Quantitative measurement of L1 HPV16 methylation for the prediction of pre-

invasive and invasive cervical disease

Christine Kottaridi1, Maria Kyrgiou2,3, Abraham Pouliakis1, Maria Magkana1, Evangelia

Aga1, Aris Spathis1, Anita Mitra2,3, George Makris4, Charalampos Chrelias4, Vassiliki

Mpakou5, Evangelos Paraskevaidis6, John G. Panayiotides7, Petros Karakitsos1

1Department of Cytopathology, National and Kapodistrian University of Athens,

Medical School, “ATTIKON” University Hospital, 1 Rimini, Haidari, 12462, Athens,

Greece.

2Department of Surgery and Cancer, IRDB, Faculty of Medicine, Imperial College,

London W12 0NN, UK

3West London Gynaecological Cancer Center, Queen Charlotte’s and Chelsea,

Hammersmith Hospital, Imperial Healthcare NHS Trust

43rd Department of Obstetrics and Gynecology, “ATTIKON” University Hospital, 1

Rimini, Haidari, 12462, Athens, Greece

5Second Dept. of Internal Medicine and Research Institute, University of Athens,

Medical School, “ATTIKON” University Hospital, 1 Rimini, Haidari, 12462, Athens,

Greece

6Department of Obstetrics and Gynecology, University Hospital of Ioannina, 45500

Ioannina, Greece

72nd Department of Pathology, University of Athens, Medical School, “ATTIKON”

University Hospital, 1 Rimini, Haidari, 12462, Athens, Greece

1

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

Running title: HPV16 L1 methylation as a biomarker

Word count (excluding references): 3497, Abstract word count: 197

Number of figures: 4, Number of tables: 2

Supplementary data: 2 tables, 5 figures

Declaration of competing interests: The authors report no conflict of interests. All

the authors declare: no support from any organization for the submitted work; no

financial relationships with any organizations that might have an interest in the

submitted work in the previous three years; no other relationships or activities that

could appear to have influenced the submitted work.

Ethical approval: National Research Ethics Service Committee London – Fulham

(Approval number 13/LO/0126). Bioethics committee of “ATTIKON” University

Hospital (Approval number 5/14-06-2013)

Role of the funding source: This work was supported by the British Society of

Colposcopy Cervical Pathology Jordan/Singer Award (P47773)(MK); the Imperial

College Healthcare Charity (P47907) (AM, MK); Genesis Research Trust (P55549)

(MK); and the Imperial Healthcare NHS Trust NIHR Biomedical Research Centre

(P45272) (MK). None of the funders have had any influence on the study design; in

the collection, analysis, and interpretation of data; in the writing of the report; and in

the decision to submit the article for publication. This study was partially funded by

the Greek Ministry of Development (General Secretariat for Research and

2

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

Technology-GSRT) and European Union. Project acronym: HPV-Guard (Cooperation

2011-2013, code: 11ΣΥΝ_10_250 http://HPVGuard.org).

Corresponding author:

Dr Maria Kyrgiou, MSc, PhD, MRCOG, 3rd Floor, Institute of Reproductive and

Developmental Biology, Department of Surgery and Cancer, Imperial College,

Hammersmith Campus, Du Cane Road, W12 0NN, London, UK. Tel: +44 (0)20 7594

2177; Email: m.kyrgiou@imperial.ac.uk

3

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

ABSTRACT

Background: Methylation of the HPV DNA has been proposed as a novel biomarker.

Here, we correlated the mean methylation level of 12 CpG sites within L1 gene, to

the histological grade of cervical precancer and cancer. We assessed whether HPV L1

gene methylation can predict the presence of high-grade disease at histology in

women testing positive for HPV 16 genotype.

Methods: Pyrosequencing was used for DNA methylation quantification and 145

women were recruited.

Results: We found that the L1 HPV16 mean methylation (+/-SD) significantly

increased with disease severity [CIN3=17.9%(±7.2) vs CIN2=11.6%(±6.5), p<0.001 or

vs CIN1 =9.0%(±3.5), p<0.001). Mean methylation was a good predictor of CIN3+

cases; the Area Under the Curve (AUC) was higher for sites 5611 in the prediction of

CIN2+ and higher for position 7145 for CIN3+. The evaluation of different

methylation thresholds for the prediction of CIN3+, showed that the optimal balance

of sensitivity and specificity (75.7% and 77.5%, respectively), PPV and NPV (74.7%

and 78.5%, respectively) was achieved for a methylation of 14.0% with overall

accuracy of 76.7%.

Conclusion: Elevated methylation level is associated with increased disease severity

and has good ability to discriminate HPV16 positive women that have high-grade

disease or worse.

