Epigenetic variability in cells of normal cytology is associated with the risk of future morphological transformation
© Teschendorff et al.; licensee BioMed Central Ltd. 2012
Received: 17 January 2012
Accepted: 27 March 2012
Published: 27 March 2012
Recently, it has been proposed that epigenetic variation may contribute to the risk of complex genetic diseases like cancer. We aimed to demonstrate that epigenetic changes in normal cells, collected years in advance of the first signs of morphological transformation, can predict the risk of such transformation.
We analyzed DNA methylation (DNAm) profiles of over 27,000 CpGs in cytologically normal cells of the uterine cervix from 152 women in a prospective nested case-control study. We used statistics based on differential variability to identify CpGs associated with the risk of transformation and a novel statistical algorithm called EVORA (Epigenetic Variable Outliers for Risk prediction Analysis) to make predictions.
We observed many CpGs that were differentially variable between women who developed a non-invasive cervical neoplasia within 3 years of sample collection and those that remained disease-free. These CpGs exhibited heterogeneous outlier methylation profiles and overlapped strongly with CpGs undergoing age-associated DNA methylation changes in normal tissue. Using EVORA, we demonstrate that the risk of cervical neoplasia can be predicted in blind test sets (AUC = 0.66 (0.58 to 0.75)), and that assessment of DNAm variability allows more reliable identification of risk-associated CpGs than statistics based on differences in mean methylation levels. In independent data, EVORA showed high sensitivity and specificity to detect pre-invasive neoplasia and cervical cancer (AUC = 0.93 (0.86 to 1) and AUC = 1, respectively).
We demonstrate that the risk of neoplastic transformation can be predicted from DNA methylation profiles in the morphologically normal cell of origin of an epithelial cancer. Having profiled only 0.1% of CpGs in the human genome, studies of wider coverage are likely to yield improved predictive and diagnostic models with the accuracy needed for clinical application.
The ARTISTIC trial is registered with the International Standard Randomised Controlled Trial Number ISRCTN25417821.
It has been proposed that epigenetic variation may contribute to the risk of complex genetic diseases like cancer and that differential exposure to environmental risk factors may underlie much of this variation [1, 2]. Consistent with this view, a recent study has shown that regions that are differentially methylated between normal and cancer tissue appear to be highly variable in cancer itself, and that identification of cancer-relevant markers may therefore benefit from statistics that measure differential variability . Based on these insights, we here aimed to demonstrate that analysis of epigenetic variability in prospectively collected normal cells can predict the risk of future morphological transformation.
In order to demonstrate this in humans, two requirements are mandatory: the cells that are used for epigenetic analyses need to be (i) the cells of origin for the studied cancer, and (ii) they need to be collected years in advance of the onset of cytological and morphological signs of cancer. Currently, the only human organ that meets these two requirements is the uterine cervix. Thus, we used Illumina Infinium technology  to measure DNA methylation (DNAm) at 27,578 CpG sites in cytologically entire normal cells (liquid-based cytology (LBC) samples) from the uterine cervix of 152 women (aged 19 to 55 years) in a nested prospective case-control study within ARTISTIC (A Randomised Trial of HPV Testing in Primary Cervical Screening [5, 6]). Prospective cases were women who developed a cervical intraepithelial neoplasia of grade 2 or higher (CIN2+) within 3 years of sample draw, while controls were women who remained disease-free. To further support our data we used completely independent LBC samples with abnormal cytology and associated controls, as well as cervical cancer tissue and normal cervix specimens.
The ARTISTIC trial
The LBC samples we analyzed were collected from women as part of the ARTISTIC trial [5, 6]. All women underwent two screening rounds with an interval of 3 years. Within the ARTISTIC trial, women, aged 19 to 64 years who were undergoing routine screening as part of the English National Health Service Cervical Screening Programme in Greater Manchester were randomly assigned in a ratio of 3:1 to either combined LBC and human papilloma virus (HPV) testing where the results were revealed and acted on, or to combined LBC and HPV testing where the HPV result was concealed from the patient and investigator. There were a total of 24,510 eligible women at entry. In the first round of screening 453 women had CIN2+. In the second round of screening 75 women (who were screen-negative in the first round and who had a sample stored from the first round) had developed CIN2+ (44 were HPV-positive and 31 were HPV-negative in the first round). Seventy-seven women who had not developed any cytological changes were matched (age and HPV status in round 1) to the cases. The cytologically normal samples from round 1 from these 152 women were used for DNAm analysis. Further details, cytology and HPV scoring are described in Additional file 1. This trial is registered with the International Standard Randomised Controlled Trial Number ISRCTN25417821.
DNA methylation nested case control study
A total of 152 samples in a prospective nested case control study within ARTISTIC were selected for DNAm analysis. Cases were 75 women who had normal cytology in screening round 1 but demonstrated CIN2+ after 3 years in round 2. Controls were 77 women who had normal cytology at entry and in the second screening round. Cases and controls were matched for age (Wilcox rank sum test P = 0.95) and HPV status: 92 were HPV-positive (44 cases and 48 controls) and 60 were HPV-negative (31 cases and 29 controls) at entry (Fisher test P = 0.74). Informed consent was obtained from the main study population and this study has been approved by the ethical committee (National Research Ethics Service Reference Number 10/H1107/15).
