Improved high-dimensional prediction with Random Forests by the use of co-data

Background

Prediction in high dimensional settings is difficult due to the large number of variables relative to the sample size. We demonstrate how auxiliary ‘co-data’ can be used to improve the performance of a Random Forest in such a setting.

Results

Co-data are incorporated in the Random Forest by replacing the uniform sampling probabilities that are used to draw candidate variables by co-data moderated sampling probabilities. Co-data here are defined as any type information that is available on the variables of the primary data, but does not use its response labels. These moderated sampling probabilities are, inspired by empirical Bayes, learned from the data at hand. We demonstrate the co-data moderated Random Forest (CoRF) with two examples. In the first example we aim to predict the presence of a lymph node metastasis with gene expression data. We demonstrate how a set of external p-values, a gene signature, and the correlation between gene expression and DNA copy number can improve the predictive performance. In the second example we demonstrate how the prediction of cervical (pre-)cancer with methylation data can be improved by including the location of the probe relative to the known CpG islands, the number of CpG sites targeted by a probe, and a set of p-values from a related study.

Conclusion

The proposed method is able to utilize auxiliary co-data to improve the performance of a Random Forest.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-017-1993-1) contains supplementary material, which is available to authorized users.

Better diagnostic signatures from RNAseq data through use of auxiliary co-data

SUMMARY

Our aim is to improve omics based prediction and feature selection using multiple sources of auxiliary information: co-data. Adaptive group regularized ridge regression (GRridge) was proposed to achieve this by estimating additional group-based penalty parameters through an empirical Bayes method at a low computational cost. We illustrate the GRridge method and software on RNA sequencing datasets. The method boosts the performance of an ordinary ridge regression and outperforms other classifiers. Post-hoc feature selection maintains the predictive ability of the classifier with far fewer markers.

AVAILABILITY AND IMPLEMENTATION

GRridge is an R package that includes a vignette. It is freely available at ( https://bioconductor.org/packages/GRridge/ ). All information and R scripts used in this study, including those on retrieval and processing of the co-data, are available from http://github.com/markvdwiel/GRridgeCodata .

CONTACT

mark.vdwiel@vumc.nl.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Ridge estimation of the VAR(1) model and its time series chain graph from multivariate time-course omics data

Omics experiments endowed with a time-course design may enable us to uncover the dynamic interplay among genes of cellular processes. Multivariate techniques (like VAR(1) models describing the temporal and contemporaneous relations among variates) that may facilitate this goal are hampered by the high-dimensionality of the resulting data. This is resolved by the presented ridge regularized maximum likelihood estimation procedure for the VAR(1) model. Information on the absence of temporal and contemporaneous relations may be incorporated in this procedure. Its computational efficient implemention is discussed. The estimation procedure is accompanied with an LOOCV scheme to determine the associated penalty parameters. Downstream exploitation of the estimated VAR(1) model is outlined: an empirical Bayes procedure to identify the interesting temporal and contemporaneous relationships, impulse response analysis, mutual information analysis, and covariance decomposition into the (graphical) relations among variates. In a simulation study the presented ridge estimation procedure outperformed a sparse competitor in terms of Frobenius loss of the estimates, while their selection properties are on par. The proposed machinery is illustrated in the reconstruction of the p53 signaling pathway during HPV-induced cellular transformation. The methodology is implemented in the ragt2ridges R-package available from CRAN.

Interplay between promoter methylation and chromosomal loss in gene silencing at 3p11-p14 in cervical cancer

Loss of 3p11-p14 is a frequent event in epithelial cancer and a candidate prognostic biomarker in cervical cancer. In addition to loss, promoter methylation can participate in gene silencing and promote tumor aggressiveness. We have performed a complete mapping of promoter methylation at 3p11-p14 in two independent cohorts of cervical cancer patients (n = 149, n = 121), using Illumina 450K methylation arrays. The aim was to investigate whether hyperm-ethylation was frequent and could contribute to gene silencing and disease aggressiveness either alone or combined with loss. By comparing the methylation level of individual CpG sites with corresponding data of normal cervical tissue, 26 out of 41 genes were found to be hypermethylated in both cohorts. The frequency of patients with hypermethylation of these genes was found to be higher at tumor stages of 3 and 4 than in stage 1 tumors. Seventeen of the 26 genes were transcriptionally downregulated in cancer compared to normal tissue, whereof 6 genes showed a significant correlation between methylation and expression. Integrated analysis of methylation, gene dosage, and expression of the 26 hypermethylated genes identified 3 regulation patterns encompassing 8 hypermethylated genes; a methylation driven pattern (C3orf14, GPR27, ZNF717), a gene dosage driven pattern (THOC7, PSMD6), and a combined methylation and gene dosage driven pattern (FHIT, ADAMTS9, LRIG1). In survival analysis, patients with both hypermethylation and loss of LRIG1 had a worse outcome compared to those harboring only hypermethylation or none of the events. C3orf14 emerged as a novel methylation regulated suppressor gene, for which knockdown was found to promote invasive growth in human papilloma virus (HPV)-transformed keratinocytes. In conclusion, hypermethylation at 3p11-p14 is common in cervical cancer and may exert a selection pressure during carcinogenesis alone or combined with loss. Information on both events could lead to improved prognostic markers.