Keywords: cervical intraepithelial neoplasia; CIN; HPV L1 gene methylation;

pyrosequencing; Human Papillomavirus

4

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

INTRODUCTION

The introduction of systematic call and recall screening programmes has resulted in a

profound decrease in the incidence and mortality from invasive cervical cancer as

pre-invasive lesions (cervical intra-epithelial neoplasia;CIN) can be treated

appropriately [1, 2]. Despite the high efficacy of screening in preventing cervical

cancer, this has traditionally relied on cytology that is known to have limitations.

Establishing that persistent infection with high-risk human papilloma-virus (HPV) is

causally associated with cervical cancer has led to major advances in primary and

secondary prevention. It is now recognized that the use of the HPV DNA test in

primary screening is likely to offer 60-70% greater protection against invasive cancer

compared with cytology-based screening [3], while the best policy in further triaging

women with positive results at HPV-based screening is unclear.

New biomarkers, exploring the viral genome and life cycle together with its

interactions with the host, have the potential to permit a more comprehensive

understanding of the disease process. These tests may enable a more accurate

detection of clinically significant pre-invasive lesions and a more personalised

identification and management of lesions with true progressive carcinogenic

potential [4-6]. Some target genes in the HPV genome play pivotal roles in the viral

life cycle and ongogenesis [7, 8]. Methylation of the viral DNA has been recently

proposed as a novel biomarker with encouraging results. The quantification of the

percentage of cytosines with a covalently added methyl-group at individual CpG

dinucleotides reflects the degree of epigenetic changes of the viral genome.

5

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

Research in the quantification of methylation at different CpG sites and genes

produced contradictory results, while the methylation threshold varied across

studies [9-11]. Although two small studies suggested that increased methylation was

associated with high-grade intra-epithelial lesions (HSIL) [9, 11], others examining

CpG sites within the upstream regulatory region (URR) documented that it is the lack

of methylation that associates with disease progression [12, 13]. Two further studies

on the Guanacaste Costa Rica cohort reported that high methylation at L1, L2, E2-4

CpG sites increased the risk of CIN3 by 50 times when compared to low methylation

(Odds Ratio(OR)=52, 95%CI:4.0–670). Increased methylation levels predicted both

the presence of HSIL in women with positive HPV16 infections as well as the risk for

development of future CIN2+ [10, 14].

This prospective study aims to assess whether pyrosequencing of the HPV16 L1 gene

and quantification of the methylation level of specific CpGs could predict the

presence of high-grade disease at histology in women testing positive for HPV16

genotype. More specifically, we aim to correlate the mean methylation level to the

histological grade and to determine its accuracy in predicting the disease severity by

establishing optimum methylation cut-offs.

METHODS

Study population

We conducted a multicentric prospective study that recruited in three University

Hospitals (Ioannina/Athens, Greece and Imperial College London, UK). Ethical

6

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

approval was obtained and all patients gave consent (Bioethics Committee, Greece

and National Research Ethics Service Committee London–Fulham (13/LO/0126).

We included non-pregnant women, 21–67 years of age who attended the

gynaecology clinics (May 2013 to 2015). All cervical samples that tested positive for

HPV16 DNA typing and had available histology (punch biopsy or cone - gold

standard) were included. If histology was available from both biopsies and cones, the

most severe lesion was used. The histological diagnoses included the following

groups: normal, CIN1, CIN2, CIN3, squamous cell(SCC) and adenocarcinoma (Adeno-

ca).

Women were included irrespective of their ethnicity and smoking habits.. Women

who were HIV or hepatitis B/C positive, with autoimmune disorders, or had a

previous history of cone were excluded..

Sample collection and methods

We prospectively collected patient characteristics and recorded the cytological,

colposcopic and histological findings. We obtained a liquid-based cytology sample

(LBC, ThinPrep® Pap Test, Hologic) that was prepared on a TP2000 Processor and was

reported by trained cytopathologists according to the Bethesda 2001 system

(TBS2001) [15]. The residual LBC sample was aliqouted and stored at 4°C until

further use.

7

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

The HPV DNA typing was performed with CLART® HPV2 kit (Genomica, Spain) which

is validated to genotype 20 HR HPV types (16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53,

56, 58, 59, 66, 68, 70, 73, 82 and 85) and 15 LR HPV types (6, 11, 40, 42, 43, 44, 54,

61, 62, 71, 72, 81, 83, 84 and 89). The extracted DNA concentration’s were measured

with Quant-iT™ PicoGreen® dsDNA Assay Kit (Thermo Fisher Scientific Inc., Waltham,

MA, USA).

