Sage Journals: Discover world-class research

Abstract

Multicomponent molecular modifications such as DNA methylation may offer sensitive and specific cervical intraepithelial neoplasia and cervical cancer biomarkers. In this study, we tested cervical tissues at various stages of tumor progression for 5-methylcytosine and 5-hydroxymethylcytosine levels and also DNA promoter methylation profile of a panel of genes for its diagnostic potential. In total, 5-methylcytosine, 5-hydroxymethylcytosine, and promoter methylation of 33 genes were evaluated by reversed-phase high-performance liquid chromatography, enzyme-linked immunosorbent assay based technique, and bisulfate-based next generation sequencing. The 5-methylcytosine and 5-hydroxymethylcytosine contents were significantly reduced in squamous cell carcinoma and receiver operating characteristic curve analysis showed a significant difference in (1) 5-methylcytosine between normal and squamous cell carcinoma tissues (area under the curve = 0.946) and (2) 5-hydroxymethylcytosine levels among normal, squamous intraepithelial lesions and squamous cell carcinoma. Analyses of our next generation sequencing results and data from five independent published studies consisting of 191 normal, 10 low-grade squamous intraepithelial lesions, 21 high-grade squamous intraepithelial lesions, and 335 malignant tissues identified a panel of nine genes (ARHGAP6, DAPK1, HAND2, NKX2-2, NNAT, PCDH10, PROX1, PITX2, and RAB6C) which could effectively discriminate among the various groups with sensitivity and specificity of 80%–100% (p < 0.05). Furthermore, 12 gene promoters (ARHGAP6, HAND2, LHX9, HEY2, NKX2-2, PCDH10, PITX2, PROX1, TBX3, IKBKG, RAB6C, and DAPK1) were also methylated in one or more of the cervical cancer cell lines tested. The global and gene-specific methylation of the panel of genes identified in our study may serve as useful biomarkers for the early detection and clinical management of cervical cancer.

Keywords

Cervical cancer DNA methylation methylation biomarkers sensitivity specificity diagnosis

Introduction

Cervical cancer is one of the major causes of cancer-related deaths among women worldwide and is strongly associated with high-risk human papillomavirus (HPV) infection. Although early cytomorphological screening and vaccines against HPV are anticipated to reduce the cervical cancer burden,¹ it is projected that by the year 2020, there will be 42% increase in cervical cancer incidence globally (http://globocan.iarc.fr/Default.aspx). In India, cervical cancer alone accounted for 24% (62,416) of all cancer cases in women who succumbed to the disease in the period 2015–2016. Epigenetic mechanisms leading to global hypomethylation and gene-specific hypermethylation play a key role in carcinogenesis, including cervical cancer.^2,3 Removal of methyl group from the cytosine residue of the DNA resulting in global hypomethylation is mediated by several non-enzymatic mechanisms or by the action of ten–eleven translocation (TET) family of alpha-glutarate-dependent oxygenases.⁴ Oxidative by-products of 5-hydroxymethylcytosine (5-hmC), such as 5-formylcytosine (5-fC) and 5-carboxylcytosine (5-caC), are part of cytosine deaminase pathway and participate in DNA repair process.⁵ Recently, role of 5-fC and 5-caC in affecting the rate of polymerase II nucleotide incorporation, thereby influencing the rate of transcription, is also reported.⁶ It is now understood that “cytosine codes” (various modifications of cytosine such as 5-mC, 5-hmC, 5-fC or 5-caC) govern epigenetic plasticity that may regulate gene expression.

The aberrant methylation of genes has been reported as an early and frequent event during cervical carcinogenesis.⁷ In addition to their role in transcription silencing and tumor development, the hypermethylation of promoters at specific CpG sites can be sensitive and specific diagnostic and prognostic markers.^8,9 For example, DNA promoter methylation of glutathione S-transferase pi 1 (GSTP1) and O-6-methylguanine-DNA methyltransferase (MGMT1) is reported as a potential diagnostic marker for glioblastoma and prostate cancer.^10–12 It has been shown that 95% of the adenocarcinoma of the prostate have methylated GSTP1 promoter, while the normal tissues were apparently negative.¹³ The MGMT1 DNA methylation at specific CpG sites is shown to be an important prognostic marker for gliobastoma and response to treatment.¹⁰ The diagnostic and prognostic significance of methylated genes in cervical cancer have been extensively reviewed.^2,14,15 A majority of the cervical cancer DNA methylation studies have validated genes which are also reported in other cancers, and only a few of these markers are specific for cervical cancer with high sensitivity and specificity.¹⁶ In addition, studies have implicated ethnicity as one of the confounding factors for the development of DNA methylation–based biomarkers. For example, cyclin-dependent kinase inhibitor 2A (p16^INK4a) gene is hypermethylated in individuals with hepatocellular carcinoma (HCC) from Japan and is unmethylated in individuals from Taiwan.^17,18

Although DNA methylation changes represent one of the stable and earliest changes during carcinogenesis, efforts are needed to discover and validate these findings using population-based studies. There is also a need for the novel targets useful for diagnostics, prognostics, and therapeutic interventions, which can be used uniformly across the population. In this study, we have identified new DNA methylation markers with high sensitivity and specificity for cervical cancer screening. Furthermore, we have tested the global 5-methylcytosine (5-mC) and 5-hmC levels, evaluated the methylation status of large number of gene promoters in a panel of tissues representing progressive stages of cervical cancer, and performed in silico validation in independent datasets.

Materials and methods

Cell line maintenance and patient cohort

Three cervical cancer cell lines (SiHa, HeLa, and CaSki) were purchased from American Type Culture Collection (ATCC, USA) and subcultured according to the standard protocol (http://www.atcc.org/). Samples of the human subjects who visited either for screening or for the diagnosis or treatment of cervical cancer were obtained from hospitals affiliated to the Kasturba Medical College (KMC) (Manipal and Mangalore, Karnataka, India). In total, three lymphoblastoid cell lines (LCLs) were generated in our laboratory by infecting peripheral blood mononuclear cells (PBMCs) with Epstein–Barr virus (EBV), as published earlier.¹⁹ In brief, PBMCs were separated from 10 mL of blood by Ficoll-Hypaque lymphocyte separation medium (GE Healthcare, USA), according to manufacturer’s instructions, and infected with EBV isolated from marmoset B-cell line (B95-8), as published earlier.¹⁹ The transformed cell lines were cultured in RPMI containing 10% fetal bovine serum (FBS), confirmed by light microscopy and immunophenotyping using CD23, CD58, and CD19 antibodies.

Individuals who were already on the treatment or those who had synchronous cancers were excluded from the study. Samples were collected after obtaining the written informed consent from the participants in accordance with ethics committee of Kasturba Hospital (KH) Manipal, Manipal University. The samples were histologically reviewed by pathologists and classified according to the Bethesda system.²⁰ Histopathologically confirmed premalignant and squamous cell carcinoma (SCC) cells that contained more than 80% of abnormal and tumor cells were used for DNA isolation. The global 5-mC levels were estimated in the DNA of 20 normal, 35 SCC, 3 cervical cancer cell lines (SiHa, HeLa, and CaSki), and 3 LCLs. The 5-hmC content was estimated in 10 each of normal, squamous intraepithelial lesions (SILs), SCC, and 3 cervical cancer cell lines (SiHa, HeLa, and CaSki). For next generation sequencing (NGS)-based bisulfite genomic sequencing, 20 each of normal, SIL, SCC, and 3 cervical cancer cell lines were used.

DNA extraction and bisulfite conversion

DNA was isolated by digesting the cells/tissues with proteinase K followed by standard phenol chloroform and ethanol precipitation method, while bisulfite modification was performed using EZ DNA Methylation Kit (Zymo Research, USA). Bisulfite-converted DNA (100 ng) was amplified using forward and reverse primers under the following conditions: 95°C for 5 min; 34 cycles of 95°C for 30 s, respective annealing temperature for 45 s, 72°C for 1 min; and a final extension of 72°C for 5 min (Table 1). The polymerase chain reaction (PCR) products were gel purified and used for NGS.^21,22

Table 1.

The list of primers used in this study.

