Limits...
Unique DNA methylation signature in HPV-positive head and neck squamous cell carcinomas

View Article: PubMed Central - PubMed

ABSTRACT

Background: Head and neck squamous cell carcinomas (HNSCCs) represent a heterogeneous group of cancers for which human papilloma virus (HPV) infection is an emerging risk factor. Previous studies showed promoter hypermethylation in HPV(+) oropharyngeal cancers, but only few consistent target genes have been so far described, and the evidence of a functional impact on gene expression is still limited.

Methods: We performed global and stratified pooled analyses of epigenome-wide data in HNSCCs based on the Illumina HumanMethylation450 bead-array data in order to identify tissue-specific components and common viral epigenetic targets in HPV-associated tumours.

Results: We identified novel differentially methylated CpGs and regions associated with viral infection that are independent of the anatomic site. In particular, most hypomethylated regions were characterized by a marked loss of CpG island boundaries, which showed significant correlations with expression of neighbouring genes. Moreover, a subset of only five CpGs in a few hypomethylated regions predicted HPV status with a high level of specificity in different cohorts. Finally, this signature was a better predictor of survival compared with HPV status determined by viral gene expression by RNA sequencing in The Cancer Genome Atlas cohort.

Conclusions: We identified a novel epigenetic signature of HPV infection in HNSCCs which is independent of the anatomic site, is functionally correlated with gene expression and may be leveraged for improved stratification of prognosis in HNSCCs.

Electronic supplementary material: The online version of this article (doi:10.1186/s13073-017-0419-z) contains supplementary material, which is available to authorized users.

No MeSH data available.


Related in: MedlinePlus

DNA methylation predictive signature of HPV status in HNSCCs. a Overall and class-specific misclassification errors based on the number of CpGs selected to predict HPV status. b AUC values from PAM algorithm according to the number of probes selected to predict the HPV status. The red dot indicates the AUC for 10-CpG signatures. Similar results were obtained using RF algorithm (data not shown). c Receiver operating characteristic (ROC) curves (AUC) using the training set data. d ROC curves (AUC) using the test set data. e Average methylation index (AMI) of the 5-CpG signature across the different datasets (TCGA cohort, UCL cohort, FITMANET cohort 450 K, FITMANET validation cohort). f AMI of the 5-CpG signature in cervical carcinomas from TCGA. Similar to HNSCCs signature, high methylation is considered when the AMI is higher than 0.75
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC5382363&req=5

Fig5: DNA methylation predictive signature of HPV status in HNSCCs. a Overall and class-specific misclassification errors based on the number of CpGs selected to predict HPV status. b AUC values from PAM algorithm according to the number of probes selected to predict the HPV status. The red dot indicates the AUC for 10-CpG signatures. Similar results were obtained using RF algorithm (data not shown). c Receiver operating characteristic (ROC) curves (AUC) using the training set data. d ROC curves (AUC) using the test set data. e Average methylation index (AMI) of the 5-CpG signature across the different datasets (TCGA cohort, UCL cohort, FITMANET cohort 450 K, FITMANET validation cohort). f AMI of the 5-CpG signature in cervical carcinomas from TCGA. Similar to HNSCCs signature, high methylation is considered when the AMI is higher than 0.75

Mentions: To test whether HPV-associated methylation changes may be predictive of HPV status, we next analysed the normalized data using a prediction analysis of microarrays (PAM) algorithm and a random forest (RF) algorithm (see Methods section), performing five cross-validations for each method. ROC curve (AUC) analysis showed that methylation values of as low as 10 CpGs (Fig. 5a, b) were able to give AUC >0.95 in most of the training and the test sets (Fig. 5c, d). Predictive signatures of 10 CpGs for each cross-validation resulted in an average sensitivity of 89% (87–94%) and an average specificity of 96% (95–97%) in the training test. In the test set, the average sensitivity was 89% (83–94%) and the average specificity was 95% (range 94–97%). The recurrent CpGs across the five cross-validated signatures of 10 CpGs were localized in the B3GALT6-SDF4 locus (3 CpGs), in the SYCP2-FAM217B locus (1 CpG), the HLTF-HLTF-AS1 locus (from 1 to 3 CpGs), TLX2 gene (from 1 to 2 CpGs), LOC729683 (from 1 to 2 CpGs), IL4I4 gene (1 CpG) and LINC00925 (1 CpG).Fig. 5