DNAs were then bisulphite converted using the EpiTect Bisulfite Kit (Qiagen, Hilden,

Germany), according to the manufacturer’s instructions and stored at -80°C. Biotin-

labelled primer sets and PCR conditions [14] were used as previously described to

amplify HPV16 L1 region. DNA from the cervical cancer cell line SiHa that contains

integrated HPV16 genome [16] was extracted and bisulfite converted. The

methylation quantification was performed by Pyrosequencing technology (PyroMark

Q24 Qiagen, Hilden, Germany), which provides a site-specific quantification of

methylation at individual CpG sites. The analysis performed in the present study

included 12 CpGs mapped in the HPV16 L1 ORF and specifically 5611, 5726, 5927,

6367, 6389, 6457, 6581, 6650, 6796, 7091, 7136, 7145 of clinical samples’ and SiHa

DNA. The pyrograms were analyzed using the CpG mode of the PyroMark Q24

software, to determine the percentage methylation at each site as well as the overall

mean methylation.

Statistical Analysis

For each included sample, we calculated the mean methylation level (i.e. the mean

value of the measured methylation percentages) by dividing the sum of the

8

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

percentage methylation levels for all positions (P01M-P12M) by the number of

positions (i.e.12). We produced box and whisker plots of the mean methylation, the

standard deviation (SD) and range for each histological group. Each histological

category was given a number starting from 1 (NEGATIVE) and ending to 6 (SCC). We

performed a regression analysis [17] to identify and describe the linear mathematical

equation governing the correlation between mean methylation percentage and

histological grade that includes the intercept and slope of the line. We used t-test to

determine significance. This mathematical formula was developed with the aim to

predict the histological grade from the mean methylation measured. We further

produced a fit plot of the mean methylation and the histological grades to describe

the linear model, the confidence limits of the line and the limits of the prediction

accuracy. We used unadjusted and adjusted R-square to describe the performance of

the regression analysis and the fit of the line to the data with a value near to 100%

describing the perfect fit. We calculated the modified adjusted R-squared that

adjusts for several predictors of the model and has been shown to be a better

performance metric [17]. These analyses were performed for each of the 12 CpG

sites. We further performed a linear regression analysis to explore whether the

mean methylation level is affected by age.

We further assessed the discriminative capability for each of the twelve methylation

positions (P01M–P12M) in predicting the disease at four different histological cut-

offs (CIN1+, CIN2+, CIN3+, Cancer). We produced Receiver Operating Characteristics

(ROC) curves [18] and calculated the area under the curve (AUC), the standard error

(SE) and the 95% confidence intervals (CI). The largest AUC (near to 100%) indicated

9

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

methylation positions that had the highest prediction of the histological outcome for

the chosen cut-off.

We calculated different accuracy parameters for the ability of the mean methylation

to detect the presence of disease for 4 histological cut-offs: CIN1+, CIN2+, CIN3+ or

cancer. These included the sensitivity (S), specificity (Sp), positive (PPV) and negative

predictive value (NPV), the false positive (FP) and false negative (FN) rate, the overall

accuracy (OA) and the positive (PLR) and negative likelihood ratio (NLR). We applied

various threshold values starting from 0% and increasing up to 100% using an

increment step of 0.1%, as described in our previous studies [19-21]. We further

described the accuracy parameters for different methylation thresholds at each

histological cut-offs, and calculated the mean methylation thresholds that optimize

the overall accuracy and/or the balance between sensitivity and specificity. We

calculated the odds ratio (OR) to describe the likelihood of the histological diagnosis

if the mean methylation was above or below a chosen threshold. We subsequently

tested the produced algorithm in a small validation set to confirm reproducibility of

the results.

We used SAS9.4 for the statistical analysis (SAS Institute Inc. NC, USA) [23, 24] and

the algorithms for the determination of the optimum threshold values were

developed in-house within the MATLAB software environment and programming

language (The MathWorks, Inc. Natick, Massachusetts, U.S.A.).

RESULTS

10

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

We identified a total of 151 women that tested positive for HPV16 type and had

histology; 1 was excluded as the histology revealed VAIN2 (Table 1). The mean age

differed amongst histological groups; women with invasive disease tended to be

older (P-value<0.0001). All women with normal cytology had normal findings at

histology. None of the women presented with ASCUS cytology had HSIL in histology,

although one in four women presenting with LSIL (12/20, 60%) had CIN2+ at

histology. One in ten women (11/105, 10.4%) that presented with HSIL cytology had

CIN1 or normal histology. All cases of invasive cancer presented with HSIL or possible

invasion on cytology. Approximately half of the population (46.7%) had HPV16 only

at genotyping, while the remaining women were also infected with other high-risk

(44.6) or low-risk (8.7%) subtypes.

We found that the mean L1 HPV16 methylation increased with increasing disease

severity. The mean methylation (SD, range) was 8.1 (3.7, 5.1-15.7) for normal, 9.0

(3.4, 4.5-15.8) for CIN1, 11.6 (6.5, 4.2-34.5) for CIN2, 17.9 (7.2, 6-41.5) for CIN3, 38.1

(10.8, 27-48.6) for adenocarcinoma and 58.1 (17.3, 42.5-86.2) for SCC (Figure 1A).