Other DNA methylation data sets
In addition to the nested case control prospective study within ARTISTIC, we used two additional DNAm data sets that were also generated using the same Illumina Infinium 27 k platform. Set 1 comprised a total of 30 LBC samples (19 HPV-negatives and 11 HPV-positives) with normal cytology and 18 LBC samples (all HPV-positive) with CIN2+, as described in . Set 2 comprised a total of 63 cervical tissue samples: 48 cervical cancers, 15 normals. The normal cervical tissue samples were from women (mean age 55.4 years) who underwent a hysterectomy for fibroids of the uterine corpus, that is, these women did not have fibroids in the uterine cervix. The 48 cervical cancer specimens were from women (matched for age with mean age 56.8 years) who were treated at the Innsbruck Medical University. Of these 48 cancers, 26 were of stage 1, 10 of stage 2, 4 of stage 3 and 7 of stage 4 (1 sample had missing stage information). In terms of grade, 7 were of grade 1, 28 of grade 2 and 11 of grade 3 (2 samples had no grading). All specimens were obtained with informed consent and approval from the ethics committee UN4044-290/4.9.
DNA extraction and methylation assay
The DNA extraction protocol is described in Additional file 1. Genome-wide methylation analysis using the Illumina Infinium Methylation27K beadchip (Illumina Inc., USA, WG-311-1201) was performed. This chip interrogates the methylation status of over 27,000 CpG sites throughout the human genome, covering the promoters of over 14,000 genes . Further details are in Additional file 1.
All data in this manuscript have been deposited in the Gene Expression Omnibus repository  under accession number [GSE30760].
mRNA expression data of normal cervical and cervical cancer tissue
We used the publicly available normalized expression data (all Affymetrix) from the Gene Expression Omnibus with accession numbers [GSE9750] , [GSE7803] , and [GSE6791] . For each downloaded data set we only selected the normal squamous cervix epithelial and squamous cervical cell carcinoma samples: for [GSE9750] there were 24 normal and 33 cancers, for [GSE7803] there were 10 normals and 21 cancers, and for [GSE6791] there were 8 normals and 20 cancers. Probes mapping to the same Entrez ID were averaged, resulting in 13,213 genes [GSE9750], 13,262 genes [GSE7803] and 20,827 genes [GSE6791].
mRNA expression data analysis
From the above three expression data sets we built an integrated (merged) expression set over 42 normal cervical epithelial samples and 74 cervical cancers using a procedure that we have validated many times previously [12–14]. Briefly, there were 13,213 genes in common between all three expression arrays. For each of these genes and for each expression set we then renormalized the gene expression profile to mean zero and unit variance, yielding 'z-profiles'. For each gene, its z-profile in each of the three studies was then merged. This resulted in a merged expression set over 13,213 genes and 116 samples (42 normal and 74 cancers). For each gene, we then computed a t-statistic P-value against normal/cancer status. Of the 140 risk genes, 86 were found in the merged expression set. Of these 86, 46 exhibited differential expression P-values < 0.05. A binomial test was used to test the significance of the skew towards differential under- or overexpression. To adjust for any global (that is, over all 13,213 genes) skew towards under- or overexpression, we also estimated the P-value using a Monte Carlo procedure (1,000 Monte Carlo runs) in which 86 genes were selected at random and a binomial test P-value was recomputed. The fraction of the 1,000 runs with a binomial test P-value more extreme than the observed gives an independent P-value estimate.
Full details of statistical methods are in Additional file 1. Brief descriptions of the different parts of the statistical analysis are given below.
Quality control and inter-array normalization
The raw DNAm data were subject to a similar quality control procedure as used in our previous publication .
To identify CpGs associated with age (aCpGs) we used surrogate variable analysis . False discovery rate (FDR) estimation was implemented as in the q-value package . To identify age-independent variable CpGs (vCpGs), we adjusted the data for age, and subsequently estimated the variances for each CpG. Because of the heteroscedasticity of β-values , we also estimated the variance using R-values (defined as R = M/U) . To identify differentially variable CpGs (DVCs) between prospective CIN2+ cases and normals, we used Bartlett's test . In doing so, variances were estimated after the methylation profiles were linearly adjusted for age within each phenotype.
EVORA: Epigenetic Variable Outliers for Risk Prediction Analysis
Full details and the model assumptions on which EVORA is based are described in Additional file 1. There are three important statistical aspects to EVORA: (i) identification of candidate risk CpGs; (ii) identification of samples that constitute methylation outliers; and (iii) a method for assigning risk to each sample, which is robust and independent of the scale used. For (i) we use Bartlett's test , since our hypothesis is that DVCs defined by outlier profiles are more likely to define risk CpGs . To define outliers in a scale-independent fashion (ii) we use the COPA (Cancer Outlier Profile Analysis) transformation . Lastly, to assign a risk score to a sample, we use an adaptive index algorithm framework . EVORA is freely available as an R-package (evora) .