Gene	Primer sequence (5′–3′)	Annealing temperature (°C)	Amplicon length (bp)	Chromosomal coordinates	CpG island	Position with respect to Transcription start site (TSS)
ARHGAP6	F: TTTATTAAGAAAGTGGAGGGG	55	439	ChrX:11,684,340–11,684,779	Yes	−958 to −519
ARHGAP6	R: AAAAACACAAACACAAAATAAACA	55	439	ChrX:11,684,340–11,684,779	Yes	−958 to −519
ATM	F: GTTTTYGGAATTGTYGTTATTT	55	454	Chr11:108,093,462–108,093,916	Yes	−96 to 358
ATM	R: ACCACTACACTCRAAAAATCAAAAT	55	454	Chr11:108,093,462–108,093,916	Yes	−96 to 358
CDK6	F: GYGGGGAGTAAGTAAAGTTAGAT	57.5	439	Chr7:92,462,325–92,462,764	Yes	467 to 906
CDK6	R: TCTAAACCTAAAAATTCCCRACT	57.5	439	Chr7:92,462,325–92,462,764	Yes	467 to 906
DAPK1	F: GGTAGTTTAGTAATGTGTTATAGGTGG	57.5	391	Chr9:90,112,486–90,112,877	Yes	−171 to 220
DAPK1	R: AAACCCTACCRCTACRAAT	57.5	391	Chr9:90,112,486–90,112,877	Yes	−171 to 220
DDB1	F: TTTTTAGGAYGAAAGGGGAT	57.5	361	Chr11:61,100,153–61,100,514	Yes	170 to 531
DDB1	R: CACCTATCTTTTCRCTTATATCCC	57.5	361	Chr11:61,100,153–61,100,514	Yes	170 to 531
DGKQ	F: GGGCCTAACTGTGGATAGGAGT	57.5	401	Chr4:967,179–967,580	Yes	−232 to 169
DGKQ	R: ACAGCCCCTAGGACAGGT	57.5	401	Chr4:967,179–967,580	Yes	−232 to 169
DLX1	F: AGAGGTAGAGGTTAGTTTTTTTTTG	55	443	Chr2:172,949,774–172,950,217	Yes	−433 to 10
DLX1	R: ATTCAAAACTTTAACTCAAAC	55	443	Chr2:172,949,774–172,950,217	Yes	−433 to 10
DPYD	F: AGTYGTAGGATYGAGAGYGTAG	55	439	Chr1:98,386,231–98,386,670	Yes	−55 to 384
DPYD	R: AACRAAAAATCCTCCCCTAAACT	55	439	Chr1:98,386,231–98,386,670	Yes	−55 to 384
FGF18	F: AGGAAYGAAGTTGTTTAGTATGT	55	355	Chr5:170,845,717–170,846,072	Yes	−949 to −594
FGF18	R: AACCACCCTCTTAACTAACTACC	55	355	Chr5:170,845,717–170,846,072	Yes	−949 to −594
GBF1	F: TTTTGGTTTAGGTAATTTTGATAATA	57.5	462	Chr10:104,004,859–104,005,321	Yes	−395 to 67
GBF1	R: ACCTTCCCTATCAACTCACTA	57.5	462	Chr10:104,004,859–104,005,321	Yes	−395 to 67
HAND2	F: GAGACTGCTGGCCTTGGTG	55	399	Chr4:174,452,424–174,452,823	Yes	−2293 to −1894
HAND2	R: ATACTCAAATAAAATCTTTAAATC	55	399	Chr4:174,452,424–174,452,823	Yes	−2293 to −1894
HEY2	F: TAATTTGGAGAGAATTAGTAGTTAGA	55	390	Chr6:126,069,672–126,070,062	Yes	−1059 to −669
HEY2	R: TACTTCCATTAAATCTTCAACC	55	390	Chr6:126,069,672–126,070,062	Yes	−1059 to −669
HOXA7	F: GTTGTTGTAGATAGAAGGAATTTTAAA	57.5	389	Chr7:27,200,737–27,201,126	Yes	−4830 to −4441
HOXA7	R: CTCTTACATAAAACCACACTACCA	57.5	389	Chr7:27,200,737–27,201,126	Yes	−4830 to −4441
HOXC9	F: GTTAAGTTTTAGTTTTTTTATT	52.5	440	Chr12:54,393,460–54,393,900	Yes	−416 to 24
HOXC9	R: CAAATCAAATAATTATTTTTCCAACC	52.5	440	Chr12:54,393,460–54,393,900	Yes	−416 to 24
IKBKG	F: GTTTTAGGGGTTTAAGAGAGGAGG	60	503	ChrX:153,774,837–153,775,340	Yes	−724 to −221
IKBKG	R: AAACACTACRCRAAACTACACC	60	503	ChrX:153,774,837–153,775,340	Yes	−724 to −221
IREB2	F: AGAAAATTTAAGGTYGTAATTAAGTAA	57.5	402	Chr15:78,730,302–78,730,704	Yes	−215 to 187
IREB2	R: AAACCCRAAACTAACCTACTTTTAA	57.5	402	Chr15:78,730,302–78,730,704	Yes	−215 to 187
ITGAV	F: AGTYGGAGGGAAGTAAAGGAT	57.5	377	Chr2:187,454,817–187,455,194	Yes	28 to 405
ITGAV	R: TCCCTCRAAACCAAAATACTC	57.5	377	Chr2:187,454,817–187,455,194	Yes	28 to 405
ITPA	F: GGAGGGGTGGGTTATTTAGT	60	426	Chr20:3,189,964–3,190,390	Yes	−41 to 385
ITPA	R: CATTCTTCCRAACTAAATTCTATTC	60	426	Chr20:3,189,964–3,190,390	Yes	−41 to 385
LHX9	F: AAGAGGTTAAATTYGGTTGTTTT	57.5	404	Chr1:197,879,639–197,880,043	Yes	−1995 to −1591
LHX9	R: TCCRCCATACAATTTAAAATC	57.5	404	Chr1:197,879,639–197,880,043	Yes	−1995 to −1591
LRP5	F: GGAGTYGGTTYGTAAGTTTGTTT	55	329	Chr11:68,080,660–68,080,989	Yes	553 to 882
LRP5	R: CTACCTCTCCCCACRAAAATATAC	55	329	Chr11:68,080,660–68,080,989	Yes	553 to 882
LRPPRC	F: TGGGGAAGAGAGTTAGTTTAATTT	57.5	443	Chr2:44,223,363–44,223,806	Yes	−662 to −219
LRPPRC	R: CCCCTACACTCCTCTATTAATAAAAC	57.5	443	Chr2:44,223,363–44,223,806	Yes	−662 to −219
NKX2-2	F: TTTTTTAGGGGAAAAAAGGG	60	409	Chr20:21,496,571–21,496,980	Yes	−2316 to −1907
NKX2-2	R: CCAAATTAAAACAAACCCCTC	60	409	Chr20:21,496,571–21,496,980	Yes	−2316 to −1907
NME6	F: AATGTATGYGTTTGGAAAATT	55	426	Chr3:48,342,596–48,343,022	Yes	−174 to 252
NME6	R: CCACTCRCTTCCTAAAAAAC	55	426	Chr3:48,342,596–48,343,022	Yes	−174 to 252
NNAT	F: ATTTATTAGGGTTTGGGGG	57.5	424	Chr20:36,148,834–36,149,258	Yes	−772 to −348
NNAT	R: ATCATCTACCCCATAAAACAAA	57.5	424	Chr20:36,148,834–36,149,258	Yes	−772 to −348
PCDH10	F: TTTTAGGATTGGTAGAAATTAGTG	55	405	Chr4:134,069,437–134,069,842	Yes	−1032 to −627
PCDH10	R: CCTAAAAAACRAAAAAATTCA	55	405	Chr4:134,069,437–134,069,842	Yes	−1032 to −627
PCGF2	F: AYGGGGTAAGTTGGGTTAGG	57.5	409	Chr17:36,904,279–36,904,688	Yes	−127 to 282
PCGF2	R: ATCTAAAAAAAAAAAACCATTTTATC	57.5	409	Chr17:36,904,279–36,904,688	Yes	−127 to 282
PITX2	F: TTTTTTATTTTTAGGGGAAATTGT	60	461	Chr4:111,549,645–111,550,106	Yes	−5852 to −5391
PITX2	R: AAAAAACTTTCCTAATCAAAACCC	60	461	Chr4:111,549,645–111,550,106	Yes	−5852 to −5391
POLG	F: TTAAGGAGTAGGTGGATTGGTTT	57.5	429	Chr15:89,878,217–89,878,646	Yes	−620 to −191
POLG	R: TAAAATTTCTATACAAAATACCCCTCC	57.5	429	Chr15:89,878,217–89,878,646	Yes	−620 to −191
PROX1	F: AGTTTGTGTTAATTTGTATATAGYGTT	60	465	Chr1:214,158,638–214,159,103	Yes	−2639 to −2174
PROX1	R: CTCRAACCTCAAATCTTATCCA	60	465	Chr1:214,158,638–214,159,103	Yes	−2639 to −2174
RAB6C	F: GATGTTGTTTTAGTYGGTTGA	57.5	395	Chr2:132,121,344–132,121,739	Yes	−739 to −345
RAB6C	R: AACTCRCRCACTCAACAA	57.5	395	Chr2:132,121,344–132,121,739	Yes	−739 to −345
RBBP8	F: GATTGTTTTTGGTTTGGAGTT	60	330	Chr18:20,513,576–20,513,906	Yes	282 to 612
RBBP8	R: AAAACTTTCCAACCAAATCC	60	330	Chr18:20,513,576–20,513,906	Yes	282 to 612
SLC5A6	F: ATAGGGAAGGTTTTTATTTTTGAAGA	57.5	343	Chr2:27,435,469–27,435,812	Yes	−637 to −294
SLC5A6	R: CAACCAAATCCTATTCCCAA	57.5	343	Chr2:27,435,469–27,435,812	Yes	−637 to −294
TBX3	F: GTTTGAAGGTGATAGGGATTT	57.5	400	Chr12:115,106,906–115,107,306	No	14,663 to 15,063
TBX3	R: AATAATTCCAAAATACACATTTC	57.5	400	Chr12:115,106,906–115,107,306	No	14,663 to 15,063

F: forward primer; R: reverse primer; TSS: Transcription start site.