The HPV16 L1 mean methylation levels (SD) of CIN3+ cases (17.9%±7.2%) were

significantly higher compared to CIN2 (11.6%±6.5%) or CIN1 (9.0%±3.5%) histological

groups (t-test for CIN3+ vs CIN2: t=-4.6, p<0.001; CIN3+ vs CIN1: t=-8.9, p<0.001).

The regression analysis produced a formula to predict the correlation between the

mean methylation and the histological grade (Mean Methylation=-

7.65+7.24*Histological grade (expressed as number: starting from 1 (NEGATIVE) to 6

(SCC))(Figure 1B). The fitted line showed no tolerance as the intercept (-7.65) and

slope (7.24) was statistically significant (t-test:p<0.05 for both). The positive slope of

11

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

this line (7.65) confirmed that the mean methylation increases with increasing

disease grade. Similar trends were found for the disease severity and the mean

methylation in each of the individual CpG sites (results not shown). We found that

for the 3 subgroups of women with CIN1, 2 and 3, there was no evidence that the

mean methylation level was affected by the age (adjusted R-square: -0.047, 0.036

and -0.016, respectively). There was a trend towards lower methylation levels in

older women with CIN2, but the results were not statistically significant.

We quantified the methylation of CpG sites 5611, 5726, 5927, 6367, 6389, 6457,

6581, 6650, 6796, 7091, 7136 and 7145 in the HPV16 L1 gene. We constructed ROC

curves to assess the diagnostic utility of the methylation levels of each individual CpG

site (assigned P01M-P12M, respectively) and of the mean methylation in predicting

the histological diagnosis for the different cut-offs described (CIN1+, CIN2+, CIN3+,

Cancer)(Figure 2A and B, Supplementary Figures 1-5, Supplementary Table 1). We

found the ability to predict the histological diagnosis increased with increased cut-off

used and this was consistent for different methylation positions. The AUCs for CIN2+

lesions ranged from 0.66 to 0.82. The lowest AUC value (AUC:0.66,95%CI%:0.57-

0.76) observed in site: 6367, the highest (AUC:0.82,95%CI:0.75-0.90) observed in site

5611 for CIN2+ lesions, while the mean methylation for CIN2+ was 0.81 (95%CI:0.74-

0.88) (Figure 2A, Supplementary Figure 3). For CIN3+ lesions, the AUCs ranged from

0.71 to 0.86 with the highest value assigned to site 7145 (AUC:0.86,,95%CI:0.80-

0.92), the lowest to site 5726 (AUC:0.7195%CI:0.73-0.87) with the one for the mean

methylation at 0.85 (95%CI:0.79-0.91)(Figure 2B, Supplementary Figure 4). The

highest performance was for the histological cut-off of cancer as expected (mean

12

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

methylation AUC:0.99,95%CI:0.98-1.00), with limited though clinical use

(Supplementary Figure 1, 5).

We further determined accuracy parameters for different histological cut-offs and

described the mean methylation thresholds that optimized overall accuracy and/or

the balance between sensitivity and specificity (Table 2, Figure 3, Supplementary

Figure 6). We found that for a histological cut-off of CIN2+ (Figure 3A), a mean

methylation of 5.6% maximised the OA (82.7%) and achieved optimal sensitivity

(97.4%) at the loss of specificity (34.3%). We therefore applied a different mean

methylation threshold (10.8%) that optimized the balance between sensitivity

(67.8%) and specificity (74.3%) with an overall accuracy of 69.3%. For the prediction

of CIN3+, a methylation level of 15.9% optimized the overall accuracy 79.3%,

although the sensitivity dropped to 67.1% with a specificity being high at 90.0%

(Figure 3B). The optimal balance of sensitivity and specificity (75.7% and 77.5%,

respectively), PPV and NPV (74.7% and 78.5%, respectively) was achieved for a

methylation of 14.0% and OA of 76.7%. Higher methylation levels were used for the

detection of cancer cases, where a threshold of 37.9% achieved an OA of 98.6% with

a sensitivity of 88.9% and a specificity of 99.3%. The use of a mean methylation

threshold of 27.0% optimized sensitivity (100%) at a small loss in specificity (94.3%)

and OA (94.7%). With the exception of the histological cut-off of invasive cancer, the

odds ratio was higher (OR:18.39, 95%CI:7.59-44.54) for a mean methylation

threshold of 15.9% for the prediction of CIN3+ (ie. odds for CIN3+ are 18.4 times

higher if the mean methylation is above as opposed to less than 15.9%).