DNA methylation variability is increased in cytologically normal cells predisposed to neoplasia
Age-associated variation in DNA methylation also correlates with the risk of neoplasia
An unsupervised singular value decomposition and a supervised linear regression analysis (adjusted for CIN2+ status) both confirmed a strong age-associated signature (644 CpGs at FDR < 0.05) with the majority (392, 61%) hypermethylated with age (Additional files 7 and 8). Interestingly, we observed that the 644 CpGs undergoing age-associated changes in DNAm (aCpGs) showed specific directional changes associated with the future development of CIN2+ even though these changes were not individually significant (Additional file 9). Specifically, those CpGs undergoing significant hypermethylation with age were generally also hypermethylated in future CIN2+ cases (Additional file 9). To test further if age-hypermethylated aCpGs are indeed associated with CIN2+, we compared their methylation levels in an independent data set (also generated with the Illumina 27 k platform) consisting of 30 normal and 18 age-matched CIN2+ samples (Set 1, Methods) . The mean methylation level of these aCpGs was also significantly higher in the CIN2+ samples of this set (Additional file 10).
Significant overlap of hypervariable DVCs and age-hypermethylated aCpGs
Differentially hypervariable CpGs predict risk of intraepithelial cervical neoplasia
Risk CpGs identified in normal cells can detect intraepithelial neoplasia and cervical cancer
Next, we explored the methylation profiles of the 140 risk CpGs in an independent set of cervical cancers and corresponding age-matched normal tissue (set 2; Methods). The outlier risk scores were significantly reduced (Figure 4c) owing to the fact that risk CpGs exhibited much more of a bi-modal methylation pattern between cancer and normal tissue; that is, risk CpGs were invariably either hemi- or fully methylated across a high proportion of the cancers (Figure 4d). Nevertheless, a small subset of risk CpGs retained their outlier profiles in cervical cancer and EVORA was therefore still highly predictive of cancer status (Figure 4c). Adapting EVORA to the original beta methylation scale to better capture the observed bi-modal methylation profiles yielded a perfect discrimination of normal and cancer tissue (Figures 3e and 4d).
Risk genes are underexpressed in cervical cancer
To evaluate if the genes associated with the identified risk CpGs show expression changes in cervical cancer, we built an integrated data set of mRNA expression profiles over 13,213 genes and encompassing 42 normal cervical epithelial and 74 cervical cancer (squamous cell carcinoma) specimens (Methods). The integration was done using a renormalization procedure that we have validated previously [12–14]. Of the 140 risk genes, 86 could be mapped to this merged data set. Comparison of average mRNA levels of our 86 risk genes between normal and cancer tissue confirmed that risk genes showed lower average expression levels in cervical cancer (Additional file 12). Analysis of individual gene statistics further showed that 46 were differentially expressed and that there was a significant skew towards underexpression in cancer (Additional file 12). Moreover, in only 13 of 1,000 random selections of 86 gene sets (Monte Carlo analysis described in Methods) did we observe a skew as significant as the one observed for the risk genes (Additional file 12).
In this work we have addressed two fundamental questions. First, do DNAm changes precede the morphological signs of neoplastic transformation? And second, can these epigenetic changes (a) predict the risk of neoplastic development, and (b) allow detection of early (non-invasive) cancers? We have addressed these questions in the context of the uterine cervix, currently the only human organ allowing relatively easy access to the cell of origin of the associated cancer well in advance of the first morphological signs of neoplastic transformation.
Based on the hypothesis that epigenetic variability may reflect exposure to various genetic and non-genetic risk factors (including age), and that differential epigenetic variability may reflect underlying differential exposure to these factors [1, 3], we posited that differentially variable methylated CpGs (DVCs) might be potential risk indicators of neoplastic transformation. Our prediction algorithm based on epigenetic variance (EVORA; Figure 3A) could predict, with low but statistically significant accuracy (AUC = 0.66, P < 0.05), prospective CIN2+ status in blind test sets (Figure 3b). In contrast, an analogous classifier that uses differential methylation instead of differential variability statistics could not predict prospective CIN2+ status (Figure 3c). Using EVORA, the set of risk CpGs could also accurately identify independent CIN2+ and cervical cancer samples (AUC > 0.9; Figures 3d, e and 4). Using these additional data sets, we also showed that differential variability is a more important characteristic of risk CpGs (supporting the findings in ) than PCGT status (Figure 6). Thus, these results support the view that differential variability is a key characteristic of cells predisposed to morphological transformation.
We have obtained these results in the context of cervical cancer using a study design that was matched for age and HPV status. Since HPV infection is a well-known necessary causal factor for cervical cancer , it is thus reasonable to assume that those HPV-negative samples that later developed a CIN2+ will have done so only by acquiring an HPV infection after sample collection . It is important, therefore, to point out that the increased DNAm variability at the identified loci only increases the risk of CIN2+ in the background of an existing HPV infection. Given that HPV infection is a necessary but not sufficient factor to cause cervical carcinogenesis (most HPV-infected women do not develop cervical cancer), it is indeed very likely that unknown 'tumor-suppressor' factors play a role in determining which HPV-infected women develop a CIN2+. Thus, our data support a model in which increased DNAm variability at the identified loci may compromise specific tumor suppressor functions, predisposing HPV-infected women to develop a CIN2+. In this context, it is also important to note that we did not observe HPV infection to have a major effect on DNAm patterns and the EVORA risk score itself was not correlated with HPV status (Additional file 13). Nevertheless, we did observe a very weak marginal association of DNAm with HPV status within only the prospective CIN2+ sample subgroup (Additional file 13). Thus, while increased DNAm variability may synergize with a background HPV infection to increase the risk of CIN2+, HPV infection itself only appears to have a minor effect on DNAm patterns in the earliest stages of cervical carcinogenesis. It is possible, however, that HPV infection may cause significant DNAm and mRNA changes during the latter stages of carcinogenesis, as reported, for instance, in the case of head and neck cancers [11, 28, 29].