Methylation analysis by NGS

The bisulfite PCR product was purified and used for library construction according to Ion Xpress™ Plus Fragment Library Kit (Thermo Fisher Scientific, USA). In brief, PCR products of eight genes/samples were pooled at a final concentration of 25 nM in 100 µL and were fragmented using Ion Shear™ Plus Reagents according to manufacturer’s protocol (Thermo Fisher Scientific, USA). The fragmented DNAs were purified using Agencourt AMPure XP beads (Beckman Coulter, USA), ligated with Ion P1 adapter, and individual samples were barcoded with Ion Xpress barcode. The adapter-ligated, nick-translated, repaired DNA was purified using Agencourt AMPure XP beads. Approximately, 330 bp target library was prepared from unamplified library (200 base read length) using E-Gel SizeSelect™ 2% Agarose Gel (Thermo Fisher Scientific, USA). The size selected product (200 base read with target length of 330 bp) was then amplified and purified using Ion Xpress™ Plus Fragment Library Kit. The quality and quantity of the amplified library were analyzed by High Sensitivity DNA kit using Bioanalyzer (Agilent Technologies, USA) and Qubit^® dsDNA HS Assay Kit using Qubit fluorometer (Thermo Fisher Scientific, USA). Each pooled library (100 pM) was subjected to emulsion PCR, and enrichment was carried out using Ion PGM™ Hi-Q™ OT2 reagent in Ion OneTouch™ 2 system (Thermo Fisher Scientific, USA). The templates for sequencing were enriched by Ion OneTouch™ ES and sequenced using 316v2 Chips in the Ion Personal Genome Machine (PGM) Sequencer (Thermo Fisher Scientific, USA).

Analysis of NGS data

We have used the following pipeline to analyze bisulfite sequencing raw reads. In brief, the raw reads were analyzed for the sequence quality using FastQC. Nucleotides with quality scores <20 were trimmed from the ends using FASTX-Toolkit (version 0.0.14; http://hannonlab.cshl.edu/fastx_toolkit/). After mapping the sequences to the genomic regions, it was subsequently analyzed using NextGenMap 0.4.12.²³ Sequences were also aligned against converted and unconverted reference genomic regions to analyze the extent of bisulfite conversion. The degree of methylation at each CpG site was estimated using BiQAnalyzer HT software (http://biq-analyzer-ht.bioinf.mpi-inf.mpg.de/).²⁴

HPV DNA typing

The HPV genotyping was performed, as published earlier, with appropriate positive and negative controls.^25,26 The HPV strains were identified by National Center for Biotechnology Information’s Basic Local Alignment Search Tool (NCBIBLAST) search and HPV database (https://pave.niaid.nih.gov/). HPV negative samples were scored as negative after at least two independent rounds of testing.

Global methylation analysis by reversed-phase high-performance liquid chromatography

Global 5-mC levels were estimated using reversed-phase high-performance liquid chromatography (RP-HPLC) as described earlier.^27–29 Briefly, 2 µg of genomic DNA was digested with DNase1 (Ambion, USA), denatured at 100°C for 5 min, chilled on ice, and incubated with 2 U each of Nuclease P1 (Sigma, USA) and calf intestinal phosphatase (New England BioLabs, USA) according to manufacturer’s instruction. The sample was passed through Quia quick spin column (Qiagen, USA), lyophilized, and subjected to chromatographic separation (10 µL) under ambient column temperature, maintaining the samples at 4°C using isocratic conditions with a flow rate of 1 mL/min in HPLC instrument (Waters, USA), consisting of dipotassium hydrogen phosphate buffer (K₂HPO₄, pH 4.1) supplemented with 10% methanol and C-18 Grace Vydac smart column at 260 and 280 nm. Standards such as dC, 5-dmC, dT, dA, and dG were used to identify the retention time of these nucleosides. Peaks corresponding to 5-mC and other nucleosides were monitored over 30 min interval. The relative degree of methylation in the DNA samples was measured using the following formula: % of 5-mdC = A5mdC / (A5mdC + AdC) × 100.

Global 5-hmC estimation

The global 5-hmc was estimated using MethylFlash™ Hydroxymethylated DNA Quantification Kit (EpiGentek, USA) according to manufacturer’s instruction. The 5-hmC was estimated in 200 ng of DNA using the capture and detection antibodies followed by colorimetric detection of absorbance at 450 nm using Varioskan™ Flash Multimode Reader (Thermo Fisher Scientific, USA). The amount of hydroxymethylated DNA is calculated using the following formula according to the manufacturer’s instruction

5 - hmC % = \frac{(Sample OD - HC 4 OD) / S}{(HC 5 OD-HC 4 OD) \times 5 * / P} \times 100 %

where S is the amount of input sample DNA in nanograms; P is the amount of input positive control (HC5) in nanograms; 5* is a factor to normalize 5-hmC in the positive control to 100%, as the positive control contains only 20% of 5-hmC; HC4 is the negative control which is a methylated polynucleotide containing 20% of 5-mC. The HC5 positive control is a hydroxymethylated polynucleotide containing 20% of 5-hmC.

Public datasets

Independent validation of our findings was performed using cervical cancer methylation and gene expression microarray datasets, available at ArrayExpress (https://www.ebi.ac.uk/arrayexpress/) and the Cancer Genome Atlas (TCGA; https://tcga-data.nci.nih.gov/tcga/) using cancer browser (https://genome-cancer.ucsc.edu) or cBioPortal (www.cbioportal.org). A total of four independent datasets were created consisting of 557 microarrays for methylation and 66 microarrays for gene expression. The methylation dataset consisted of 191 normal, 10 low-grade SILs (LSILs), 21 high-grade SILs (HSILs), and 335 tumor samples, while the gene expression dataset consisted of 24 normal, 33 tumor, and 3 cervical cancer cell lines (Supplementary Table S1). DNA methylation and expression analysis were also performed using Methylation and Expression Database of Normal and Tumor Tissues (MENT; http://mgrc.kribb.re.kr:8080/MENT/), a database of DNA Methylation and gene expression in Human Cancer (MethHC; http://methhc.mbc.nctu.edu.tw/), Gene Expression across of Normal and Tumor Tissues (GENT; http://medical-genome.kribb.re.kr/GENT/), and Wanderer (http://maplab.imppc.org/wanderer/) databases. HumanMethylation27 and HumanMethylation450 BeadChip data were extracted from ArrayExpress (E-GEOD-30760, E-GEOD-30759, E-GEOD-41384, and E-GEOD-46306 datasets) database. The clinical information and detailed description of the dataset are available at ArrayExpress. The cervical SCC and endocervical adenocarcinoma (CESC) dataset consisting of DNA methylation data (n = 260; Illumina Infinium Human methylation 450 K bead array, Illumina, USA) and RNA-seq data (n = 256) were available at the time of our analysis. A β value ≥0.2 was considered as methylated. The β value was considered as significantly hypermethylated only if the value was found in more than 5% of tumors, as published earlier.³⁰

Bioinformatics and statistical analysis

Circos and CRAN Package “pheatmap” (https://cran.r-project.org/web/packages/pheatmap/index.html) were used for circular representation of the significantly differentially methylated regions and heat maps, respectively.³¹ Results of all the assays were shown as mean ± standard deviation (SD). Student’s t test, analysis of variance (ANOVA), Kruskal–Wallis test, and Fisher’s exact test were performed using Microsoft Excel (Microsoft, USA) and GraphPad Prism 5 (GraphPad Software, USA). A p value ≤ 0.05 was considered as statistically significant. The sensitivity and specificity analysis were performed by MedCalc (http://www.medcalc.org/calc/diagnostic_test.php) and GraphPad Prism 5.

For random forest (RF) analysis, we have averaged the methylation value of all the CpG sites within the given region of each gene for each sample. For RF analysis, methylation was considered as an independent variable, tumor condition was considered a dependent variable, and 1000 random trees were built using bootstrap sampling as the tree building algorithm. Herein, software used 63.2% of randomly selected samples as training set and remaining ones as validation cohort. Following this, misclassification error (OOB error) was calculated by confusion matrix. RF and area under the curve (AUC) were calculated using randomForest and ROCR using R package, respectively.^32–36

Results

Clinical characteristics of patients

This study consisted of 75 participants (20 normal, 20 SIL consisting of 6 LSIL and 14 HSIL, and 35 SCC samples). In our samples, we have combined both LSILs and HSILs together as SILs. The median age of normal, SIL, and SCC samples was 46, 53, and 58 years, respectively. All women were sexually active and non-smokers. Two in the normal group and one in the malignant group were nulliparous. Except for three women in the normal group, none had undergone a Pap test before. The HPV prevalence in normal, SIL, and tumor samples was found to be 20%, 65%, and 90%, respectively.

