Cervical cancer develops from precancerous high-grade cervical intraepithelial neoplasia (CIN) harboring a transforming infection with high-risk human papillomavirus, which is characterized by p16INK4a overexpression. Once such a lesion has developed, progression toward an invasive squamous cell carcinoma (SCC) may take one or more decades, underlining the heterogeneity of these lesions in terms of duration of existence and progression risk. We performed array-based comparative genomic hybridization (array CGH) on 46 p16INK4a immunopositive CIN2/3 lesions to determine whether this heterogeneity is reflected in their chromosomal profiles. Chromosomal profiles of CIN2/3 lesions were related to those of invasive cervical SCC and promoter methylation of CADM1, a tumor suppressor gene known to be functionally involved in the tumorigenic phenotype of cervical cancer cells. Frequent alterations found in CIN2/3 lesions included gains located at chromosome 1, 3, 7, and 20 and losses located at 4, 11, 16, 17, and 19. Unsupervised hierarchical clustering identified two subsets of CIN2/3 lesions, chromosomal profiles of one of which closely resembled invasive SCCs. Gains of 1, 3q, and 20 were characteristic for CIN2/3 lesions with chromosomal signatures resembling carcinomas. In addition, dense promoter methylation of the CADM1 gene was significantly more frequent in these CIN2/3 lesions (P = 0.004). No chromosomal alterations were detected in six CIN1 lesions, five of which were completely p16INK4a immunonegative. These findings suggest that biomarkers associated with gains at chromosomes 1, 3q, and 20 are potential hallmarks of advanced p16INK4a-positive CIN2/3 lesions with a high short-term risk of progression. [Cancer Res 2009;69(2):647–55]

Cervical cancer is the second most common malignancy in women worldwide and is caused by a persistent infection with high-risk (i.e., oncogenic) human papillomavirus (hrHPV) types. The majority (∼80%) of cervical carcinomas are squamous cell carcinomas (SCC; refs. 1, 2). Cervical SCCs develop through different stages of premalignant disease, called cervical intraepithelial neoplasia (CIN), graded 1 to 3. To reflect their relative risk of progression to cervical cancer, CIN1 lesions are nowadays often called low-grade CIN, whereas CIN2 and 3 lesions together are considered high-grade CIN. It is current practice to treat all high-grade CIN lesions (CIN2/3) radically to prevent cervical SCC development, despite the fact that only a subset of these lesions displays a short-term progression risk for invasive cancer (3). Treatment of CIN lesions can cause preterm delivery, underlining the necessity of careful consideration of the moment of therapeutic intervention in women of reproductive age (4).

In CIN lesions, generally two distinct patterns of hrHPV infections can be recognized. A productive infection, which is mostly associated with CIN1 lesions, is characterized by the occurrence of most viral activity and vegetative viral replication in differentiating, nondividing suprabasal epithelial cells, finally resulting in the formation of new viral particles in the upper layers of the epithelium. In CIN2/3 lesions, a different pattern of hrHPV infection is frequently observed, in which the viral life cycle is aborted and no new viruses are produced. This condition, which can be called an abortive or transforming infection, is characterized by aberrant expression of the viral oncogenes E6 and E7 in proliferating cell layers (5, 6). E6 and E7-mediated degradation of the tumor suppressor genes p53 and pRb in these dividing cells triggers uncontrolled proliferation and subsequent chromosomal instability. Khleif and colleagues (7) reported that inactivation of pRb by mutations or viral oncogenes results in increased expression of its transcriptional target p16INK4a. Because, based on these data, p16INK4a overexpression can be considered a consequence of deregulated E6/E7 expression, overexpression of p16INK4a potentially is an attractive marker of transforming hrHPV infections. In support of this, the percentage of lesions with diffuse p16INK4a immunostaining in at least the basal and parabasal cells increases proportional to the severity of CIN lesions (811). Moreover, CIN1 lesions with p16INK4a overexpression were shown to have a higher risk of persistence and progression to high-grade lesions (12). On the other hand, expression of the viral capsid protein L1, which marks productive infections, was recently shown to decrease with increasing severity of the lesion (13). Overall, sufficient evidence has now been collected in favor of the concept that p16INK4a overexpression is a marker for CIN lesions that harbor a transforming hrHPV infection and consequently should be considered the true precursor lesions of cervical cancer (14, 15).