In line with our hypothesis that epigenetic variable CpGs are involved in carcinogenesis, we find that they become more homogeneously methylated (that is, either hemi- or fully methylated) across cancers (Figure 4d), in stark contrast to their heterogeneous (that is, unmethylated versus hemi- or fully methylated) outlier profiles in potent CIN2+ and established CIN2+ cells (Figure 4a, b). This transition from heterogeneous outlier profiles in the earliest stages of cervical carcinogenesis to more homogeneous methylation profiles in the more advanced cancer stage points towards a dynamic process in which DNAm is initially acquired in a largely stochastic manner, as hypothesized in , but that later becomes more homogeneous and deterministic. However, we should point out that risk CpGs remain highly variably methylated across the invasive cancers, as previously reported in .
Interestingly, the heatmaps of the prospective and pre-invasive cancer series (Figures 4a, b) also revealed the existence of a striking bi-modality within the cases only, with a small subset of cases exhibiting a highly correlated pattern of CpG methylation, in contrast to the more stochastic pattern of methylation seen in the majority of samples. It is particularly interesting that the risk CpGs identified in ARTISTIC generated this same characteristic stochastic versus correlated bi-modal pattern in the CIN2+ samples of set 1. To investigate this further, we used an unsupervised clustering algorithm to divide up the CIN2+ samples (of set 1) and the cervical cancers (of set 2) each into two clusters according to the average level of methylation over all CpGs mapping to CpG islands excluding the 140 risk CpGs (Additional file 14). Interestingly, in both sets, the EVORA risk scores obtained from the 140 riskCpGs were significantly higher for the cases exhibiting the higher global levels of CpG island methylation (Additional file 14). Thus, the risk CpGs appear to correlate particularly strongly with CIN2+ samples and cervical cancers of overall higher CpG island methylation levels. It will be interesting, therefore, to explore the relationship, if any, of the identified risk CpGs to a potential CpG island methylator phenotype (CIMP) [30, 31] in cervical cancer. Using an unsupervised consensus clustering over the 1,000 most variable CpGs in the ARTISTIC set - that is, the unsupervised procedure used in [31, 32] to define a CIMP, did not yield a clear CIMP, nor did the clusters correlate with the EVORA risk scores, suggesting that there is no CIMP (in the sense defined in [31, 32]) present three years in advance of morphological transformation (Additional file 15).
Over the years extensive evidence has accumulated to demonstrate that PCGT methylation is crucially involved in carcinogenesis [33–37]. Although we have here provided substantial evidence that differential variability is a more fundamental characteristic of transformation than PCGT status (Figure 6), our analyses nevertheless implicate a subset of PCGTs in the earliest stages of carcinogenesis (69 out of the 140 risk CpGs mapped to PCGTs). Numerous risk CpGs were also annotated to genes that have been previously demonstrated to be linked to cervical cancer, including CALCA , CXCR4 , DCC , HOXC10 , MYOD1 , SFRP4 , SOX1 , TWIST1  and WT1 . However, we note that only HOXC10 had its risk CpG site located within 200 bp upstream of the transcription start site. In fact, of the 140 risk CpG sites, only 64 were within 200 bp of the start site, and of these only 24 were upstream. Thus, even though we observed the genes associated with our risk CpGs to be frequently underexpressed in cervical cancer, it remains to be seen if methylation of the identified risk CpG sites plays a part in the underlying regulatory mechanism.
Our data also revealed a remarkably strong overlap between CpGs undergoing age-associated hypermethylation in normal tissue and CpGs that were hypervariable in prospective CIN2+ cases. PCGTs made a large component of this overlap and age-associated hypermethylated CpGs also correlated significantly with CIN2+ status, further strengthening our previous findings . We stress that even those PCGT CpGs that were identified as undergoing age-associated hypermethylation in other tissues (for example, whole blood)  were also more hypervariable in prospective CIN2+ samples and could indicate the risk of neoplastic transformation. This strong overlap between hypermethylated aCpGs and hypervariable DVCs is consistent with a model in which differential exposure to risk factors accumulates with age, thus generating the observed stochastic epigenetic variability. That these methylation changes and overlaps are not due to changes in cell-type composition is supported by many studies [7, 46–48]. We also observed here that changes in cell-type composition were more likely to be captured by age-independent variable CpGs (vCpGs), defined as CpGs that showed the largest (age-adjusted) variation across all normal samples. Specifically, these maximally varying vCpGs exhibited fairly large (> 80%) changes in methylation, did not overlap with DVCs, and accordingly were neither enriched for PCGTs nor discriminatory of prospective CIN2+ status (Additional file 9). In fact, vCpGs were enriched for Gene Ontology terms related to extracellular space and mesenchymal features, including also many cell differentiation markers (Additional file 16), and thus it is possible that methylation variation in these CpGs reflects variations in the epithelial to mesenchymal and stromal cell ratio. Since risk CpGs were not enriched for these stromal and mesenchymal features, it supports the view that their observed methylation changes reflect clonal heterogeneity within the epithelial cell population only.