Unique DNA methylation signature in HPV-positive head and neck squamous cell carcinomas
DNA methylation predictive signature of HPV status in HNSCCs. a Overall and class-specific misclassification errors based on the number of CpGs selected to predict HPV status. b AUC values from PAM algorithm according to the number of probes selected to predict the HPV status. The red dot indicates the AUC for 10-CpG signatures. Similar results were obtained using RF algorithm (data not shown). c Receiver operating characteristic (ROC) curves (AUC) using the training set data. d ROC curves (AUC) using the test set data. e Average methylation index (AMI) of the 5-CpG signature across the different datasets (TCGA cohort, UCL cohort, FITMANET cohort 450 K, FITMANET validation cohort). f AMI of the 5-CpG signature in cervical carcinomas from TCGA. Similar to HNSCCs signature, high methylation is considered when the AMI is higher than 0.75
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC5382363&req=5

Fig5: DNA methylation predictive signature of HPV status in HNSCCs. a Overall and class-specific misclassification errors based on the number of CpGs selected to predict HPV status. b AUC values from PAM algorithm according to the number of probes selected to predict the HPV status. The red dot indicates the AUC for 10-CpG signatures. Similar results were obtained using RF algorithm (data not shown). c Receiver operating characteristic (ROC) curves (AUC) using the training set data. d ROC curves (AUC) using the test set data. e Average methylation index (AMI) of the 5-CpG signature across the different datasets (TCGA cohort, UCL cohort, FITMANET cohort 450 K, FITMANET validation cohort). f AMI of the 5-CpG signature in cervical carcinomas from TCGA. Similar to HNSCCs signature, high methylation is considered when the AMI is higher than 0.75
Mentions: To test whether HPV-associated methylation changes may be predictive of HPV status, we next analysed the normalized data using a prediction analysis of microarrays (PAM) algorithm and a random forest (RF) algorithm (see Methods section), performing five cross-validations for each method. ROC curve (AUC) analysis showed that methylation values of as low as 10 CpGs (Fig. 5a, b) were able to give AUC >0.95 in most of the training and the test sets (Fig. 5c, d). Predictive signatures of 10 CpGs for each cross-validation resulted in an average sensitivity of 89% (87–94%) and an average specificity of 96% (95–97%) in the training test. In the test set, the average sensitivity was 89% (83–94%) and the average specificity was 95% (range 94–97%). The recurrent CpGs across the five cross-validated signatures of 10 CpGs were localized in the B3GALT6-SDF4 locus (3 CpGs), in the SYCP2-FAM217B locus (1 CpG), the HLTF-HLTF-AS1 locus (from 1 to 3 CpGs), TLX2 gene (from 1 to 2 CpGs), LOC729683 (from 1 to 2 CpGs), IL4I4 gene (1 CpG) and LINC00925 (1 CpG).Fig. 5

View Article: PubMed Central - PubMed

ABSTRACT

Background: Head and neck squamous cell carcinomas (HNSCCs) represent a heterogeneous group of cancers for which human papilloma virus (HPV) infection is an emerging risk factor. Previous studies showed promoter hypermethylation in HPV(+) oropharyngeal cancers, but only few consistent target genes have been so far described, and the evidence of a functional impact on gene expression is still limited.

Methods: We performed global and stratified pooled analyses of epigenome-wide data in HNSCCs based on the Illumina HumanMethylation450 bead-array data in order to identify tissue-specific components and common viral epigenetic targets in HPV-associated tumours.

Results: We identified novel differentially methylated CpGs and regions associated with viral infection that are independent of the anatomic site. In particular, most hypomethylated regions were characterized by a marked loss of CpG island boundaries, which showed significant correlations with expression of neighbouring genes. Moreover, a subset of only five CpGs in a few hypomethylated regions predicted HPV status with a high level of specificity in different cohorts. Finally, this signature was a better predictor of survival compared with HPV status determined by viral gene expression by RNA sequencing in The Cancer Genome Atlas cohort.

Conclusions: We identified a novel epigenetic signature of HPV infection in HNSCCs which is independent of the anatomic site, is functionally correlated with gene expression and may be leveraged for improved stratification of prognosis in HNSCCs.

Electronic supplementary material: The online version of this article (doi:10.1186/s13073-017-0419-z) contains supplementary material, which is available to authorized users.

No MeSH data available.


Related in: MedlinePlus