigations into other precursor micro RNAs for miR124 are required to define any roles in the development of HSIL. Combination algorithms utilising CADM1, HRHPV detection and cytology diagnosis marginally improved the predictive value (from 61.2% alone to 61.6% combined) with cytology alone outperforming all other combinations at 68.8% AUROC in differentiating HSIL from LSIL and normal samples. Similar findings were also observed utilising MAL, in HIV positive participants. These findings indicate that complex screening algorithms may be required to achieve similar results to those seen in the cervix. For example, if all samples are screened by cytology to filter out all normal samples then CADM1 and HRHPV testing can be added for disease grade diagnosis (between LSIL, HSIL and ASCC) then the predictive value increases to 78.2% (similar to findings in the cervix at 78% specificity and 70% sensitivity in women with HRHPV infections) 30 .
Limitations to this study were the use of single promoter regions for methylation measurements and the absence of mRNA measurements to assess the level of gene silencing at a transcriptional level. Future studies should also assess the mRNA effects due to methylation of all possible promoter regions. The methylation patterns of CADM1 and MAL will be assessed for changes in longitudinal studies of HPV persistence utilising the SPANC cohort of samples. The detection cut off for the internal control was also increased from the published > 32 to > 40 cycle threshold which is a considerably low threshold. However, through modifications to the PCR reactions and confirmation through serial dilutions of the SiHa cell line (inter and intra assays) we confirmed that our modifications increased the assays sensitivity allowing for the increased cycle threshold cut-off.
In conclusion, this study demonstrates methylation of the promoter region for gene CADM1, when used in a panel of biomarkers, is a promising candidate for use as part of an anal screening algorithm for HSIL in HRHPV positive GBM. Furthermore, in HIV positive GBM, methylation of the promoter region for gene MAL may be important in HIV related immune modulation. It is clear from these findings that biomarkers utilised in the cervix are of limited use within other mucosa tissue sites highlighting the need to further explore new genes of interest specific for HPV related disease of the anal mucosa.   www.nature.com/scientificreports/ positive for any HPV type that was detected on either Linear Array or Anyplex™ to achieve a high sensitivity for all HPV genotypes. Briefly, each PCR run had an old bisulphite modified SiHa cell line that had been previously validated for methylation analysis and a new (identical DNA concentration) SiHa cell line that was bisulphite modified at the same time and with the same bisulphite reagents that the samples to be analysed by qPCR that day (in general 9 samples). The objective of this procedure was to analyse the variation in the bisulphite modification procedure between assays and to validate the methylation analysis between different runs (reproducibility of % of methylation of the positive control for each gene). Our results showed very high reproducibility in the % of methylation in the SiHa cell lines for each gene between runs (data not shown). www.nature.com/scientificreports/ Data analysis. A sample was deemed satisfactory if the crossing threshold value for ACTB was < 40. Any sample that was negative for ACTB was not considered for analysis regardless of methylation marker status. The percentage methylation in each individual sample (% meCpG) was calculated as described based on the house keeping gene ACTB.

Isolation of nucleic acids for methylation PCR.
Percentage methylation for each marker (CADM1, MAL and miR124-2), for each composite disease grade (HSIL, LSIL and normal) was compared using two non-parametric comparative analysis, Wilcoxon test for between groups and Kruskal-Wallis for overall analysis: this was visualised using box and whisker plots. These analyses were further stratified by HRHPV positivity and HIV status.
The diagnostic performance of each methylation marker was evaluated by receiver operating characteristics (ROCs) analyses. Area under the curve (AUROC) was the main measure used to assess the ability of the genes to distinguish/differentiate HSIL from LSIL/normal by using the best cut-off with the maximum sum of sensitivity and specificity. Different screening algorithms were also assessed utilising gene methylation for markers that were significantly associated with HSIL detection, within a sub cohort of HRHPV positive samples and anal cytological HSIL, to predict histologically confirmed HSIL. All %meCpG calculations were performed in Microsoft Excel v16.39 (Microsoft, Redmond, Washington, USA). The results were then analysed and figures produced using the statistical platform R studio (v4.0.1) 49

Data availability
All data and materials are available within this document and the supplementary files. www.nature.com/scientificreports/