Molecular events leading to HPV-induced high grade neoplasia

Cervical cancer is initiated by high-risk types of the human papillomavirus (hrHPV) and develops via precursor stages, called cervical intraepithelial neoplasia (CIN). High-grade CIN lesions are considered true precancerous lesions when the viral oncogenes E6 and E7 are aberrantly expressed in the dividing cells. This results in abolishment of normal cell cycle control via p53 and pRb degradation. However, it has become clear that these viral oncogenes possess additional oncogenic properties, including interference with the DNA methylation machinery and mitotic checkpoints. Identification of the resulting molecular events leading to high-grade neoplasia will 1) increase our understanding of cervical carcinogenesis, 2) yield biomarkers for early diagnosis, and 3) identify therapeutic targets for HPV-induced (pre) cancerous lesions.

This review will briefly summarise current advances in our understanding of the molecular alterations in the host cell genome that occur during HPV-induced carcinogenesis.

Better prediction by use of co-data: adaptive group-regularized ridge regression

For many high-dimensional studies, additional information on the variables, like (genomic) annotation or external p-values, is available. In the context of binary and continuous prediction, we develop a method for adaptive group-regularized (logistic) ridge regression, which makes structural use of such 'co-data'. Here, 'groups' refer to a partition of the variables according to the co-data. We derive empirical Bayes estimates of group-specific penalties, which possess several nice properties: (i) They are analytical. (ii) They adapt to the informativeness of the co-data for the data at hand. (iii) Only one global penalty parameter requires tuning by cross-validation. In addition, the method allows use of multiple types of co-data at little extra computational effort. We show that the group-specific penalties may lead to a larger distinction between 'near-zero' and relatively large regression parameters, which facilitates post hoc variable selection. The method, termed GRridge, is implemented in an easy-to-use R-package. It is demonstrated on two cancer genomics studies, which both concern the discrimination of precancerous cervical lesions from normal cervix tissues using methylation microarray data. For both examples, GRridge clearly improves the predictive performances of ordinary logistic ridge regression and the group lasso. In addition, we show that for the second study, the relatively good predictive performance is maintained when selecting only 42 variables.

Longitudinal assessment of DNA methylation changes during HPVE6E7-induced immortalization of primary keratinocytes

High-risk human papillomavirus (hrHPV)-induced immortalization and malignant transformation are accompanied by DNA methylation of host genes. To determine when methylation is established during cell immortalization and whether it is hrHPV-type dependent, DNA methylation was studied in a large panel of HPVE6E7-immortalized keratinocyte cell lines. These cell lines displayed different growth behaviors, i.e., continuous growth versus crisis period prior to immortalization, reflecting differential immortalization capacities of the 7 HPV-types (16/18/31/33/45/66/70) studied. In this study, cells were monitored for hypermethylation of 14 host genes (APC, CADM1, CYGB, FAM19A4, hTERT, mir124-1, mir124-2, mir124-3, MAL, PHACTR3, PRDM14, RASSF1A, ROBO3, and SFRP2) at 4 different stages during immortalization. A significant increase in overall methylation levels was seen with progression through each stage of immortalization. At stage 1 (pre-immortalization), a significant increase in methylation of hTERT, mir124-2, and PRDM14 was already apparent, which continued over time. Methylation of ROBO3 was significantly increased at stage 2 (early immortal), followed by CYGB (stage 3) and FAM19A4, MAL, PHACTR3, and SFRP2 (stage 4). Methylation patterns were mostly growth behavior independent. Yet, hTERT methylation levels were significantly increased in cells that just escaped from crisis. Bisulfite sequencing of hTERT confirmed increased methylation in immortal cells compared to controls, with the transcription core and known repressor sites remaining largely unmethylated. In conclusion, HPV-induced immortalization is associated with a sequential and progressive increase in promoter methylation of a subset of genes, which is mostly independent of the viral immortalization capacity.

tigaR: integrative significance analysis of temporal differential gene expression induced by genomic abnormalities

Background

To determine which changes in the host cell genome are crucial for cervical carcinogenesis, a longitudinal in vitro model system of HPV-transformed keratinocytes was profiled in a genome-wide manner. Four cell lines affected with either HPV16 or HPV18 were assayed at 8 sequential time points for gene expression (mRNA) and gene copy number (DNA) using high-resolution microarrays. Available methods for temporal differential expression analysis are not designed for integrative genomic studies.

Results

Here, we present a method that allows for the identification of differential gene expression associated with DNA copy number changes over time. The temporal variation in gene expression is described by a generalized linear mixed model employing low-rank thin-plate splines. Model parameters are estimated with an empirical Bayes procedure, which exploits integrated nested Laplace approximation for fast computation. Iteratively, posteriors of hyperparameters and model parameters are estimated. The empirical Bayes procedure shrinks multiple dispersion-related parameters. Shrinkage leads to more stable estimates of the model parameters, better control of false positives and improvement of reproducibility. In addition, to make estimates of the DNA copy number more stable, model parameters are also estimated in a multivariate way using triplets of features, imposing a spatial prior for the copy number effect.