13

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

We evaluated the algorithm for a validation set of 22 (normal:5; CIN1:3; CIN2:4;

CIN3:8 and SCC:2)(data not shown). The results were consistent with the training set,

although the number of patients was small. For all thresholds, the sensitivity was

excellent, although the specificity varied. For the cut-off of CIN2+, the sensitivity was

100% and the specificity was 75.0%, compared to 67.8% and 74.3% in the training set (best

balance at 10.8%). All 14 CIN2+ cases were correctly assigned, while 2 out of 8 normal/CIN1

cases were misclassified as CIN2+. For the CIN3+ cut-off (maximum accuracy threshold at

15.9%) the sensitivity was 100% and the specificity 66.67% (training sets 67.1% and 90.0%,

respectively). The mean methylation enabled the correct classification of the 10 CIN3+ cases,

while misclassified as high-grade 4 out of 12 CIN2- cases.

DISCUSSION

Main findings and Interpretation in light of other evidence

The knowledge that oncogenic HPV types are causally associated with invasive

cervical cancer has initiated research into viral genome’s alterations during the viral

lifecycle and their interactions with the host genome [25-27]. The identification of

women screened that have CIN2+ has been one of the major challenges of concerted

efforts over the last decade, particularly as we are moving towards the use of HPV

test at primary screening. The second major challenge was to identify biomarkers

and molecular determinants that could distinguish the rare HPV infections that have

a true oncogenic malignant potential from those common infections from the same

types that resolve spontaneously without leading to disease.

14

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

Our study suggests that quantification of the methylation levels of 12 CpG sites in

the L1 gene with pyrosequencing has the potential to serve as a molecular marker in

the prediction of the severity of cervical disease. We studied 150 HPV16 positive

cases with histologically confirmed diagnosis. We found that mean methylation level

increased with increasing disease severity, while the methylation patterns of each

CpG studied did not differ substantially, as previously reported [14]. The mean

methylation levels for CIN3+ cases were significantly higher than CIN1 or CIN2 cases,

in agreement with previously reported results [14, 28]. Furthermore, we calculated

accuracy parameters for various methylation thresholds and defined those that

would optimize the diagnostic accuracy of the test.

Various methods of evaluation of viral DNA methylation have been published [9-11,

14, 28-31]. Some researchers used restriction endonucleases in non-bisulfite based

studies [13, 32, 33], whilst others employed bisulfite-based assays where cytosine

residues from single-stranded DNA are deaminated by sodium bisulfate and

converted to uracils with the exception of 5-methyl cytosines that are protected

from conversion [34-36]. In the present study we used bisulfite-treated DNA and the

methylation patterns were analysed by pyrosequencing which is a method with

reproducible and accurate measures of the degree of methylation at several CpGs in

the same amplicon [37].

There have been considerable efforts to study epigenetic factors affecting viral genes

and their associations to cervical cancer and precancer. Researchers have studied

CpG sites along HPV genes L1, L2, E2-E4, E5 and URR [9-11, 14, 29, 30, 38]. The

15

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

results of methylation status vary between studies. Cross-sectional studies in

diagnostic samples have correlated elevated methylation levels of various genes such

as L1, L2 and E2-E4 to the disease severity, while those from enhancer and promoter

region of the URR region being less reliable, inconsistent and heterogenous often

indicating higher methylation frequencies in normal as opposed to HSIL or invasive

samples [28].

Mirabello et al. [14] demonstrated that methylation of several CpG sites within L1, L2

and E2-E4 predicted the presence of CIN2 or worse in women with HPV16 infection

as well as the risk for future HSIL disease in women testing positive for HPV16. The

association between increased DNA methylation in the L1 regions and CIN3+ was

also confirmed by Sun et al. [30]. Another study that explored 13 CpG sites within

the L1 gene reported that higher methylation levels at certain L1 sites was associated

with HPV persistence and cervical precancerous progression [38]. Bryant et al. [39],

however, stated that although HPV DNA methylation may be a promising biomarker

in the triage of HPV-positive cytology samples, its value was less pronounced in

young women. Our results did not confirm a correlation between HPV methylation

and age, although the analysis was limited due to the small sample size in each

group.

We quantified the methylation of 12 CpGs within L1 gene and the highest accuracy

was found for sites 5611, for the prediction of CIN2+, and for site 7145 for the

prediction of CIN3+, which is consistent with other reports. Louvanto et al. [40]

reported high accuracy for sites CpG 6367 and 6389 of the L1 HPV16 region in the

16

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

prediction of CIN2+ (AUC>0.7). Mirabello et al. [14] reported an AUC=0.82 for CpG

6457, while for Niyazi et al. [38] the position 6650 appeared to be the most robust

CpG site (AUC>0.82). In our study, the estimation of the mean methylation from

12CpG sites did not achieve better accuracy in the prediction of the disease status

when compared to loci 5611 and 7145, suggesting that the selective use of few sites

that optimize the accuracy consistently across different studies may be the most

cost-effective option for the future. The reasons why different L1 sites performed

better for CIN2+ and CIN3+ are unclear, although this may be due to the modest

sample size.