Cervical cancer genome is globally hypomethylated

Global DNA methylation at the cytosine residues was estimated in normal (n = 20), SCC (n = 35), three cervical cancer cell lines (SiHa, HeLa, and CaSki), and three LCLs by RP-HPLC. There was a significant reduction in global methylation content between normal and tumor samples (p < 0.0001; Figure 1(a)–(c)). The 5-mC content was significantly lower in cervical cancer cell lines (SiHa = 1.8%, HeLa = 1.9%, and CaSKi = 2.1%; Figure 1(d)). However, the 5-mC levels in LCLs were higher than the cervical cancer cell lines (3.84 vs 1.93; Figure 1(d)). The receiver operating characteristic (ROC) curve analysis showed a significant difference in 5-mC content between normal and tumor samples with an AUC = 0.946 and p < 0.0001 (Figure 1(e)). Taken together, our result shows that the cervical cancer genome is globally hypomethylated and can be useful for discriminating tumor from normal tissues.

Figure 1.

Cervical cancer genome shows reduced 5-mC and 5-hmC level. The 5-mC content was estimated by RP-HPLC in 20 normal, 35 SCC, 3 cervical cancer cell lines (SiHa, HeLa, and CaSki), and 3 LCLs. (a and b) The HPLC profile from normal and tumor samples. (c) The 5-mC levels in 20 normal and 35 SCC samples. The mean 5-mC content in normal and SCC samples was significantly different (normal = 3.5 vs SCC = 2.6) analyzed by t test (***p < 0.0001). (d) The 5-mC estimation in cervical cancer cell lines and LCLs. The 5-mC levels were also found to significantly lower in cervical cancer cell lines on comparison with LCLs (1.93 vs 3.84). (e) The diagnostic potential of 5-mC to discriminate SCC from normal samples was performed using ROC curve analysis. The AUC was 0.94 with p < 0.0001. (f) The 5-hmC estimation in 10 each of normal, SIL, and SCC, and 3 cervical cancer cell lines (SiHa, HeLa, and CaSki), respectively. The mean 5-hmC level in normal, SIL, and SCC samples were 0.40, 0.16, and 0.08, respectively. The 5-hmC levels were significantly lower in SIL and SCC samples when compared with normal samples as well as between SIL and SCC samples (p < 0.05). The mean fold difference between SIL versus normal, tumor versus normal, and SCC versus SIL was 2.5-, 5.0-, and 2.0-folds, respectively (p < 0.05). The 5-hmC level in cervical cancer cell lines, namely SiHa, HeLa, and CaSki was 0.017, 0.007, and 0.022, respectively. (g) The ROC analysis of 5-hmC in normal versus SIL, normal versus SCC, and SIL versus SCC, respectively. The AUC was 0.79, 0.92, and 0.63 in normal versus SIL, normal versus SCC, and SIL versus SCC, respectively.

Decreased 5-hmC levels in SIL and tumor sample

The 5-hmC levels were estimated in 10 each of normal, SIL, SCC, and cell lines. The mean 5-hmC levels in normal, SIL, and SCC samples were 0.40, 0.16, and 0.08, respectively (Figure 1(f)). The 5-hmC levels were significantly lower in SIL and SCC when compared to normal samples as well as between SIL and SCC samples (p < 0.05). The mean fold difference between various groups (SIL vs normal, SCC vs normal, and tumor vs SIL) was 2.5-, 5.0-, and 2.0-folds, respectively (p < 0.05). The 5-hmC levels in SiHa, HeLa, and CaSKi cervical cancer cell lines were 0.017, 0.007, and 0.022, respectively (Figure 1(f)). The AUC of 5-hmC between normal and SIL, normal and tumor, and SIL and tumor tissues were 0.79, 0.92, and 0.64, respectively (Figure 1(g)).

Methylation analysis in cervical cancer cell lines

We have analyzed the methylation status of 33 gene promoters in SiHa, CaSki, and HeLa cell lines and compared against the gene expression data from Gene Expression Omnibus (GEO) and GENT databases (Table 2). A total of 14 gene promoters were identified as methylated in at least one of the three cell lines tested (Table 2). Gene expression levels in normal, tumor, and cervical cancer cell lines were analyzed using the Affymetrix expression array data from NCBI (GSE9750 dataset) and GENT database (Table 2). Among the 33 genes tested, 21 of them showed significantly differential expression between normal and tumor tissues (p < 0.05; Table 2). Among the 15 methylated genes tested, 11 of them showed inverse relationship between methylation levels and gene expression.

Table 2.

DNA methylation and its correlation with gene expression.

Genes	Methylation			Expression			Normal	Tumor	p value
Genes	SiHa	HeLa	CaSki	SiHa	HeLa	CaSki	Normal	Tumor	p value
ARHGAP6	M	M	M	10.2	0.4	3.5	55.99	36.45	0.0243
ATM	U	U	U	240.8	237.7	304.5	322.67	432.16	0.0201
CDK6	U	U	U	10.2	26.3	111	21.32	30.43	0.0837
DAPK1	M	U	U	9.2	12.4	25.8	47.10	17.17	<0.0001
DDB1	U	ND	ND	581.3	1228	1232.9	737.31	697.38	0.3968
DLX1	ND	ND	U	463	570	NA	50.92	51.84
DPYD	U	U	U	167.5	56.7	93.7	192.05	182.00	0.9635
FGF18	U	U	U	3.5	1.6	6	17.50	14.79	0.4983
GBF1	U	U	U	234.1	203.4	268.2	164.10	172.84	0.5468
HAND2	M	M	M	3.2	2.6	1.5	18.60	8.60	0.006
HEY2	U	M	M	9.9	9.9	2	45.98	52.06	0.6052
HOXA7	U	ND	ND	79.7	16.8	45.6	44.33	67.78	0.0367
HOXC9	ND	ND	ND	378	272	NA	180.38	54.42
IKBKG	M	M	M	151.8	218.8	217.3	160.50	270.18	<0.0001
IREB2	U	U	U	12.7	17.3	11.8	38.67	23.42	0.0026
ITGAV	U	U	U	771.5	996.1	890.7	656.46	823.51	0.0813
ITPA	U	U	U	577.8	731.6	980	573.49	719.97	0.0007
LHX9	M	M	M	NA	NA	NA	NA	NA	NA
LRP5	ND	ND	ND	65.2	70.3	94.6	71.11	89.30	0.0722
LRPPRC	U	U	U	733	788.8	709.1	349.15	537.35	<0.0001
NKX2-2	M	M	M	11.1	0.6	0.7	8.74	11.90	0.3648
NME6	U	U	U	101	90.3	116.4	161.19	145.65	0.1889
NNAT	ND	ND	ND	47.4	39.7	41	102.53	59.37	0.0005
PCDH10	M	M	M	NA	NA	NA	NA	NA	NA
PCGF2	U	U	U	231	174.3	322	209.87	274.70	0.0724
PITX2	M	M	M	3.9	13.2	0.2	30.89	16.80	0.001
POLG	U	ND	ND	129.1	100.8	54.3	86.93	92.41	0.6852
PROX1	M	M	M	24.1	8.7	6.8	39.10	23.02	0.0017
RAB6C	M	M	M	912.5	799.8	1093.6	1220.30	885.50	0.0011
SLC5A6	U	U	U	312.9	216	372	180.81	267.46	0.0075
TBX3	M	M	M	100.6	8.5	54	335.11	99.46	<0.0001

M: methylated; U: unmethylated; ND: not determined; NA: not available.

Methylation data are from our study, gene expression is from public datasets.

p < 0.05 statistically significant.

DNA promoter methylation analysis in cervical samples

Initial screening of 33 gene promoters in normal, SIL, and SCC tissues identified 8 genes as differentially and significantly methylated. These were subsequently screened in another set of 10 samples belonging to each category. Among the 33 gene promoters tested, 8 of them (ARHGAP6, DAPK1 (death-associated protein kinase 1), HAND2 (heart and neural crest derivatives expressed 2), NKX2-2 (NK2 homeobox 2), PCDH10 (Protocadherin 10), PITX2 (paired-like homeodomain 2), PROX1, and RAB6C) showed significant difference in methylation either between normal and SCC or between SIL and SCC tissues (Figures 2 and 3; Table 3). The frequency of DNA methylation in all the 33 genes is shown in Supplementary Table S2. Furthermore, except DAPK1, remaining seven genes showed a significant potential to discriminate tumor from normal samples or tumor from SIL samples (Table 3). However, DAPK1 was found to discriminate only normal tissues from SCC. Furthermore, all the eight genes tested did not discriminate the normal from SIL samples. LHX9 (LIM homeobox 9), IKBKG, and TBX3 (T-box protein 3) were commonly methylated in normal, SIL, and SCC tissues, while ITPA, HOXC9, HEY2, and HOXA7 were methylated in very low percentage of SCC tissues. The gene promoters such as ATM (ataxia-telangiectasia mutated), BTG2, CDK6, DDB1, DPYD, FGF18 (fibroblast growth factor 18), GBF1, IREB2, LRP5, PCGF2, and POLG were unmethylated in all the categories of samples tested. When the individual CpG sites of eight significantly differentially methylated genes were analyzed, several of them showed significant hypermethylation among normal versus tumor or SIL versus SCC tissues (Supplementary Tables S2 and S3).