The reliability of the methylation data generated from the ARTISTIC cohort is strongly supported by emerging biology and successful validations in two independent data sets. Nevertheless, to further check the reliability of the Infinium methylation data, we compared the β-values to those obtained using Methylight. Matched Infinium-Methylight data were available for SOX1 and WT1 in 48 LBC samples (set 1) . We found statistically significant agreement between the two data types for both SOX1 and WT1 (Additional file 17).
Unfortunately, the limited coverage of the Infinium 27 K array (only 0.1% of the CpGs present in the human genome ) means that we could not fully explore the spatial methylation patterns around the identified risk CpG sites. As shown in Hansen et al.  in the context of cancer, differential variability is associated with increased spatial variability and loss of stability of the sharp methylation boundaries. Thus, it will be interesting to explore the spatial variability around the identified CpG risk sites with a more comprehensive and unbiased technology such as the Infinium 450 K methylation beadarray , as this may reveal a similar loss of methylation boundaries surrounding these sites.
We have demonstrated that variability in DNAm, which is associated with age and other factors, and which occurs well in advance (at least 3 years) of the first morphological neoplastic changes, is associated with the risk of neoplastic transformation. More generally, we have demonstrated that inter-individual epigenetic variance is an intrinsic characteristic of cells that become neoplastic.
A Randomised Trial of HPV Testing in Primary Cervical Screening
area under the curve
CpG island methylator phenotype
cervical intraepithelial neoplasia of grade 2 or higher
Cancer Outlier Profile Analysis
differentially methylated CpG
differentially variable CpG
Epigenetic Variable Outliers for Risk Prediction Analysis
false discovery rate
human papilloma virus
polycomb group target
This work was funded by the Eve Appeal, a grant from the UCLH/UCL Comprehensive Biomedical Research Centre, the NIHR Health Technology Assessment Programme, the Department of Health NIHR Biomedical Research Centres (BRC) funding scheme. AET was supported by a Heller Research Fellowship. HCK receives support from the Manchester NIHR BRC. We wish to thank the University of Manchester for financial support. We are particularly grateful to the women who have participated in the ARTISTIC trial and to the entire medical, nursing, and administrative staff who work on the ARTISTIC trial.
- Feinberg AP, Irizarry RA: Evolution in health and medicine Sackler colloquium: Stochastic epigenetic variation as a driving force of development, evolutionary adaptation, and disease. Proc Natl Acad Sci USA. 2010, 107 (Suppl 1): 1757-1764.View ArticlePubMed CentralPubMedGoogle Scholar
- Feinberg AP, Irizarry RA, Fradin D, Aryee MJ, Murakami P, Aspelund T, Eiriksdottir G, Harris TB, Launer L, Gudnason V, Fallin MD: Personalized epigenomic signatures that are stable over time and covary with body mass index. Sci Transl Med. 2010, 2: 49ra67Google Scholar
- Hansen KD, Timp W, Bravo HC, Sabunciyan S, Langmead B, McDonald OG, Wen B, Wu H, Liu Y, Diep D, Briem E, Zhang K, Irizarry RA, Feinberg AP: Increased methylation variation in epigenetic domains across cancer types. Nat Genet. 2011, 43: 768-775. 10.1038/ng.865.View ArticlePubMed CentralPubMedGoogle Scholar
- Bibikova M, Fan JB: Genome-wide DNA methylation profiling. Wiley Interdiscip Rev Syst Biol Med. 2010, 2: 210-223.View ArticlePubMedGoogle Scholar
- Kitchener HC, Almonte M, Gilham C, Dowie R, Stoykova B, Sargent A, Roberts C, Desai M, Peto J: ARTISTIC: a randomised trial of human papillomavirus (HPV) testing in primary cervical screening. Health Technol Assess. 2009, 13: 1-150, iii-iv.View ArticlePubMedGoogle Scholar
- Kitchener HC, Almonte M, Thomson C, Wheeler P, Sargent A, Stoykova B, Gilham C, Baysson H, Roberts C, Dowie R, Desai M, Mather J, Bailey A, Turner A, Moss S, Peto J: HPV testing in combination with liquid-based cytology in primary cervical screening (ARTISTIC): a randomised controlled trial. Lancet Oncol. 2009, 10: 672-682. 10.1016/S1470-2045(09)70156-1.View ArticlePubMedGoogle Scholar