Strengths and limitations

This is a large cohort of patients assessing the methylation levels in the L1 gene in

women with histologically confirmed diagnosis of the disease status that also

provides data on the methylation thresholds that maximize accuracy of the

molecular marker. All samples were analysed in a single laboratory to minimize bias.

The findings add valuable information towards the better understanding of the

biological behavior of HPV16 and the ‘cross-talk’ between the viral and host genome

leading to increased methylated-CpG content in the L1 gene and disease

progression. Although this was a sizeable cohort of patients, there were only a

limited number of samples for healthy controls and for some of the histological

groups making interpretation of the results for these groups difficult. Correlation of

methylation levels to age was limited due to the limited number of samples in the

different histological subgroups. We tested the algorithm in a small validation

cohort. Although the results have shown similar trends as the training set, the

17

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

number of patients was small and we were not able to assess whether there were

statistically significant differences. Future studies should assess the reproducibility of

these results in large validation sets. Future studies should also analyze serial

samples from larger cohorts to further assess the value of methylation as a

predictive and diagnostic molecular determinant. The accuracy of HPV methylation

should be also compared to that of other currently used tests and biomarkers such

as cytology and HPV mRNA.

Conclusions

Elevated methylation levels within the L1 region of the viral DNA are associated with

increased disease severity. This molecular determinant may have a role in the triage

of screen-detected HPV positive women that warrant colposcopic investigation, but

has the potential to further distinguish the infections and pre-invasive lesions most

likely to progress allowing more advanced prognostic risk stratification. This will

allow the identification of women that would benefit from colposcopic assessment

with or without treatment with a reduction in the unnecessary visits and the

reproductive morbidity associated with interventions and treatment [41-45]. Further

serial samples from well-established biobanks should assess the impact of the viral

methylation levels to the infection/disease progression at various stages of the

natural history of the disease. Studies should assess its diagnostic accuracy for

different clinical groups and applications.

18

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

Acknowledgements: We thank all the participants of the study.

Authors’ contribution: The study was conceived and designed by CK, AP, MK and PK.

The samples and data was acquired and collated by MK, AM, CC, GM, VM, EP and

analyzed by CK, AP, MM, EA, AS, GJP and MK. The manuscript was drafted and

revised critically for important intellectual content by all authors (CK, MK, AP, MM,

EA, AS, AM, VM, EP, JGP, PK). CK and MK are joint first authors. All authors gave final

approval of the version to be published and have contributed to the manuscript.

19

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

References

1. Saslow D, Solomon D, Lawson HW, et al. American Cancer Society, American Society for

Colposcopy and Cervical Pathology, and American Society for Clinical Pathology screening

guidelines for the prevention and early detection of cervical cancer. Am J Clin Pathol 2012;

137:516-42.

2. Quinn M, Babb P, Jones J, Allen E. Effect of screening on incidence of and mortality from

cancer of cervix in England: evaluation based on routinely collected statistics. BMJ 1999;

318:904-8.

3. Koliopoulos G, Sotiriadis A, Kyrgiou M, Martin-Hirsch P, Makrydimas G, Paraskevaidis E.

Conservative surgical methods for FIGO stage IA2 squamous cervical carcinoma and their

role in preserving women's fertility. Gynecol Oncol 2004; 93:469-73.

4. Kyrgiou M, Pouliakis A, Panayiotides JG, et al. Personalised management of women with

cervical abnormalities using a clinical decision support scoring system. Gynecol Oncol 2016;

141:29-35.

5. Nasioutziki M, Daniilidis A, Dinas K, et al. The evaluation of p16INK4a

immunoexpression/immunostaining and human papillomavirus DNA test in cervical liquid-

based cytological samples. Int J Gynecol Cancer 2011; 21:79-85.

6. Tsoumpou I, Valasoulis G, Founta C, et al. High-risk human papillomavirus DNA test and

p16(INK4a) in the triage of LSIL: a prospective diagnostic study. Gynecol Oncol 2011; 121:49-

53.

7. Gustinucci D, Giorgi Rossi P, Cesarini E, et al. Use of Cytology, E6/E7 mRNA, and p16INK4a-

Ki-67 to Define the Management of Human Papillomavirus (HPV)-Positive Women in Cervical

Cancer Screening. Am J Clin Pathol 2016; 145:35-45.

8. Gage JC, Hunt WC, Schiffman M, et al. Risk Stratification Using Human Papillomavirus

Testing among Women with Equivocally Abnormal Cytology: Results from a State-Wide

Surveillance Program. Cancer Epidemiol Biomarkers Prev 2016; 25:36-42.

20

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

9. Fernandez AF, Rosales C, Lopez-Nieva P, et al. The dynamic DNA methylomes of double-

stranded DNA viruses associated with human cancer. Genome Res 2009; 19:438-51.