Figure 2.

The methylation analysis of eight differentially methylated gene promoters identified in our study groups. The methylation analysis was performed in 20 each of normal, SIL, and SCC samples by NGS-based bisulfite genomic sequencing. The box plot represents methylation level averaged over all CpG sites analyzed for each categories of samples. The line graph represents the methylation levels averaged over each CpG site of individual categories of samples. Asterisks (*) indicate statistical significance as calculated by ANOVA and Kruskal–Wallis test (p value ≤ 0.05 was considered as statistically significant). The significantly differentially methylated individual CpG sites are shown in Supplementary Table S3.

Figure 3.

Heat map of eight differentially methylated genes in 20 each of normal, SIL, and SCC samples, and 3 cervical cancer cell lines (SiHa, HeLa, and CaSki). Samples arranged as normal, SIL, SCC, and SiHa; HeLa and CaSki (horizontal); and CpG sites (vertical). The frequency of methylation is represented by colors with blue and red representing 0%–100% methylation. The heat map was drawn using R package (CRAN) of pheatmap tool.

Table 3.

The list of eight significantly differentially methylated genes and their diagnostic significance.

	ARHGAP6	DAPK1	HAND2	NKX2-2	PCDH10	PIXT2	PROX1	RAB6C
CpG sites Screened	23	45	27	18	26	19	40	40
Normal
U	12	20	18	17	20	16	19	20
M	8	0	2	3	0	4	1	0
SIL
U	4	18	15	13	8	13	10	13
M	16	2	5	7	12	7	10	7
Tumor
U	2	7	1	1	1	2	2	6
M	18	13	19	19	19	18	18	14
Normal vs SIL
Sensitivity	80	10	25	35	60	35	50	35
Specificity	60	100	90	85	100	80	95	100
p value	0.0225	0.4872	0.4075	0.2733	<0.0001	0.4801	0.0033	0.0083
AUC	0.73	0.62	0.65	0.59	0.6	0.51	0.67	0.71
Normal vs SCC
Sensitivity	90	65	95	95	95	90	90	70
Specificity	60	100	90	85	100	80	95	100
p value	0.0022	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001
AUC	0.88	0.80	0.83	0.90	0.95	0.87	0.96	0.89
SIL vs SCC
Sensitivity	90	65	95	95	95	90	90	35
Specificity	20	90	75	65	40	65	50	100
p value	0.6614	0.0008	<0.0001	0.0001	0.0197	0.0008	0.0138	0.0083
AUC	0.74	0.6	0.82	0.79	0.74	0.92	0.86	0.64

U: unmethylated; M: methylated; SIL: squamous intraepithelial lesion; AUC: area under the curve; SCC: squamous cell carcinoma.

DNA promoter methylation analysis of public datasets

We have analyzed the methylation status of 33 genes in four public datasets. In Dataset 1 (E-GEOD-30760), promoter CpG sites belonging to six (ARHGAP6, DAPK1, HAND2, NKX2-2, NNAT (Neuronatin), and PITX2) and three (SLC5A6, ITPA, and ITGAV) genes were hypermethylated and hypomethylated, respectively, in tumor samples (153 normal vs 62 tumors). In Dataset 2 (E-GEOD-30759), consisting of 15 normal and 48 tumor samples, six gene promoters such as ARHGAP6, DAPK1, HAND2, NKX2-2, NNAT, and PITX2 were significantly hypermethylated in tumor tissues. However, DLX1, HOXA7, HOXC9, LHX9, PROX1, and TBX3 gene promoters were also hypermethylated in less than 50% of the samples analyzed. In Dataset 3 (E-GEOD-41384), seven gene promoters (DAPK1, DLX1, HOXA7, LRPPRC, NNAT, PITX2, and SLC5A6) were hypermethylated in LSIL samples while eight gene promoters (ARHGAP6, DAPK1, HAND2, HOXA7, NKX2-2, NNAT, PITX2, and PROX1) were hypermethylated in HSIL when compared with normal tissues. The promoters of ARHGAP6, DAPK1, DLX1, HAND2, HOXA7, FGF18, LRP5, NKX2-2, NNAT, PITX2, and TBX3 were significantly hypermethylated in tumors when compared to the normal tissues. In Dataset 4 (E-GEOD-46306), which consisted of 20 normal, 18 HSIL, and 6 tumor samples, DAPK1, NNAT, and SLC5A6 gene promoters were hypermethylated in HSILs, while ARHGAP6, DAPK1, FGF18, HAND2, NNAT, and PITX2 were hypermethylated in tumors when compared with normal tissues. Although there was a difference in the extent of methylation in each study for the same gene, the pattern of methylation was consistent when compared to the normal tissues (Supplementary Table S4).

Sensitivity, specificity, and diagnostic utility of methylation markers

In order to identify the diagnostic significance and robustness of the markers, sensitivity and specificity analysis was performed. First, we investigated the diagnostic significance of methylation markers in our samples followed by independent validation in four public datasets. In our samples and among the eight significantly differentially methylated genes, the sensitivity of PROX1, PCDH10, and ARHGAP6 was 50%, 60%, and 80%, respectively, while the specificity was 95%, 100%, and 100%, respectively, for distinguishing SIL from normal tissues. For normal versus tumor samples, ARHGAP6, DAPK1, HAND2, NKX2-2, PCDH10, PITX2, PROX1, and RAB6C showed sensitivity of 65%–95% and specificity of 60%–100% (p < 0.05). For SIL versus tumor tissues, five genes (DAPK1, PITX2, PROX1, HAND2, and NKX2-2) showed sensitivity of 65%, 90%, 90%, 95%, and 95% and specificity of 90%, 65%, 50%, 75%, and 65%, respectively (p < 0.05; Figure 4; Table 3). The diagnostic potential of genes identified as aberrantly methylated in public datasets is summarized in Supplementary Tables and Figures.

Figure 4.

(a) Circular representation of the eight significantly differentially methylated genes. In total, 33 genes were profiled in 20 each of normal, SIL, SCC, and 3 cervical cancer cell lines (SiHa, HeLa, and CaSki) by NGS-based bisulfite genomic sequencing. The circle represents the eight significantly differentially methylated genes. Blue, yellow, and red represent the normal, SIL, and SCC samples. The percentage indicates the extent of methylation. (b–d) The ROC curve analysis of average methylation of the eight genes. The AUC was 0.65, 0.93, and 0.87 for normal versus SIL, normal versus SCC, and SIL versus SCC, respectively (p value < 0.05 was considered as statistically significant).

Discussion

The aim of this study was to investigate the diagnostic relevance of (1) global cytosine methylation, (2) global 5-hmC, and (3) promoter DNA methylation profile of a panel of genes in SIL and SCC. We have also compared our results with the published large datasets from other geographical regions and ethnicity. This is probably the first report to show the clinical utility of 5-mC, 5-hmC, and DNA methylation in a panel of genes in a single study. By ROC curve and sensitivity and specificity analyses, we have identified a panel of genes with potential for clinical management of cervical cancer. Cytosine methylation is an important and widely studied epigenetic modification that is frequently altered in diverse cancers.^37–39 Global hypomethylation and gene-specific hypermethylation are the most common and frequent epigenetic modifications known to regulate diverse aspects of tumorigenesis.^37,38 Although methylation in mammals is mostly seen at 5′ of cytosine at CpG dinucleotide context, there are also reports of less frequent non-CpG cytosine methylation.⁴⁰ During carcinogenesis, both global loss of DNA methylation and gene-specific hypermethylation of anticancer genes have been reported in different types of cancer, including cervical cancer.^2,37–39

Global loss of DNA methylation is primarily seen in repetitive elements, comprising approximately half of the human genome. Besides, progressive loss of DNA methylation may lead to activation of proto-oncogenes and transposable elements resulting in genomic instability and thus contributing to carcinogenesis.⁴¹ Similar to other published studies, we have identified significant loss of 5-mC content in cervical cancer tissues and cell lines compared to normal tissues.^42,43 Discovery of 5-hmC was considered as “sixth base” and is formed as an oxidative product of 5-mC by a highly regulated enzymatic process controlled by TET family (TET1, TET2, and TET3) of proteins.⁴ The distribution of 5-hmC is reported in many different tissues and cell types, and is abundant in embryonic stem cells and adult neuronal cells.⁴⁴ Although the biological function of 5-hmC is yet to be completely understood, many studies have reported the global loss of 5-hmC in cancer^45–47 and this may be related to reduced 5-mc levels. More recently, existence of 5-hmC inside mitochondria has also been reported.⁴⁸ The ROC curve analysis of 5-hmC levels was significant in distinguishing all the pertinent tissue categories tested. A recent study showed that global reduction of 5-hmC in cervical carcinoma is correlated with International Federation of Gynecology and Obstetrics (FIGO) stage, histologic grade, metastatic lymph node, and vascular invasion.⁴⁹ Furthermore, the reduced 5-hmC levels were associated with significant decrease in overall survival rate of the patients.⁴⁹ Taken together, our study along with other reports suggest that the global loss of 5-hmC during cervical carcinogenesis has clinical relevance. Further study needs to be performed in a large number of samples to elaborate these conclusions.