Conclusion

With the proposed method for analysis of time-course multilevel molecular data, more profound insight may be gained through the identification of temporal differential expression induced by DNA copy number abnormalities. In particular, in the analysis of an integrative oncogenomics study with a time-course set-up our method finds genes previously reported to be involved in cervical carcinogenesis. Furthermore, the proposed method yields improvements in sensitivity, specificity and reproducibility compared to existing methods. Finally, the proposed method is able to handle count (RNAseq) data from time course experiments as is shown on a real data set.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2105-15-327) contains supplementary material, which is available to authorized users.

Methylation marker analysis of self-sampled cervico-vaginal lavage specimens to triage high-risk HPV-positive women for colposcopy

Methylation markers were studied for their suitability to triage human papillomavirus (HPV)-positive women by testing self-collected cervico-vaginal lavage specimens. For this purpose, we analyzed 355 hrHPV-positive self-collected specimens with three methylation markers, that is, CADM1-m18, MAL-m1 and miR-124-2 by quantitative methylation-specific PCR. The areas under the receiver-operating characteristic (ROC) curve for end-point cervical intraepithelial neoplasia grade 3 or worse (CIN3+) were 0.637 for CADM1-m18, 0.767 for MAL-m1 and 0.762 for miR-124-2. This indicates that CADM1-m18 is not suitable as single marker. By varying the thresholds of both markers in the bi-marker panels CADM1-m18/MAL-m1, CADM1-m18/miR-124-2 and MAL-m1/miR-124-2 upper and lower ROC curves were obtained, depicting the maximum and minimum CIN3+ sensitivity, respectively, at given specificity. For all these bi-marker combinations, the upper curves were similar. However, for the MAL-m1/miR-124-2 panel, the distance between upper and lower ROC curves was closest and this panel displayed the highest assay thresholds, indicating that this combination was most robust. At clinical specificities of 50 and 70%, the MAL-m1/miR-124-2 sensitivity for detection of CIN3+ ranged from 77.0 to 87.8% and from 64.9 to 71.6%, respectively. At 70% specificity thresholds no carcinomas were missed. By comparison, the CIN3+ sensitivity of HPV16/18 genotyping on the self-sampled lavage specimens was 58.1% (95%CI: 46.6-68.8) at a specificity of 87.7% (95%CI: 83.2-91.2). In conclusion, methylation analysis is a promising triage tool that in combination with HPV-DNA testing offers feasible, full molecular screening on self-collected cervico-vaginal lavage specimens.

Modeling the DNA copy number aberration patterns in observational high-throughput cancer data

The process of occurrence of genomic aberrations over time in the genetic material of cancer cells reflects the progression of the cancer. Modern technologies like aCGH (array Comparative Genomic Hybridization) and MPS (Massive Parallel Sequencing) provide high-resolution measurements of DNA copy number aberrations, that reveal the full scale of genomic aberrations. A continuous time Markov chain model is proposed to describe the accumulation of aberrations over time. Time however is a latent variable (with the number of aberrations as a proxy). Integrating out time, yields the distribution of the observed DNA copy number data. The model parameters are estimated from high-dimensional DNA copy number data by means of penalized maximum pseudo- and likelihood and method of moments procedures. Having fitted the model, posterior time estimates of the advancement of each sample's cancer are obtained and the most likely locations of a sample's aberrations are predicted. The three estimation methods are compared in a simulation study. The paper closes with an application of the proposed methodology on cancer data.

Identification of eight candidate target genes of the recurrent 3p12-p14 loss in cervical cancer by integrative genomic profiling

The pathogenetic role, including its target genes, of the recurrent 3p12-p14 loss in cervical cancer has remained unclear. To determine the onset of the event during carcinogenesis, we used microarray techniques and found that the loss was the most frequent 3p event, occurring in 61% of 92 invasive carcinomas, in only 2% of 43 high-grade intraepithelial lesions (CIN2/3), and in 33% of 6 CIN3 lesions adjacent to invasive carcinomas, suggesting a role in acquisition of invasiveness or early during the invasive phase. We performed an integrative DNA copy number and expression analysis of 77 invasive carcinomas, where all genes within the recurrent region were included. We selected eight genes, THOC7, PSMD6, SLC25A26, TMF1, RYBP, SHQ1, EBLN2, and GBE1, which were highly down-regulated in cases with loss, as confirmed at the protein level for RYBP and TMF1 by immunohistochemistry. The eight genes were subjected to network analysis based on the expression profiles, revealing interaction partners of proteins encoded by the genes that were coordinately regulated in tumours with loss. Several partners were shared among the eight genes, indicating crosstalk in their signalling. Gene ontology analysis showed enrichment of biological processes such as apoptosis, proliferation, and stress response in the network and suggested a relationship between down-regulation of the eight genes and activation of tumourigenic pathways. Survival analysis showed prognostic impact of the eight-gene signature that was confirmed in a validation cohort of 74 patients and was independent of clinical parameters. These results support the role of the eight candidate genes as targets of the 3p12-p14 loss in cervical cancer and suggest that the strong selection advantage of the loss during carcinogenesis might be caused by a synergetic effect of several tumourigenic processes controlled by these targets.