10. Mirabello L, Sun C, Ghosh A, et al. Methylation of human papillomavirus type 16 genome

and risk of cervical precancer in a Costa Rican population. Journal of the National Cancer

Institute 2012; 104:556-65.

11. Brandsma JL, Sun Y, Lizardi PM, et al. Distinct human papillomavirus type 16 methylomes

in cervical cells at different stages of premalignancy. Virology 2009; 389:100-7.

12. Piyathilake CJ, Macaluso M, Alvarez RD, et al. A higher degree of methylation of the HPV

16 E6 gene is associated with a lower likelihood of being diagnosed with cervical

intraepithelial neoplasia. Cancer 2011; 117:957-63.

13. Hublarova P, Hrstka R, Rotterova P, et al. Prediction of human papillomavirus 16 e6 gene

expression and cervical intraepithelial neoplasia progression by methylation status. Int J

Gynecol Cancer 2009; 19:321-5.

14. Mirabello L, Schiffman M, Ghosh A, et al. Elevated methylation of HPV16 DNA is

associated with the development of high grade cervical intraepithelial neoplasia.

International journal of cancer Journal international du cancer 2013; 132:1412-22.

15. Solomon D, Davey D, Kurman R, et al. The 2001 Bethesda System: terminology for

reporting results of cervical cytology. JAMA 2002; 287:2114-9.

16. el Awady MK, Kaplan JB, O'Brien SJ, Burk RD. Molecular analysis of integrated human

papillomavirus 16 sequences in the cervical cancer cell line SiHa. Virology 1987; 159:389-98.

17. Huang N. Paper 1760-2014 Scenarios Where Utilizing a Spline Model in Developing a

Regression Model Is Appropriate. Available at:

http://support.sas.com/resources/papers/proceedings14/1760-2014.pdf. Accessed 28 June

2016.

21

479

480

481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

18. Shin S, Coulter B. Paper SP09-2009 ROC analysis for the evaluation of continuous

biomarkers: Existing tools and new features in SAS® 9.2 Available at:

http://pharmasug.org/download/papers/SP09.pdf. Accessed 28 June 2016.

19. Varlatzidou A, Pouliakis A, Stamataki M, et al. Cascaded learning vector quantizer neural

networks for the discrimination of thyroid lesions. Analytical and quantitative cytology and

histology / the International Academy of Cytology [and] American Society of Cytology 2011;

33:323-34.

20. Pouliakis A, Margari C, Margari N, et al. Using classification and regression trees, liquid-

based cytology and nuclear morphometry for the discrimination of endometrial lesions.

Diagn Cytopathol 2014; 42:582-91.

21. Zygouris D, Pouliakis A, Margari N, et al. Classification of endometrial lesions by nuclear

morphometry features extracted from liquid-based cytology samples: a system based on

logistic regression model. Anal Quant Cytopathol Histpathol 2014; 36:189-98.

22. Pouliakis A, Margari C, Margari N, et al. Using classification and regression trees, liquid-

based cytology and nuclear morphometry for the discrimination of endometrial lesions.

Diagnostic cytopathology 2013:In press (early view available).

23. DiMaggio C. SAS for epidemiologists : applications and methods. New York: Springer,

2013.

24. SAS Institute. SAS Home Page. Available at: http://www.sas.com.

25. Xu L, Verdoodt F, Wentzensen N, Bergeron C, Arbyn M. Triage of ASC-H: A meta-analysis

of the accuracy of high-risk HPV testing and other markers to detect cervical precancer.

Cancer Cytopathol 2016; 124:261-72.

26. Arbyn M, Roelens J, Simoens C, et al. Human papillomavirus testing versus repeat

cytology for triage of minor cytological cervical lesions. Cochrane Database Syst Rev 2013;

3:CD008054.

22

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

27. Lynge E, Antilla A, Arbyn M, Segnan N, Ronco G. What's next? Perspectives and future

needs of cervical screening in Europe in the era of molecular testing and vaccination. Eur J

Cancer 2009; 45:2714-21.

28. Clarke MA, Wentzensen N, Mirabello L, et al. Human papillomavirus DNA methylation as

a potential biomarker for cervical cancer. Cancer Epidemiol Biomarkers Prev 2012; 21:2125-

37.

29. Kalantari M, Garcia-Carranca A, Morales-Vazquez CD, et al. Laser capture microdissection

of cervical human papillomavirus infections: copy number of the virus in cancerous and

normal tissue and heterogeneous DNA methylation. Virology 2009; 390:261-7.

30. Sun C, Reimers LL, Burk RD. Methylation of HPV16 genome CpG sites is associated with

cervix precancer and cancer. Gynecol Oncol 2011; 121:59-63.

31. Sun C, McAndrew T, Smith BC, Chen Z, Frimer M, Burk RD. Characterization of HPV DNA

methylation of contiguous CpG sites by bisulfite treatment and massively parallel

sequencing-the FRAGMENT approach. Front Genet 2014; 5:150.