In our study, we have used an unbiased approach to select the candidate gene promoters to discover their diagnostic significance in cervical cancer. This was accomplished first by identification and validation of panel of gene promoters followed by confirmation of their significance from independent studies consisting of five large panels of datasets. Combined analysis of our study and public datasets identified ARHGAP6, DAPK1, HAND2, PROX1, PCDH10, NKX2-2, NNAT, RAB6C, and PITX2 gene promoters with more than 80% sensitivity and specificity to discriminate between normal, premalignant, or tumor tissues. Moreover, some of the gene promoters (DAPK1, HEY2, LRPPRC, PCDH10, PROX1, RAB6C, and RBBP8) showed more than 80% sensitivity to discriminate normal, premalignant, and malignant tissues but these were not considered due to their lower specificity.

Among the 33 gene promoters screened, PROX1, PCDH10, and RAB6C showed very high sensitivity (80%–100%) and specificity (70%–100%) in distinguishing normal versus premalignant tissues: PROX1, PCDH10, DAPK1, HAND2, PITX2, and NKX2-2 (sensitivity: 80%–100% and specificity: 80%–100%; p < 0.05) for normal versus tumor tissues; and DAPK1, HAND2, and NKX2-2 (sensitivity: 70%–90% and specificity: 80%–90%; p < 0.05) for premalignant versus tumor tissues, respectively. The results of this study are promising and encouraging to deploy in the clinical practice for cervical cancer diagnosis. When these gene promoters were mapped using publicly available independent studies, we were able to identify a consistent pattern of methylation for selected CpG sites. In addition, comparison of these gene promoters in public datasets allowed us to identify additional genes with high sensitivity and specificity. Gene promoters such as ARHGAP6, HAND2, DAPK1, NNAT, NKX2-2, and PITX2 exhibited remarkably high sensitivity and specificity of >90% in distinguishing one among the three categories. Although several gene promoters validated in our dataset were significantly hypermethylated when compared to normal tissues, they were inconsistent in public datasets due to low methylation status and lower sensitivity and specificity. For example, while PROX1 showed very high sensitivity and specificity in our samples, these were absent in public datasets. However, NNAT promoter showed very high sensitivity and specificity in the public datasets and was less sensitive in our samples. This may be attributed to different methods employed, region of the gene used for analysis, and ethnic variations among the population tested. Based on our results, we propose that the hypermethylation of ARHGAP6, DAPK1, HAND2, NKX2-2, NNAT, PROX1, PCDH10, PITX2, and RAB6C makes them potential targets, as they were significantly methylated in more than two datasets.

Several studies have used high throughput microarray or NGS technologies to identify the differentially methylated signatures of clinical relevance.^50–52 Despite the differences in the method used for discovery, ethnic differences among the populations, and other confounding factors, the studies have still identified common methylation patterns among the identified genes.^53,54 Our study has identified co-methylation and co-expression of multiple genes affecting the same pathway. Moreover, the gene ontology and pathway analysis identified functionally and biologically relevant processes and pathways in cancer.

Among the 33 gene promoters screened, only 3 of them (NNAT, PCDH10, and DAPK1) were previously reported as hypermethylated in cervical cancer, while rest of the 30 gene promoters were found to be novel targets for cervical cancer.^55,56 One of the limitations of this study is the moderate clinical sample size with limited samples in the premalignant group. In India, unlike in Western countries, routine screening is not popular, and therefore, few women are diagnosed at the stage of premalignant stage. Moreover, samples which did not contain more than 80% of diseased cells were excluded from the study. Some of the genes identified in our analysis were novel and showed high sensitivity and specificity to cervical cancer progression. However, their mechanism of action and role in cervical cancer progression need to be explored.

Rho GTPase–activating protein 6 (ARHGAP6), a member of rhoGAP family is involved in actin polymerization and cell motility. ARHGAP6 is aberrantly expressed in colorectal cancer and might serve as a biomarker for colorectal cancer.⁵⁷ ATM protein belongs to PI3/PI4–kinase family and is an important cell cycle–checkpoint kinase reported as hypermethylated in human colorectal cancer cell lines, oral SCCs, breast cancers, and brain tumors.^58–61 In addition to this, methylation of ATM promoter is correlated with radiosensitivity in colorectal cancers.⁵⁹ FGF18 is a fibroblast growth factor family member possessing cell survival and mitogenic activities and is involved in cell growth, morphogenesis, tumor growth, and invasion by regulation of actin cytoskeleton, PI3K, and fibroblast growth factor receptor (FGFR) signaling. FGF18 is downregulated in response to demethylating agent treatment in bladder cancer.⁶² HAND2 encodes transcription factors belonging to basic helix-loop-helix (bHLH) family and is downregulated in endometrial cancer by promoter hypermethylation.⁶³ It is essential for cardiac morphogenesis, development, and regulation of angiogenesis and is a part of nuclear factor of activated T cells (NFAT) and cardiac hypertrophy signaling pathway. Hypermethylation of HAND2 is also reported in lung cancer.⁶⁴ Hes-related family bHLH transcription factor with YRPW motif 2 (HEY2), a transcription factor belonging to the bHLH-type family, acts as a downstream effector of Notch signaling.⁶⁵ The downregulation of HOXA7, belonging to the homeobox group of transcription factors and essential for embryonic development, is reported to be associated with leukemia. The hypermethylation of HOXA7 is reported in lung carcinoma and is associated with aggressive behavior in meningiomas.^39,66

LHX9, a developmentally expressed transcription factor, participates in the development of bi-potential gonads in mice and humans.⁶⁷ The aberrant methylation and silencing of LHX9 is reported in childhood malignant gliomas.⁶⁸ NKX2-2 is a homeobox-containing gene involved in the morphogenesis of central nervous system (CNS) and is required for maintenance of NEUROD1 expression in pancreas.⁶⁹ PCDH10, belonging to the protocadherin family, is a potential calcium-dependent cell adhesion molecule associated with Dravet syndrome. It inhibits cancer cell migration and is hypermethylated in cervical and gastric cancers.^55,70 Hypermethylation and reduced expression of PCDH10 show poor survival in gastric cancer, metastasis in colorectal cancer, and overall poor survival in bladder cancer.⁷¹ PITX2, belonging to the RIEG/PITX homeobox family, is involved in control of cell proliferation and morphogenesis.^72,73 PITX2 is associated with Wnt and transforming growth factor (TGF)-beta signaling and is hypermethylated in breast, prostate, and may other cancers. Hypermethylation of PITX2 is associated with poor survival in breast cancer, and it is a prognostic marker of biochemical recurrence in prostate cancer.^74,75 Propero homeobox 1 (PROX1) is a homeobox transcription factor and is associated with breast angiosarcoma and kaposiform hemangioendothelioma.⁷⁶ PROX1 is a tumor suppressor gene often hypermethylated in hematological malignancies, sporadic breast cancer, and oral cancer.^77,78 PROX1 expression in oral cancer has shown to reduce cell proliferation.⁷⁹

TBX3 is involved in the regulation of developmental process and is associated with ulnar–mammary syndrome and holt–oram syndrome.⁸⁰ It is hypermethylated in bladder cancer and acts as pTa-specific prognostic marker.⁸¹ TBX3 overexpression is also reported in a number of cancers. In melanoma, TBX3 overexpression is associated with promotion and invasion of melanoma, while in breast cancer, it is shown to interact with histone deacetylases (HDACs).^82,83 NNAT, an imprinted gene, is involved in the regulation of ion channels during brain development, and hypermethylation of NNAT locus occurs frequently in pediatric acute leukemia.⁸⁴ RAB6C, a member of RAS oncogene family, regulates centrosome duplication and cell cycle progression by guanosine triphosphate (GTP) binding and GTPase activity. RAB6C is hypermethylated and downregulated in drug resistant breast cancer cell lines and is known to inhibit p53 regulated expression.⁸⁵ DAPK1 is a calcium/calmodulin-dependent serine/threonine kinase which inhibits cell survival and triggers apoptosis and autophagy. Hypermethylation leading to transcriptional silencing of DAPK1 and reactivation by demethylating agents is reported in many cancers.^56,86,87

Conclusion

In conclusion, we have shown that during cervical carcinogenesis, the genome undergoes significant reduction in 5-mC and 5-hmC content. We have also shown that estimation of 5-mC or 5-hmC or profiling for promoter methylation of a panel of genes (ARHGAP6, DAPK1, HAND2, NKX2-2, NNAT, PROX1, PCDH10, PITX2, and RAB6C) may act as a potential tool for early detection and screening of cervical cancer. Furthermore, these methylation analyses showed high sensitivity and specificity, which are equivalent or higher than cytology-based screening. However, a uniform assay and population-wide screening are required before drawing further conclusions. DNA methylation–based biomarkers combined with cytology screening might help to improve the sensitivity and specificity of current screening methodology in cervical cancer. Further studies are required to translate these DNA methylation markers for early detection and prognosis of cervical neoplasia for better clinical management.