- Teschendorff AE, Menon U, Gentry-Maharaj A, Ramus SJ, Weisenberger DJ, Shen H, Campan M, Noushmehr H, Bell CG, Maxwell AP, Savage DA, Mueller-Holzner E, Marth C, Kocjan G, Gayther SA, Jones A, Beck S, Wagner W, Laird PW, Jacobs IJ, Widschwendter M: Age-dependent DNA methylation of genes that are suppressed in stem cells is a hallmark of cancer. Genome Res. 2010, 20: 440-446. 10.1101/gr.103606.109.View ArticlePubMed CentralPubMedGoogle Scholar
- Edgar R, Domrachev M, Lash AE: Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res. 2002, 30: 207-210. 10.1093/nar/30.1.207.View ArticlePubMed CentralPubMedGoogle Scholar
- Scotto L, Narayan G, Nandula SV, Arias-Pulido H, Subramaniyam S, Schneider A, Kaufmann AM, Wright JD, Pothuri B, Mansukhani M, Murty VV: Identification of copy number gain and overexpressed genes on chromosome arm 20q by an integrative genomic approach in cervical cancer: potential role in progression. Genes Chromosomes Cancer. 2008, 47: 755-765. 10.1002/gcc.20577.View ArticlePubMedGoogle Scholar
- Zhai Y, Kuick R, Nan B, Ota I, Weiss SJ, Trimble CL, Fearon ER, Cho KR: Gene expression analysis of preinvasive and invasive cervical squamous cell carcinomas identifies HOXC10 as a key mediator of invasion. Cancer Res. 2007, 67: 10163-10172. 10.1158/0008-5472.CAN-07-2056.View ArticlePubMedGoogle Scholar
- Pyeon D, Newton MA, Lambert PF, den Boon JA, Sengupta S, Marsit CJ, Woodworth CD, Connor JP, Haugen TH, Smith EM, Kelsey KT, Turek LP, Ahlquist P: Fundamental differences in cell cycle deregulation in human papillomavirus-positive and human papillomavirus-negative head/neck and cervical cancers. Cancer Res. 2007, 67: 4605-4619. 10.1158/0008-5472.CAN-06-3619.View ArticlePubMed CentralPubMedGoogle Scholar
- Teschendorff AE, Miremadi A, Pinder SE, Ellis IO, Caldas C: An immune response gene expression module identifies a good prognosis subtype in estrogen receptor negative breast cancer. Genome Biol. 2007, 8: R157-10.1186/gb-2007-8-8-r157.View ArticlePubMed CentralPubMedGoogle Scholar
- Teschendorff AE, Gomez S, Arenas A, El-Ashry D, Schmidt M, Gehrmann M, Caldas C: Improved prognostic classification of breast cancer defined by antagonistic activation patterns of immune response pathway modules. BMC Cancer. 2010, 10: 604-10.1186/1471-2407-10-604.View ArticlePubMed CentralPubMedGoogle Scholar
- Teschendorff AE, Naderi A, Barbosa-Morais NL, Caldas C: PACK: Profile Analysis using Clustering and Kurtosis to find molecular classifiers in cancer. Bioinformatics. 2006, 22: 2269-2275. 10.1093/bioinformatics/btl174.View ArticlePubMedGoogle Scholar
- Leek JT, Storey JD: A general framework for multiple testing dependence. Proc Natl Acad Sci USA. 2008, 105: 18718-18723. 10.1073/pnas.0808709105.View ArticlePubMed CentralPubMedGoogle Scholar
- Storey JD, Tibshirani R: Statistical significance for genomewide studies. Proc Natl Acad Sci USA. 2003, 100: 9440-9445. 10.1073/pnas.1530509100.View ArticlePubMed CentralPubMedGoogle Scholar
- Du P, Zhang X, Huang CC, Jafari N, Kibbe WA, Hou L, Lin SM: Comparison of Beta-value and M-value methods for quantifying methylation levels by microarray analysis. BMC Bioinformatics. 2010, 11: 587-10.1186/1471-2105-11-587.View ArticlePubMed CentralPubMedGoogle Scholar
- Laird PW: Principles and challenges of genomewide DNA methylation analysis. Nat Rev Genet. 2010, 11: 191-203.View ArticlePubMedGoogle Scholar
- Snedecor GW, Cochran WG: Statistical Methods. 1989, Iowa State University PressGoogle Scholar
- Tomlins SA, Rhodes DR, Perner S, Dhanasekaran SM, Mehra R, Sun XW, Varambally S, Cao X, Tchinda J, Kuefer R, Lee C, Montie JE, Shah RB, Pienta KJ, Rubin MA, Chinnaiyan AM: Recurrent fusion of TMPRSS2 and ETS transcription factor genes in prostate cancer. Science. 2005, 310: 644-648. 10.1126/science.1117679.View ArticlePubMedGoogle Scholar
- Tian L, Tibshirani R: Adaptive index models for marker-based risk stratification. Biostatistics. 2011, 12: 68-86. 10.1093/biostatistics/kxq047.View ArticlePubMed CentralPubMedGoogle Scholar
- R: A language and environment for statistical computing. R Foundation for Statistical Computing. [http://www.r-project.org]