32. Badal V, Chuang LS, Tan EH, et al. CpG methylation of human papillomavirus type 16 DNA

in cervical cancer cell lines and in clinical specimens: genomic hypomethylation correlates

with carcinogenic progression. J Virol 2003; 77:6227-34.

33. Badal S, Badal V, Calleja-Macias IE, et al. The human papillomavirus-18 genome is

efficiently targeted by cellular DNA methylation. Virology 2004; 324:483-92.

34. Eads CA, Danenberg KD, Kawakami K, et al. MethyLight: a high-throughput assay to

measure DNA methylation. Nucleic Acids Res 2000; 28:E32.

35. Herman JG, Graff JR, Myohanen S, Nelkin BD, Baylin SB. Methylation-specific PCR: a novel

PCR assay for methylation status of CpG islands. Proc Natl Acad Sci U S A 1996; 93:9821-6.

36. Fraga MF, Esteller M. DNA methylation: a profile of methods and applications.

Biotechniques 2002; 33:632, 4, 6-49.

37. Tost J, Gut IG. DNA methylation analysis by pyrosequencing. Nat Protoc 2007; 2:2265-75.

23

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

38. Niyazi M, Sui S, Zhu K, Wang L, Jiao Z, Lu P. Correlation between Methylation of Human

Papillomavirus-16 L1 Gene and Cervical Carcinoma in Uyghur Women. Gynecol Obstet Invest

2016.

39. Bryant D, Hibbitts S, Almonte M, Tristram A, Fiander A, Powell N. Human papillomavirus

type 16 L1/L2 DNA methylation shows weak association with cervical disease grade in young

women. J Clin Virol 2015; 66:66-71.

40. Louvanto K, Franco EL, Ramanakumar AV, et al. Methylation of viral and host genes and

severity of cervical lesions associated with human papillomavirus type 16. Int J Cancer 2015;

136:E638-45.

41. Kyrgiou M, Athanasiou A, Paraskevaidi M, et al. Adverse obstetric outcomes after local

treatment for cervical preinvasive and early invasive disease according to cone depth:

systematic review and meta-analysis. BMJ 2016; 354:i3633.

42. Kyrgiou M, Mitra A, Arbyn M, et al. Fertility and early pregnancy outcomes after

treatment for cervical intraepithelial neoplasia: systematic review and meta-analysis. BMJ

2014; 349:g6192.

43. Kyrgiou M, Arbyn M, Martin-Hirsch P, Paraskevaidis E. Increased risk of preterm birth

after treatment for CIN. BMJ 2012; 345:e5847.

44. Kyrgiou M, Valasoulis G, Stasinou SM, et al. Proportion of cervical excision for cervical

intraepithelial neoplasia as a predictor of pregnancy outcomes. Int J Gynaecol Obstet 2015;

128:141-7.

45. Founta C, Arbyn M, Valasoulis G, et al. Proportion of excision and cervical healing after

large loop excision of the transformation zone for cervical intraepithelial neoplasia. BJOG

2010; 117:1468-74.

24

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

Figures legends

Figure 1. A: Box and whicker plot of the mean methylation for the studied

histological categories. The mean methylation values increase with increasing

disease severity. Each box contains the range from 25% to 75%, dashes inside the

boxes indicate the median value and diamonds the mean value. Whisker limits

indicate the minimum and maximum values, circles out of the whisker ranges

indicate outliers. B: Fit plot of the mean methylation and the histological categories.

The solid line represents the graph of the equation that predicts mean methylation

from the histological category. The shaded area represents the 95 confidence

interval limits of the fitted line and the dashed lines the range of the prediction

limits. The fit performance is represented by R-square=44.33% (Adjusted R-

square=43.96%).

Figure 2: Receiver Operating Characteristic (ROC) curves of the twelve methylation

positions (P01M-P12M) each one corresponding to CpG position (5611, 5726, 5927,

6367, 6389, 6457, 6581, 6650, 6796, 7091, 7136 and 7145) and the mean

methylation (bold lines) for the discrimination of: A) CIN2+ and B) CIN3+ cases. All

methylation positions exhibit similar behavior. The Area Under Curve (AUC) increases

as the cut-off severity increased from CIN2+ to CIN3+, for the majority of the

positions and the mean methylation. Mean methylation exhibits better performance

than most individual positions for both CIN2+ and CIN3+ thresholds.

25

580

581

582

583

584

585

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

601

602

Figure 3: Overall accuracy (bold lines), sensitivity (dashed lines) and specificity (solid

lines) for different mean methylation threshold values. Performance characteristics

are presented for: A) CIN2+ B) CIN3+. The horizontal axis represents different

threshold values for the mean methylation. Higher methylation thresholds increase

specificity, but at the cost of sensitivity.

26

603

604

605

606

607