Footnotes

Acknowledgements

The authors thank TIFAC-CORE for pharmacogenomics study and DST-FIST, Government of India, for infrastructure support; Manipal University and CSIR for fellowship under structured PhD program; and the Government of India for senior research fellowship. The authors also thank all the participants for their kind cooperation in this study. Samatha Bhat and Shama Prasada Kabekkodu contributed equally to this work

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Sawaya

Grimes

DA.

New technologies in cervical cytology screening: a word of caution. Obstet Gynecol 1999; 94: 307–310.

Bhat

Kabekkodu

Noronha

. Biological implications and therapeutic significance of DNA methylation regulated genes in cervical cancer. Biochimie 2016; 121: 298–311.

Mersakova

Nachajova

Szepe

. DNA methylation and detection of cervical cancer and precancerous lesions using molecular methods. Tumour Biol 2015; 37: 23–27.

Tahiliani

Koh

Shen

. Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science 2009; 324: 930–935.

Song

Potential functional roles of DNA demethylation intermediates. Trends Biochem Sci 2013; 38: 480–484.

Kellinger

Song

Chong

. 5-formylcytosine and 5-carboxylcytosine reduce the rate and substrate specificity of RNA polymerase ii transcription. Nat Struct Mol Biol 2015; 19: 831–833.

Yang

HJ.

Aberrant DNA methylation in cervical carcinogenesis. Chin J Cancer 2012; 32: 42–48.

Amornpisutt

Proungvitaya

Jearanaikoon

. DNA methylation level of OPCML and SFRP1: a potential diagnostic biomarker of cholangiocarcinoma. Tumour Biol 2015; 36: 4973–4978.

Dhir

Guzzetta

. DNA methylation biomarker candidates for early detection of colon cancer. Tumour Biol 2012; 33: 363–372.

10.

Esteller

Garcia-Foncillas

Andion

. Inactivation of the DNA-repair gene MGMT and the clinical response of gliomas to alkylating agents. N Engl J Med 2000; 343: 1350–1354.

11.

Singal

van Wert

Bashambu

Cytosine methylation represses glutathione S-transferase P1 (GSTP1) gene expression in human prostate cancer cells. Cancer Res 2001; 61: 4820–4826.

12.

Kishida

Natsume

Toda

. Correlation between quantified promoter methylation and enzymatic activity of O6-methylguanine-DNA methyltransferase in glioblastomas. Tumour Biol 2012; 33: 373–381.

13.

Harden

Sanderson

Goodman

. Quantitative GSTP1 methylation and the detection of prostate adenocarcinoma in sextant biopsies. J Natl Cancer Inst 2003; 95: 1634–1637.

14.

Duenas-Gonzalez

Lizano

Candelaria

. Epigenetics of cervical cancer. An overview and therapeutic perspectives. Mol Cancer 2005; 4: 38.

15.

Saavedra

Brebi

Roa

JC.

Epigenetic alterations in preneoplastic and neoplastic lesions of the cervix. Clin Epigenetics 2012; 4: 13.

16.

Eijsink

Lendvai

Deregowski

. A four-gene methylation marker panel as triage test in high-risk human papillomavirus positive patients. Int J Cancer 2011; 130: 1861–1869.

17.

Lin

Chen

Huang

. Infrequent mutations and no methylation of CDKN2A (P16/MTS1) and CDKN2B (p15/MTS2) in hepatocellular carcinoma in Taiwan. Eur J Cancer 1998; 34: 1789–1795.

18.

Matsuda

Ichida

Matsuzawa

. p16(INK4) is inactivated by extensive CpG methylation in human hepatocellular carcinoma. Gastroenterology 1999; 116: 394–400.

19.

Hui-Yuen

McAllister

Koganti

. Establishment of Epstein-Barr virus growth-transformed lymphoblastoid cell lines. J Vis Exp 2011; 8: 3321.

20.

Solomon

Davey

Kurman

. The 2001 Bethesda system: terminology for reporting results of cervical cytology. JAMA 2002; 287: 2114–2119.

21.

Bakshi

Ekram

Kim

Locus-specific DNA methylation analysis of retrotransposons in ES, somatic and cancer cells using high-throughput targeted repeat element bisulfite sequencing. Genom Data 2015; 3: 87–89.

22.

Ekram

Kim

High-throughput targeted repeat element bisulfite sequencing (HT-TREBS): genome-wide DNA methylation analysis of IAP LTR retrotransposon. PLoS ONE 2014; 9: e101683.

23.

Sedlazeck

Rescheneder

von Haeseler

Nextgenmap: fast and accurate read mapping in highly polymorphic genomes. Bioinformatics 2013; 29: 2790–2791.

24.

Lutsik

Feuerbach

Arand

. BiQ analyzer HT: locus-specific analysis of DNA methylation by high-throughput bisulfite sequencing. Nucleic Acids Res 2011; 39: W551–W556.

25.

Gravitt

Peyton

Alessi

. Improved amplification of genital human papillomaviruses. J Clin Microbiol 2000; 38: 357–361.

26.

Kabekkodu

Bhat

Pandey

. Prevalence of human papillomavirus types and phylogenetic analysis of HPV-16 L1 variants from southern India. Asian Pac J Cancer Prev 2015; 16: 2073–2080.

27.

Armstrong

Bermingham

Bassett

. Global DNA methylation measurement by HPLC using low amounts of DNA. Biotechnol J 2010; 6: 113–117.

28.

Cezar

Bartolomei

Forsberg

. Genome-wide epigenetic alterations in cloned bovine fetuses. Biol Reprod 2003; 68: 1009–1014.

29.

Rotti

Mallya

Kabekkodu

. DNA methylation analysis of phenotype specific stratified Indian population. J Transl Med 2015; 13: 151.

30.

Guo

Yan

. Identification and validation of the methylation biomarkers of non-small cell lung cancer (NSCLC). Clin Epigenetics 2015; 7: 3.

31.

Krzywinski

Schein

Birol

. Circos: an information aesthetic for comparative genomics. Genome Res 2009; 19: 1639–1645.

32.

Liaw

Wiener

Classification and regression by randomforest. R News 2002; 2: 18–22.

33.

Marsit

Houseman

Christensen

. Identification of methylated genes associated with aggressive bladder cancer. PLoS ONE 2010; 5: e12334.

34.

Sing

Sander

Beerenwinkel

. ROCR: visualizing classifier performance in R. Bioinformatics 2005; 21: 3940–3941.

35.

Tsou

Galler

Siegmund

. Identification of a panel of sensitive and specific DNA methylation markers for lung adenocarcinoma. Mol Cancer 2007; 6: 70.

36.

Vasiljevic

Brentnall

. Absolute quantitation of DNA methylation of 28 candidate genes in prostate cancer using pyrosequencing. Dis Markers 2011; 30: 151–161.

37.

Ning

Zhao

. Targeting epigenetic regulations in cancer. Acta Biochim Biophys Sin 2015; 48: 97–109.

38.

Paska

Hudler

Aberrant methylation patterns in cancer: a clinical view. Biochem Med 2015; 25: 161–176.

39.

Di Vinci

Brigati

Casciano

. HOXA7, 9, and 10 are methylation targets associated with aggressive behavior in meningiomas. Transl Res 2012; 160: 355–362.

40.

Pinney

SE.

Mammalian non-CpG methylation: stem cells and beyond. Biology 2014; 3: 739–751.

41.

Ehrlich

DNA hypomethylation in cancer cells. Epigenomics 2009; 1: 239–259.

42.

Kim

Giuliano

Hatch

. Global DNA hypomethylation increases progressively in cervical dysplasia and carcinoma. Cancer 1994; 74: 893–899.

43.

Missaoui

Hmissa

Dante

. Global DNA methylation in precancerous and cancerous lesions of the uterine cervix. Asian Pac J Cancer Prev 2010; 11: 1741–1744.

44.

Shen

Zhang

5-hydroxymethylcytosine: generation, fate, and genomic distribution. Curr Opin Cell Biol 2013; 25: 289–296.

45.

Chen

Zhang

Guo

. Loss of 5-hydroxymethylcytosine is linked to gene body hypermethylation in kidney cancer. Cell Res 2016; 26: 103–118.

46.

Ficz

Gribben

JG.