Methylation-mediated transcriptional repression of microRNAs during cervical carcinogenesis

Deregulated expression of microRNAs (miRNAs) is common and biologically relevant in cervical carcinogenesis and appears only partly related to chromosomal changes. We recently identified 32 miRNAs showing decreased expression in high-grade cervical intraepithelial neoplasia (CIN) and carcinomas not associated with a chromosomal loss, 6 of which were located within a CpG island. This study aimed to investigate to what extent these miRNAs are subject to DNA methylation-mediated transcriptional repression in cervical carcinogenesis.   Methylation-specific PCR (MSP) analysis on a cell line panel representing different stages of human papillomavirus (HPV) induced transformation revealed an increase in methylation of hsa-miR-149, -203 and -375 with progression to malignancy, whereas expression of these miRNAs was restored upon treatment with a demethylating agent. All three miRNAs showed significantly increased levels of methylation in cervical carcinomas, whereas methylation levels of hsa-miR-203 and -375 were also significantly increased in high-grade CIN. A pilot analysis showed that increased hsa-miR-203 methylation was also detectable in HPV-positive cervical scrapes of women with high-grade CIN compared with controls. Similar to recent findings on hsa-miR-375, ectopic expression of hsa-miR-203 in cervical cancer cells decreased both the proliferation rate and anchorage independent growth. We found evidence for methylation-mediated transcriptional repression of hsa-miR-149, -203 and -375 in cervical cancer. Methylation of the latter two was already apparent in precancerous lesions and represent functionally relevant events in HPV-mediated transformation. Increased hsa-miR-203 methylation was detectable in scrapes of women with high-grade CIN, indicating that methylated miRNAs may provide putative markers to assess the presence of (pre)cancerous lesions.

Focal aberrations indicate EYA2 and hsa-miR-375 as oncogene and tumor suppressor in cervical carcinogenesis

Cervical cancer results from persistent infection with high-risk human papillomavirus (hrHPV). Common genetic aberrations in cervical (pre)cancers encompass large genomic regions with numerous genes, hampering identification of driver genes. This study aimed to identify genes functionally involved in HPV-mediated transformation by analysis of focal aberrations (<3 Mb) in high-grade cervical intraepithelial neoplasia (hgCIN). Focal chromosomal aberrations were determined in high-resolution array comparative genomic hybridization data of 60 hgCIN. Genes located within focal aberrations were validated using 2 external gene expression datasets or qRT-PCR. Functional roles of candidate genes EYA2 (20q13) and hsa-miR-375 (2q35) were studied by siRNA-mediated knock-down and overexpression, respectively, in hrHPV-containing cell lines. We identified 74 focal aberrations encoding 305 genes. Concurrent altered expression in hgCIN and/or cervical carcinomas compared with normal cervical samples was shown for ATP13A3, HES1, OPA1, HRASLS, EYA2, ZMYND8, APOBEC2, and NCR2. Gene silencing of EYA2 significantly reduced viability, migratory capacity, and anchorage-independent growth of HPV16-transformed keratinocytes. For hsa-miR-375, a direct correlation between a (focal) loss and significantly reduced expression was found. Downregulation of hsa-miR-375 expression was confirmed in an independent series of cervical tissues. Ectopic expression of hsa-miR-375 in 2 cervical carcinoma cell lines reduced cellular viability. Our data provide a proof of concept that chromosomal aberrations are actively contributing to HPV-induced carcinogenesis and identify EYA2 and hsa-miR-375 as oncogene and tumor suppressor gene, respectively.

HPV type-related chromosomal profiles in high-grade cervical intraepithelial neoplasia

Background

The development of cervical cancer and its high-grade precursor lesions (Cervical Intraepithelial Neoplasia grade 2/3 [CIN2/3]) result from a persistent infection with high-risk human papillomavirus (hrHPV) types and the accumulation of (epi)genetic host cell aberrations. Epidemiological studies have demonstrated variable CIN2/3 and cancer risks between different hrHPV types. Recent genomic profiling studies revealed substantial heterogeneity in the chromosomal aberrations detected in morphologically indistinguishable CIN2/3 suggestive of varying cancer risk. The current study aimed to investigate whether CIN2/3 with different hrHPV types vary with respect to their chromosomal profiles, both in terms of the number of aberrations and chromosomal loci affected.

Methods

Chromosomal profiles were determined of 43 p16^INK4a-immunopositive CIN2/3 of women with long-term hrHPV infection (≥ 5 years). Sixteen lesions harboured HPV16, 3 HPV18, 14 HPV31, 1 HPV33, 4 HPV45, 1 HPV51, 2 HPV52 and 2 HPV58.

Results

Unsupervised hierarchical clustering analysis of the chromosomal profiles revealed two major clusters, characterised by either few or multiple chromosomal aberrations, respectively. A majority of 87.5% of lesions with HPV16 were in the cluster with relatively few aberrations, whereas no such unbalanced distribution was seen for lesions harbouring other hrHPV types. Analysis of the two most prevalent types (HPV16 and HPV31) in this data set revealed a three-fold increase in the number of losses in lesions with HPV31 compared to HPV16-positive lesions. In particular, losses at chromosomes 2q, 4p, 4q, 6p, 6q, 8q & 17p and gain at 1p & 1q were significantly more frequent in HPV31-positive lesions (FDR < 0.2).