- Feinberg AP, Ohlsson R, Henikoff S: The epigenetic progenitor origin of human cancer. Nat Rev Genet. 2006, 7: 21-33. 10.1038/nrg1748.View ArticlePubMedGoogle Scholar
- Lee TI, Jenner RG, Boyer LA, Guenther MG, Levine SS, Kumar RM, Chevalier B, Johnstone SE, Cole MF, Isono K, Koseki H, Fuchikami T, Abe K, Murray HL, Zucker JP, Yuan B, Bell GW, Herbolsheimer E, Hannett NM, Sun K, Odom DT, Otte AP, Volkert TL, Bartel DP, Melton DA, Gifford DK, Jaenisch R, Young RA: Control of developmental regulators by Polycomb in human embryonic stem cells. Cell. 2006, 125: 301-313. 10.1016/j.cell.2006.02.043.View ArticlePubMed CentralPubMedGoogle Scholar
- Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, Mesirov JP: Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA. 2005, 102: 15545-15550. 10.1073/pnas.0506580102.View ArticlePubMed CentralPubMedGoogle Scholar
- Walboomers JM, Jacobs MV, Manos MM, Bosch FX, Kummer JA, Shah KV, Snijders PJ, Peto J, Meijer CJ, Munoz N: Human papillomavirus is a necessary cause of invasive cervical cancer worldwide. J Pathol. 1999, 189: 12-19. 10.1002/(SICI)1096-9896(199909)189:1<12::AID-PATH431>3.0.CO;2-F.View ArticlePubMedGoogle Scholar
- Kitchener HC, Gilham C, Sargent A, Bailey A, Albrow R, Roberts C, Desai M, Mather J, Turner A, Moss S, Peto J: A comparison of HPV DNA testing and liquid based cytology over three rounds of primary cervical screening: extended follow up in the ARTISTIC trial. Eur J Cancer. 2011, 47: 864-871. 10.1016/j.ejca.2011.01.008.View ArticlePubMedGoogle Scholar
- Sartor MA, Dolinoy DC, Jones TR, Colacino JA, Prince ME, Carey TE, Rozek LS: Genome-wide methylation and expression differences in HPV(+) and HPV(-) squamous cell carcinoma cell lines are consistent with divergent mechanisms of carcinogenesis. Epigenetics. 2011, 6: 777-787. 10.4161/epi.6.6.16216.View ArticlePubMed CentralPubMedGoogle Scholar
- Slebos RJ, Yi Y, Ely K, Carter J, Evjen A, Zhang X, Shyr Y, Murphy BM, Cmelak AJ, Burkey BB, Netterville JL, Levy S, Yarbrough WG, Chung CH: Gene expression differences associated with human papillomavirus status in head and neck squamous cell carcinoma. Clin Cancer Res. 2006, 12: 701-709. 10.1158/1078-0432.CCR-05-2017.View ArticlePubMedGoogle Scholar
- Weisenberger DJ, Siegmund KD, Campan M, Young J, Long TI, Faasse MA, Kang GH, Widschwendter M, Weener D, Buchanan D, Koh H, Simms L, Barker M, Leggett B, Levine J, Kim M, French AJ, Thibodeau SN, Jass J, Haile R, Laird PW: CpG island methylator phenotype underlies sporadic microsatellite instability and is tightly associated with BRAF mutation in colorectal cancer. Nat Genet. 2006, 38: 787-793. 10.1038/ng1834.View ArticlePubMedGoogle Scholar
- Noushmehr H, Weisenberger DJ, Diefes K, Phillips HS, Pujara K, Berman BP, Pan F, Pelloski CE, Sulman EP, Bhat KP, Verhaak RG, Hoadley KA, Hayes DN, Perou CM, Schmidt HK, Ding L, Wilson RK, Van Den Berg D, Shen H, Bengtsson H, Neuvial P, Cope LM, Buckley J, Herman JG, Baylin SB, Laird PW, Aldape K, Cancer Genome Atlas Research Network: Identification of a CpG island methylator phenotype that defines a distinct subgroup of glioma. Cancer Cell. 2010, 17: 510-522. 10.1016/j.ccr.2010.03.017.View ArticlePubMed CentralPubMedGoogle Scholar
- Fang F, Turcan S, Rimner A, Kaufman A, Giri D, Morris LG, Shen R, Seshan V, Mo Q, Heguy A, Baylin SB, Ahuja N, Viale A, Massague J, Norton L, Vahdat LT, Moynahan ME, Chan TA: Breast cancer methylomes establish an epigenomic foundation for metastasis. Sci Transl Med. 2011, 3: 75ra25Google Scholar
- Sauvageau M, Sauvageau G: Polycomb group proteins: multi-faceted regulators of somatic stem cells and cancer. Cell Stem Cell. 2010, 7: 299-313. 10.1016/j.stem.2010.08.002.View ArticlePubMedGoogle Scholar
- Bracken AP, Helin K: Polycomb group proteins: navigators of lineage pathways led astray in cancer. Nat Rev Cancer. 2009, 9: 773-784. 10.1038/nrc2736.View ArticlePubMedGoogle Scholar
- Johnstone SE, Baylin SB: Stress and the epigenetic landscape: a link to the pathobiology of human diseases?. Nat Rev Genet. 2010, 11: 806-812.View ArticlePubMed CentralPubMedGoogle Scholar