Loss of 5-hydroxymethylcytosine in cancer: cause or consequence?

Genomics 2014; 104: 352–357.

47.

Lian

Ceol

. Loss of 5-hydroxymethylcytosine is an epigenetic hallmark of melanoma. Cell 2012; 150: 1135–1146.

48.

Ghosh

Sengupta

Scaria

Hydroxymethyl cytosine marks in the human mitochondrial genome are dynamic in nature. Mitochondrion 2016; 27: 25–31.

49.

Zhang

Han

. 5-hydroxymethylcytosine expression is associated with poor survival in cervical squamous cell carcinoma. Jpn J Clin Oncol 2016; 46(5): 427–434.

50.

Kron

Pethe

Briollais

. Discovery of novel hypermethylated genes in prostate cancer using genomic CpG island microarrays. PLoS ONE 2009; 4: e4830.

51.

Meldrum

Doyle

Tothill

RW.

Next-generation sequencing for cancer diagnostics: a practical perspective. Clin Biochem Rev 2011; 32: 177–195.

52.

Taylor

Kramer

Davis

. Ultradeep bisulfite sequencing analysis of DNA methylation patterns in multiple gene promoters by 454 sequencing. Cancer Res 2007; 67: 8511–8518.

53.

Ahmad

Arjumand

Seth

. Methylation of the APAF-1 and DAPK-1 promoter region correlates with progression of renal cell carcinoma in north Indian population. Tumour Biol 33: 395–402.

54.

Claus

Hackanson

Poetsch

. Quantitative analyses of DAPK1 methylation in AML and MDS. Int J Cancer 2012; 131: E138–E142.

55.

Narayan

Scotto

Neelakantan

. Protocadherin PCDH10, involved in tumor progression, is a frequent and early target of promoter hypermethylation in cervical cancer. Genes Chromosomes Cancer 2009; 48: 983–992.

56.

Banzai

Nishino

Quan

. Promoter methylation of DAPK1, FHIT, MGMT, and CDKN2A genes in cervical carcinoma. Int J Clin Oncol 2013; 19: 127–132.

57.

Jovov

Araujo-Perez

Sigel

. Differential gene expression between African American and European American colorectal cancer patients. PLoS ONE 2012; 7: e30168.

58.

Brennan

Garcia-Closas

Orr

. Intragenic ATM methylation in peripheral blood DNA as a biomarker of breast cancer risk. Cancer Res 2012; 72: 2304–2313.

59.

Kim

Shrivastav

. Aberrant methylation of the ATM promoter correlates with increased radiosensitivity in a human colorectal tumor cell line. Oncogene 2002; 21: 3864–3871.

60.

Mehdipour

Karami

Javan

. Linking ATM promoter methylation to cell cycle protein expression in brain tumor patients: cellular molecular triangle correlation in ATM territory. Mol Neurobiol 2015; 52: 293–302.

61.

Rigi-Ladiz

Kordi-Tamandani

Torkamanzehi

Analysis of hypermethylation and expression profiles of APC and ATM genes in patients with oral squamous cell carcinoma. Clin Epigenetics 2011; 3: 6.

62.

Veerla

Panagopoulos

Jin

. Promoter analysis of epigenetically controlled genes in bladder cancer. Genes Chromosomes Cancer 2008; 47: 368–378.

63.

Jones

Teschendorff

. Role of DNA methylation and epigenetic silencing of HAND2 in endometrial cancer development. PLoS Med 2013; 10: e1001551.

64.

Pradhan

Desai

Palakal

MJ.

Systems biology approach to stage-wise characterization of epigenetic genes in lung adenocarcinoma. BMC Syst Biol 2013; 7: 141.

65.

Fischer

Schumacher

Maier

. The notch target genes Hey1 and Hey2 are required for embryonic vascular development. Genes Dev 2004; 18: 901–911.

66.

Rauch

Wang

Zhang

. Homeobox gene methylation in lung cancer studied by genome-wide analysis with a microarray-based methylated CpG island recovery assay. Proc Natl Acad Sci U S A 2007; 104: 5527–5532.

67.

Eggers

Sinclair

Mammalian sex determination-insights from humans and mice. Chromosome Res 2012; 20: 215–238.

68.

Vladimirova

Mikeska

Waha

. Aberrant methylation and reduced expression of LHX9 in malignant gliomas of childhood. Neoplasia 2009; 11: 700–711.

69.

Anderson

Torres

Solomon

. Cooperative transcriptional regulation of the essential pancreatic islet gene NeuroD1 (beta2) by Nkx2.2 and neurogenin 3. J Biol Chem 2009; 284: 31236–31248.

70.

Chim

Yang

. Role of PCDH10 and its hypermethylation in human gastric cancer. Biochim Biophys Acta 2011; 1823: 298–305.

71.

. Downregulation of protocadherin-10 expression correlates with malignant behaviour and poor prognosis in human bladder cancer. J Int Med Res 2013; 41: 38–47.

72.

Lozano-Velasco

Vallejo

Esteban

. A Pitx2-MicroRNA pathway modulates cell proliferation in myoblasts and skeletal-muscle satellite cells and promotes their commitment to a myogenic cell fate. Mol Cell Biol 2015; 35: 2892–2909.

73.

Rodriguez-Leon

Rodriguez Esteban

Marti

. Pitx2 regulates gonad morphogenesis. Proc Natl Acad Sci U S A 2008; 105: 11242–11247.

74.

Strand

Orntoft

Sorensen

KD.

Prognostic DNA methylation markers for prostate cancer. Int J Mol Sci 2014; 15: 16544–16576.

75.

Weiss

Cottrell

Distler

. DNA methylation of the PITX2 gene promoter region is a strong independent prognostic marker of biochemical recurrence in patients with prostate cancer after radical prostatectomy. J Urol 2009; 181: 1678–1685.

76.

Miettinen

Wang

ZF.

Prox1 transcription factor as a marker for vascular tumors-evaluation of 314 vascular endothelial and 1086 nonvascular tumors. Am J Surg Pathol 2011; 36: 351–359.

77.

Nagai

Hatano

. Mutations and aberrant DNA methylation of the PROX1 gene in hematologic malignancies. Genes Chromosomes Cancer 2003; 38: 13–21.

78.

Versmold

Felsberg

Mikeska

. Epigenetic silencing of the candidate tumor suppressor gene PROX1 in sporadic breast cancer. Int J Cancer 2007; 121: 547–554.

79.

Rodrigues

de Oliveira Rodini

de Aquino Xavier

. PROX1 gene is differentially expressed in oral cancer and reduces cellular proliferation. Medicine 2014; 93: e192.

80.

Bamshad

Lin

Law

. Mutations in human TBX3 alter limb, apocrine and genital development in ulnar-mammary syndrome. Nat Genet 1997; 16: 311–315.

81.

Beukers

Kandimalla

Masius

. Stratification based on methylation of TBX2 and TBX3 into three molecular grades predicts progression in patients with pTa-bladder cancer. Mod Pathol 2015; 28: 515–522.

82.

Peres

Prince

The T-box transcription factor, TBX3, is sufficient to promote melanoma formation and invasion. Mol Cancer 2013; 12: 117.

83.

Yarosh

Barrientos

Esmailpour

. TBX3 is overexpressed in breast cancer and represses p14 ARF by interacting with histone deacetylases. Cancer Res 2008; 68: 693–699.

84.

Kuerbitz

Pahys

Wilson

. Hypermethylation of the imprinted NNAT locus occurs frequently in pediatric acute leukemia. Carcinogenesis 2002; 23: 559–564.

85.

Tian

Wang

Huang

. Methylation of WTH3, a possible drug resistant gene, inhibits p53 regulated expression. BMC Cancer 2008; 8: 327.

86.

Celik

Akcora

Ozkan

. Methylation analysis of the DAPK1 gene in imatinib-resistant chronic myeloid leukemia patients. Oncol Lett 2015; 9: 399–404.

87.

Xiong

Huang

. Association between DAPK1 promoter methylation and cervical cancer: a meta-analysis. PLoS ONE 2014; 9: e107272.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.27 MB

0.53 MB

0.23 MB

0.86 MB

Aberrant gene-specific DNA methylation signature analysis in cervical cancer

Abstract

Keywords

Introduction

Materials and methods

Cell line maintenance and patient cohort

DNA extraction and bisulfite conversion

Methylation analysis by NGS

Analysis of NGS data

HPV DNA typing

Global methylation analysis by reversed-phase high-performance liquid chromatography

Global 5-hmC estimation

Public datasets

Bioinformatics and statistical analysis

Results

Clinical characteristics of patients

Cervical cancer genome is globally hypomethylated

Decreased 5-hmC levels in SIL and tumor sample

Methylation analysis in cervical cancer cell lines

DNA promoter methylation analysis in cervical samples

DNA promoter methylation analysis of public datasets

Sensitivity, specificity, and diagnostic utility of methylation markers

Discussion

Conclusion

Footnotes

Acknowledgements

Declaration of conflicting interests

Funding

References

Supplementary Material