Conclusions

Chromosomal aberrations in CIN2/3 are at least in part related to the hrHPV type present. The relatively low number of chromosomal aberrations observed in HPV16-positive CIN2/3 suggests that the development of these lesions is less dependent on genetic insult than those caused by other types like HPV31.

Chromosomal profiles of high-grade cervical intraepithelial neoplasia relate to duration of preceding high-risk human papillomavirus infection

High-grade cervical intraepithelial neoplasia (CIN2/3) represents a heterogeneous disease both with respect to clinical behavior and chromosomal aberrations detected. We hypothesized that the extent of chromosomal aberrations reflects the duration of their existence. Chromosomal profiles were determined of CIN3 of women with a known 5-year history of high-risk human papillomavirus virus (hrHPV) infection, in which duration of prior hrHPV infection was considered a proxy for duration of CIN3 existence. Eleven women had a <5 year preceding hrHPV infection (CIN3<5yrPHI) and 24 had a PHI lasting ≥5 years (CIN3≥5yrPHI). For comparison, six CIN3 adjacent to squamous cell carcinomas (CIN3-SCC), the corresponding SCCs, and six CIN1 were included. Unsupervised hierarchical clustering analysis of the chromosomal profiles revealed two clusters. One was characterized by a low number of chromosomal aberrations and included all CIN1, 81.8% of CIN3<5yrPHI and 33.3% of CIN3≥5yrPHI. Samples in the second cluster, displaying multiple aberrations, included 18.2% of CIN3<5yrPHI, 66.7% CIN3≥5yrPHI, all except one CIN3-SCC and all SCCs. The number of genomic aberrations increased according to lesion grade and also with longer duration of PHI. The increase in aberrations in CIN3≥5yrPHI compared to <5yrPHI was highly significant (p = 0.001), suggesting that CIN3≥5yrPHI represent more severe lesions. In conclusion, longer duration of preceding hrHPV infection is associated with an increased number of chromosomal aberrations. Hence, CIN3 with a longer duration of existence are likely more prone to have an increased short-term risk of cervical cancer.

DNA copy number alterations in endobronchial squamous metaplastic lesions predict lung cancer

RATIONALE

Autofluorescence bronchoscopy (AFB) is a valid strategy for detecting premalignant endobronchial lesions. However, no biomarker can reliably predict lung cancer risk of subjects with AFB-visualized premalignant lesions.

OBJECTIVES

The present study set out to identify AFB-visualized squamous metaplastic (SqM) lesions with malignant potential by DNA copy number profiling.

METHODS

Regular AFB examinations in 474 subjects at risk of lung cancer identified six subjects with SqM lesions at baseline, and carcinoma in situ or carcinoma (carcinoma in situ or greater) at the initial SqM site at follow-up bronchoscopy. These progressive SqM lesions were compared for immunostaining pattern and array comparative genomic hybridization-based chromosomal profiles with 23 SqM lesions of subjects who remained cancer-free. Specific DNA copy number alterations (CNAs) linked to cancer risk were identified and accuracy of CNAs to predict endobronchial cancer in this series was determined.

MEASUREMENTS AND MAIN RESULTS

At baseline, p53, p63, and Ki-67 immunostaining were not predictive for a differential clinical outcome of SqM lesions. The mean number of CNAs in baseline SqM of cases was significantly higher compared with control subjects (P < 0.01). Chromosomal regions significantly more frequently altered in SqM of cases were 3p26.3-p11.1, 3q26.2-q29, 9p13.3-p13.2, and 17p13.3-p11.2 (family-wise error rate <0.10). CNAs were specifically detected at the site of future cancer. In cases, baseline-detected CNAs persisted in subsequent biopsies taken from the initial site, and levels increased toward cancer progression. In this series, a model based on CNAs at 3p26.3-p11.1, 3q26.2-29, and 6p25.3-24.3 predicted cancer with 97% accuracy.

CONCLUSIONS

The data suggest that the presence of specific CNAs in SqM lesions predict endobronchial cancer.

Abstract 2718: Longer duration of preceding high-risk HPV infection is associated with an increase of chromosomal aberrations in high-grade cervical intraepithelial neoplasia

High-grade cervical intraepithelial neoplastic (CIN2/3) lesions represent a heterogeneous disease both with respect to clinical behaviour and chromosomal aberrations detected. We hypothesized that the extent of chromosomal aberrations reflects the duration of their existence. To test this hypothesis, chromosomal profiles were determined of CIN3 lesions of women with a known 5 year history of high-risk human papillomavirus virus (hrHPV) infection, in which duration of prior hrHPV infection was considered a proxy for duration of CIN3 lesion existence. Eight women had a &lt;5 year preceding hrHPV infection (CIN3&lt;5yrPHI) and 24 had a PHI lasting ≥5 years (CIN3≥5yrPHI). For comparison, 6 CIN3 lesions adjacent to squamous cell carcinomas (CIN3-SCC), their corresponding SCCs, and 6 CIN1 lesions were included. Unsupervised hierarchical clustering analysis of the chromosomal profiles revealed two clusters. One was characterized by relatively quiet genomic profiles including all CIN1, 75% of CIN3&lt;5yrPHI and 37.5% of CIN3≥5yrPHI. Samples in the second cluster, displaying multiple aberrations, included 62.5% CIN3≥5yrPHI and all except one CIN3-SCC and all SCCs. The number of genomic aberrations increased according to lesion grade and also with longer duration of PHI. The increase in aberrations in CIN3≥5yrPHI compared to &lt;5yrPHI was highly significant (p=0.002), suggesting that CIN3≥5yrPHI represent more severe lesions. In conclusion, a longer duration of preceding hrHPV infection is associated with an increased number of chromosomal aberrations. Hence, CIN3 lesions with a longer duration of existence are more prone to have an increased short-term risk of cervical cancer.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 2718. doi:10.1158/1538-7445.AM2011-2718