- Cedar H, Bergman Y: Linking DNA methylation and histone modification: patterns and paradigms. Nat Rev Genet. 2009, 10: 295-304.View ArticlePubMedGoogle Scholar
- Rodriguez-Paredes M, Esteller M: Cancer epigenetics reaches mainstream oncology. Nat Med. 2011, 17: 330-339.View ArticlePubMedGoogle Scholar
- Wisman GB, Nijhuis ER, Hoque MO, Reesink-Peters N, Koning AJ, Volders HH, Buikema HJ, Boezen HM, Hollema H, Schuuring E, Sidransky D, van der Zee AG: Assessment of gene promoter hypermethylation for detection of cervical neoplasia. Int J Cancer. 2006, 119: 1908-1914. 10.1002/ijc.22060.View ArticlePubMedGoogle Scholar
- Herman L, Hubert P, Herfs M, Kustermans G, Henrotin Y, Bousarghin L, Boniver J, Delvenne P: The L1 major capsid protein of HPV16 differentially modulates APC trafficking according to the vaccination or natural infection context. Eur J Immunol. 2010, 40: 3075-3084. 10.1002/eji.201040571.View ArticlePubMedGoogle Scholar
- Kersemaekers AM, van de Vijver MJ, Kenter GG, Fleuren GJ: Genetic alterations during the progression of squamous cell carcinomas of the uterine cervix. Genes Chromosomes Cancer. 1999, 26: 346-354. 10.1002/(SICI)1098-2264(199912)26:4<346::AID-GCC9>3.0.CO;2-D.View ArticlePubMedGoogle Scholar
- Widschwendter A, Muller HM, Fiegl H, Ivarsson L, Wiedemair A, Muller-Holzner E, Goebel G, Marth C, Widschwendter M: DNA methylation in serum and tumors of cervical cancer patients. Clin Cancer Res. 2004, 10: 565-571. 10.1158/1078-0432.CCR-0825-03.View ArticlePubMedGoogle Scholar
- Lin YW, Chung MT, Lai HC, De Yan M, Shih YL, Chang CC, Yu MH: Methylation analysis of SFRP genes family in cervical adenocarcinoma. J Cancer Res Clin Oncol. 2009, 135: 1665-1674. 10.1007/s00432-009-0613-5.View ArticlePubMedGoogle Scholar
- Lai HC, Lin YW, Huang TH, Yan P, Huang RL, Wang HC, Liu J, Chan MW, Chu TY, Sun CA, Chang CC, Yu MH: Identification of novel DNA methylation markers in cervical cancer. Int J Cancer. 2008, 123: 161-167. 10.1002/ijc.23519.View ArticlePubMedGoogle Scholar
- Missaoui N, Hmissa S, Trabelsi A, Traore C, Mokni M, Dante R, Frappart L: Promoter hypermethylation of CDH13, DAPK1 and TWIST1 genes in precancerous and cancerous lesions of the uterine cervix. Pathol Res Pract. 2011, 207: 37-42. 10.1016/j.prp.2010.11.001.View ArticlePubMedGoogle Scholar
- Lin CJ, Lai HC, Wang KH, Hsiung CA, Liu HW, Ding DC, Hsieh CY, Chu TY: Testing for methylated PCDH10 or WT1 is superior to the HPV test in detecting severe neoplasms (CIN3 or greater) in the triage of ASC-US smear results. Am J Obstet Gynecol. 2011, 204: 21 e21-27.View ArticleGoogle Scholar
- Maegawa S, Hinkal G, Kim HS, Shen L, Zhang L, Zhang J, Zhang N, Liang S, Donehower LA, Issa JP: Widespread and tissue specific age-related DNA methylation changes in mice. Genome Res. 2010, 20: 332-340. 10.1101/gr.096826.109.View ArticlePubMed CentralPubMedGoogle Scholar
- Rakyan VK, Down TA, Maslau S, Andrew T, Yang TP, Beyan H, Whittaker P, McCann OT, Finer S, Valdes AM, Leslie RD, Deloukas P, Spector TD: Human aging-associated DNA hypermethylation occurs preferentially at bivalent chromatin domains. Genome Res. 2010, 20: 434-439. 10.1101/gr.103101.109.View ArticlePubMed CentralPubMedGoogle Scholar
- Teschendorff AE, Menon U, Gentry-Maharaj A, Ramus SJ, Gayther SA, Apostolidou S, Jones A, Lechner M, Beck S, Jacobs IJ, Widschwendter M: An epigenetic signature in peripheral blood predicts active ovarian cancer. PLoS One. 2009, 4: e8274-10.1371/journal.pone.0008274.View ArticlePubMed CentralPubMedGoogle Scholar
- Apostolidou S, Hadwin R, Burnell M, Jones A, Baff D, Pyndiah N, Mould T, Jacobs IJ, Beddows S, Kocjan G, Widschwendter M: DNA methylation analysis in liquid-based cytology for cervical cancer screening. Int J Cancer. 2009, 125: 2995-3002. 10.1002/ijc.24745.View ArticlePubMedGoogle Scholar
- Beck S: Taking the measure of the methylome. Nat Biotechnol. 2010, 28: 1026-1028. 10.1038/nbt1010-1026.View ArticlePubMedGoogle Scholar
- Sandoval J, Heyn HA, Moran S, Serra-Musach J, Pujana MA, Bibikova M, Esteller M: Validation of a DNA methylation microarray for 450,000 CpG sites in the human genome. Epigenetics. 2011, 6: 692-702. 10.4161/epi.6.6.16196.View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.