Combined CADM1 and MAL promoter methylation analysis to detect (pre-)malignant cervical lesions in high-risk HPV-positive women

Given the lower specificity for high-grade cervical lesions of high-risk human papillomavirus (hrHPV) testing compared to cytology, additional triage testing for hrHPV test-positive women is needed to detect high-grade cervical lesions. Here, we tested whether combined methylation analysis for cell adhesion molecule 1 (CADM1) and T-lymphocyte maturation associated protein (MAL), both functionally involved in cervical carcinogenesis, could serve as such a triage marker. Four quantitative methylation-specific PCRs (qMSP), two for CADM1 (regions M12 and M18) and MAL (regions M1 and M2) each, were applied to 261 cervical tissue specimens ranging from no neoplasia to carcinoma. When qMSPs were combined and positivity for at least one of the qMSPs in the combination was taken into account, the highest positivity rates for cervical intraepithelial neoplasia grade 3 (CIN3) lesions (97%) and squamous cell- and adeno-carcinomas (99%) were obtained by combining a single CADM1 marker with a single MAL marker. Subsequent qMSP analysis of 70 GP5+/6+-PCR hrHPV-positive scrapings revealed that a two-marker panel consisting of CADM1-M18 and MAL-M1 was most discriminative, detecting 90% of women with CIN3 (n = 30), whereas it showed a positive result in only 13.5% of women without cervical disease (n = 40). Finally, we applied hrHPV GP5+/6+-PCR testing followed by CADM1-M18/MAL-M1 methylation analysis to a cohort of 79 women visiting the outpatient colposcopy clinic. hrHPV testing revealed a sensitivity of 97% and a specificity of 33% for CIN3+. Additional CADM1-M18/MAL-M1 methylation analysis on the hrHPV-positive women increased the specificity to 78% with a sensitivity of 70%. In conclusion, the methylation marker panel CADM1-M18 and MAL-M1 may serve as an alternative molecular triage tool for hrHPV-positive women.

Methylation-mediated silencing and tumour suppressive function of hsa-miR-124 in cervical cancer

Background

A substantial number of microRNAs (miRNAs) is subject to epigenetic silencing in cancer. Although epigenetic silencing of tumour suppressor genes is an important feature of cervical cancer, little is known about epigenetic silencing of miRNAs. Since DNA methylation-based silencing of hsa-miR-124 occurs in various human cancers, we studied the frequency and functional effects of hsa-miR-124 methylation in cervical carcinogenesis.

Results

Quantitative MSP analysis of all 3 loci encoding the mature hsa-miR-124 (hsa-miR-124-1/-2/-3) showed methylation in cervical cancer cell lines SiHa, CaSki and HeLa as well as in late passages of human papillomavirus (HPV) type 16 or 18 immortalised keratinocytes. Treatment of SiHa cells with a demethylating agent reduced hsa-miR-124 methylation levels and induced hsa-miR-124 expression. In HPV-immortalised keratinocytes increased methylation levels were related to reduced hsa-miR-124 expression and higher mRNA expression of IGFBP7, a potential hsa-miR-124 target gene. Ectopic hsa-miR-124 expression in SiHa and CaSki cells decreased proliferation rates and migratory capacity. Combined hsa-miR-124-1 and/or hsa-miR-124-2 methylation analysis of 139 cervical tissue specimens showed an increasing methylation frequency from 0% in normal tissues up to 93% in cervical carcinomas. Increased methylation levels of hsa-miR-124-1 and hsa-miR-124-2 were significantly correlated with reduced hsa-miR-124 expression in cervical tissue specimens. Combined hsa-miR-124-1 and/or hsa-miR-124-2 methylation analysis of 43 cervical scrapes of high-risk HPV positive women was predictive of underlying high-grade lesions.

Conclusions

DNA methylation-based silencing of hsa-miR-124 is functionally involved in cervical carcinogenesis and may provide a valuable marker for improved detection of cervical cancer and its high-grade precursor lesions.

Genomic profiling identifies common HPV-associated chromosomal alterations in squamous cell carcinomas of cervix and head and neck

Background

It is well known that a persistent infection with high-risk human papillomavirus (hrHPV) is causally involved in the development of squamous cell carcinomas of the uterine cervix (CxSCCs) and a subset of SCCs of the head and neck (HNSCCs). The latter differ from hrHPV-negative HNSCCs at the clinical and molecular level.

Methods

To determine whether hrHPV-associated SCCs arising from different organs have specific chromosomal alterations in common, we compared genome-wide chromosomal profiles of 10 CxSCCs (all hrHPV-positive) with 12 hrHPV-positive HNSCCs and 30 hrHPV-negative HNSCCs. Potential organ-specific alterations and alterations shared by SCCs in general were investigated as well.

Results

Unsupervised hierarchical clustering resulted in one mainly hrHPV-positive and one mainly hrHPV-negative cluster. Interestingly, loss at 13q and gain at 20q were frequent in HPV-positive carcinomas of both origins, but uncommon in hrHPV-negative HNSCCs, indicating that these alterations are associated with hrHPV-mediated carcinogenesis. Within the group of hrHPV-positive carcinomas, HNSCCs more frequently showed gains of multiple regions at 8q whereas CxSCCs more often showed loss at 17p. Finally, gains at 3q24-29 and losses at 11q22.3-25 were frequent (>50%) in all sample groups.

Conclusion

In this study hrHPV-specific, organ-specific, and pan-SCC chromosomal alterations were identified. The existence of hrHPV-specific alterations in SCCs of different anatomical origin, suggests that these alterations are crucial for hrHPV-mediated carcinogenesis.

Integrated genomic and transcriptional profiling identifies chromosomal loci with altered gene expression in cervical cancer

For a better understanding of the consequences of recurrent chromosomal alterations in cervical carcinomas, we integrated genome-wide chromosomal and transcriptional profiles of 10 squamous cell carcinomas (SCCs), 5 adenocarcinomas (AdCAs) and 6 normal controls. Previous genomic profiling showed that gains at chromosome arms 1q, 3q, and 20q as well as losses at 8q, 10q, 11q, and 13q were common in cervical carcinomas. Altered regions spanned multiple megabases, and the extent to which expression of genes located there is affected remains unclear. Expression analysis of these previously chromosomally profiled carcinomas yielded 83 genes with significantly differential expression between carcinomas and normal epithelium. Application of differential gene locus mapping (DIGMAP) analysis and the array CGH expression integration tool (ACE-it) identified hotspots within large chromosomal alterations in which gene expression was altered as well. Chromosomal gains of the long arms of chromosome 1, 3, and 20 resulted in increased expression of genes located at 1q32.1-32.2, 3q13.32-23, 3q26.32-27.3, and 20q11.21-13.33, whereas a chromosomal loss of 11q22.3-25 was related to decreased expression of genes located in this region. Overexpression of DTX3L, PIK3R4, ATP2C1, and SLC25A36, all located at 3q21.1-23 and identified by DIGMAP, ACE-it or both, was confirmed in an independent validation sample set consisting of 12 SCCs and 13 normal ectocervical samples. In conclusion, integrated chromosomal and transcriptional profiling identified chromosomal hotspots at 1q, 3q, 11q, and 20q with altered gene expression within large commonly altered chromosomal regions in cervical cancer.

Association between dense CADM1 promoter methylation and reduced protein expression in high-grade CIN and cervical SCC

We previously showed that silencing of TSLC1, recently renamed CADM1, is functionally involved in high-risk HPV-mediated cervical carcinogenesis. CADM1 silencing often results from promoter methylation. Here, we determined the extent of CADM1 promoter methylation in cervical (pre)malignant lesions and its relation to anchorage-independent growth and gene silencing to select a CADM1-based methylation marker for identification of women at risk of cervical cancer. Methylation-specific PCRs targeting three regions within the CADM1 promoter were performed on high-risk HPV-containing cell lines, PBMCs, normal cervical smears, and (pre)malignant lesions. CADM1 protein expression in cervical tissues was analysed by immunohistochemistry. All statistical tests were two-sided. Density of methylation was associated with the degree of anchorage-independent growth and CADM1 gene silencing in vitro. In cervical squamous lesions, methylation frequency and density increased with severity of disease. Dense methylation (defined as >or= 2 methylated regions) increased from 5% in normal cervical samples to 30% in CIN3 lesions and 83% in squamous cell carcinomas (SCCs) and was significantly associated with decreased CADM1 protein expression (p < 0.00005). The frequency of dense methylation was significantly higher in >or= CIN3 compared with <or= CIN1 (p = 0.005), as well as in SCCs compared with adenocarcinomas (83% versus 23%; p = 0.002). Detection of dense CADM1 promoter methylation will contribute to the assembly of a valuable marker panel for the triage of high-risk HPV-positive women at risk of >or= CIN3.

Gene expression profiling to identify markers associated with deregulated hTERT in HPV-transformed keratinocytes and cervical cancer

Although high-risk human papillomavirus (HPV) infection plays a major role in the development of cervical cancer, additive oncogenic events are involved as well. One key event involves increased activity of telomerase resulting from a deregulated expression of its catalytic subunit hTERT. Our previous microcell-mediated chromosome transfer studies revealed that introduction of human chromosome 6 in the HPV16-immortalized keratinocyte cell line FK16A and in the HPV16-containing cervical cancer cell line SiHa induced growth arrest, resulting from a repression of hTERT mRNA expression and telomerase activity. Here, this model was used to analyze expression profiles associated with hTERT deregulation in HPV-transformed cells. Microarray expression analysis of 12 FK16A/chromosome 6 hybrids, 4 of which were negative for endogenous hTERT and 8 of which were positive for endogenous hTERT, resulted in the identification of 164 differentially expressed genes. Differential expression of a selection of 5 genes was verified by real-time RT-PCR. Of these 164 genes, 32 were also differentially expressed in other HPV transformed cells with deregulated hTERT. For 2 of these genes, encoding AQP3 and MGP, altered expression in hTERT positive cervical carcinomas was confirmed by real-time RT-PCR and immunohistochemistry, respectively. Moreover, increased MGP protein expression was significantly more frequent in high-grade cervical premalignant lesions with elevated hTERT mRNA expression compared to those without. In summary, we identified 32 candidate biomarkers for deregulated hTERT mRNA expression, which may enable the identification of cervical premalignant lesions that are at highest risk to progress to invasive cancer.

DNA copy number gains at loci of growth factors and their receptors in salivary gland adenoid cystic carcinoma

PURPOSE

Adenoid cystic carcinoma (ACC) is a malignant salivary gland tumor with a high mortality rate due to late, distant metastases. This study aimed at unraveling common genetic abnormalities associated with ACC. Additionally, chromosomal changes were correlated with patient characteristics and survival.

EXPERIMENTAL DESIGN

Microarray-based comparative genomic hybridization was done to a series of 18 paraffin-embedded primary ACCs using a genome-wide scanning BAC array.

RESULTS

A total of 238 aberrations were detected, representing more gains than losses (205 versus 33, respectively). Most frequent gains (>60%) were observed at 9q33.3-q34.3, 11q13.3, 11q23.3, 19p13.3-p13.11, 19q12-q13.43, 21q22.3, and 22q13.33. These loci harbor numerous growth factor [fibroblast growth factor (FGF) and platelet-derived growth factor (PDGF)] and growth factors receptor (FGFR3 and PDGFRbeta) genes. Gains at the FGF(R) regions occurred significantly more frequently in the recurred/metastasized ACCs compared with indolent ACCs. Furthermore, patients with 17 or more chromosomal aberrations had a significantly less favorable outcome than patients with fewer chromosomal aberrations (log-rank = 5.2; P = 0.02).

CONCLUSIONS

Frequent DNA copy number gains at loci of growth factors and their receptors suggest their involvement in ACC initiation and progression. Additionally, the presence of FGFR3 and PDGFRbeta in increased chromosomal regions suggests a possible role for autocrine stimulation in ACC tumorigenesis.

CGHcall: calling aberrations for array CGH tumor profiles

CGHcall achieves high calling accuracy for array CGH data by effective use of breakpoint information from segmentation and by inclusion of several biological concepts that are ignored by existing algorithms. The algorithm is validated for simulated and verified real array CGH data. By incorporating more than three classes, CGHcall improves detection of single copy gains and amplifications. Moreover, it allows effective inclusion of chromosome arm information.

AVAILABILITY

An R-package (GUI), a manual and an example data set are available at http://www.few.vu.nl/~mavdwiel/CGHcall.html.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Increased gene copy numbers at chromosome 20q are frequent in both squamous cell carcinomas and adenocarcinomas of the cervix

Genome-wide microarray-based comparative genomic hybridization (array CGH) was used to identify common chromosomal alterations involved in cervical carcinogenesis as a first step towards the discovery of novel biomarkers. The genomic profiles of nine squamous cell carcinomas (SCCs) and seven adenocarcinomas (AdCAs), as well as four human papillomavirus (HPV)-immortalized keratinocyte cell lines, were assessed. On a genome-wide scale, SCCs showed significantly more gains than AdCAs. More specifically, there was a striking and highly significant difference between the two histological types for gain at 3q12.1-28, which was predominantly observed in SCC. Other frequent alterations included gains of 1q21.1-31.1 and 20q11.21-13.33, and losses of 11q22.3-25 and 13q14.3-21.33. Subsequent FISH analysis for hTR, located at 3q26, confirmed the presence of 3q gain in SCCs and HPV-immortalized cell lines. Fine mapping of chromosome 20q using multiplex ligation-dependent probe amplification (MLPA) showed copy number increases for a number of genes located at 20q11-q12, including DNMT3B and TOP1. For DNMT3B, this correlated with elevated mRNA expression in 79% of cases. In conclusion, the assessment of frequent genomic alterations resulted in the identification of potential novel biomarkers, which may ultimately enable a better risk stratification of high-risk (hr)-HPV-positive women.

HPV-mediated transformation of the anogenital tract

Infection with high-risk human papillomavirus (HR-HPV) has been associated with intraepithelial neoplasia and carcinomas at various sites of the anogenital tract, including the cervix, vulva, vagina, penis and anus. Although HR-HPV is a necessary cause for cervical cancer, the majority of anal cancers and a subset of cancers at other genital sites, additional (epi)genetic events are required for malignant transformation. HPV-mediated transformation of human epithelial cells has been recognized as a multi-step process resulting from deregulated transcription of the viral oncogenes E6 and E7 in the proliferating cells. Interference of E6 and E7 with cell cycle regulators induces genetic instability, which drives the continuous selection of oncogenic alterations providing cells with a malignant phenotype. Early genetic events during cervical carcinogenesis associated with immortalization, include deletions at chromosomes 3p, 6 and 10p, whereas amongst others gain of chromosome 3q, loss of chromosome 11 and epigenetic alterations such as inactivation of the TSLC1 tumor suppressor gene represent later events associated with tumor invasion.