Whereas CIN2/3 lesions can develop rather fast after hrHPV infection (2–3 years), subsequent development of an invasive carcinoma may take at least one or more decades (3, 1618), indicating that even CIN2/3 lesions harboring a transforming hrHPV infection represent a heterogeneous disease in terms of duration of existence and progression risk. This is supported by the fact that some features common to cervical cancer cells, such as hTERT overexpression and CADM1 promoter methylation are only found in a subset of CIN2/3 lesions (19, 20). It is likely that, once a CIN2/3 lesion has developed, several (epi)genetic alterations have to accumulate for progression to invasive cancer. We hypothesize that chromosomal profiling of p16INK4a-positive CIN2/3 lesions will enable distinction of advanced lesions, which already have acquired a number of chromosomal alterations needed for progression, and consequently have the highest short-term progression risk.

In this study, high-resolution array-based CGH (array CGH) analysis was performed on 46 microdissected frozen CIN2/3 lesions unrelated to carcinomas, all of which contained hrHPV and were p16INK4a-positive. To estimate the invasive potential of CIN2/3 lesions included in this study independently from their genomic profiles, CADM1 promoter methylation was determined. CADM1 is a tumor suppressor gene found to be functionally involved in the late stages of cervical carcinogenesis (21). Re-expression of CADM1 suppressed the tumorigenic but not the immortal phenotype of the SiHa cervical cancer cell line. Chromosomal profiles of CIN2/3 lesions were compared with those of cervical carcinomas using unsupervised hierarchical clustering and results were related to CADM1 promoter methylation (22). In addition, six hrHPV-positive low-grade CIN1 lesions, potentially representing productive HPV infections, were analyzed.

Tissue specimens. We used both frozen and paraffin-embedded biopsy specimens of 46 CIN2/3 lesions and six CIN1 lesions unrelated to cervical SCC, all of which were collected at the Departments of Obstetrics and Gynecology of the VU University Medical Center and the Erasmus Medical Center during routine clinical practice. The mean age of women included in this study was 33.3, ranging from 22 till 54 y. Only samples that contained hrHPV DNA, as determined by a general primer GP5+/GP6+-mediated PCR-enzyme immunoassay method using a probe cocktail of 14 hrHPV types (types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68), were included (23). Presence of CIN1 or CIN2/3 in the frozen specimens used for array CGH experiments was independently validated by two experienced pathologists (FJvK and CJLMM). Because reproducibility of CIN2 versus CIN3 classification on formalin-fixed, paraffin-embedded specimens is not >70%, and distinction between both lesion grades is rather inaccurate on frozen specimens we felt that a classification of high-grade (CIN2/3) and low-grade (CIN1) lesions was more reliable (24, 25). However, to ensure inclusion of only CIN2/3 lesions harboring transforming infections, only p16INK4a-positive CIN2/3 lesions were analyzed (14, 15).

This study followed the ethical guidelines of the Institutional Review Board of the VU University Medical Center and written informed consent was obtained from all patients included.

Immunohistochemistry. Immunohistochemical staining of p16INK4a was performed on 4-μm sections of paraffin-embedded specimens of CIN1 and CIN2/3 lesions. Sections were incubated with 0.3% H2O2 in methanol for 30 min to inactivate endogenous peroxidase. Antigen retrieval was performed by heat treatment in a microwave with 10 mmol/L citrate buffer (pH 6.0). Slides were incubated with mouse monoclonal antibody for p16INK4a (MTM Laboratories) for 1 h at room temperature. For antibody detection, the EnVision horseradish peroxidase system (Dako) was used. Sections were counterstained with Mayers' hematoxylin for 1 min. Slides were examined by an experienced pathologist (MCGB) and were considered positive if diffuse staining of all dysplastic cells (nucleus and cytoplasm) was observed.

Microdissection and extraction of nucleic acids. To avoid contamination of the array CGH profile with DNA derived from surrounding normal tissue, frozen tissue specimens of CIN1 and CIN2/3 lesions were laser capture microdissected using a Leica ASLMD microscope (Leica). Ten-micrometer-thick cryosections were cut and mounted on PEN foil–coated slides (Leica). Sections were stained with Mayers' hematoxylin for 1 min and completely dehydrated. Genomic DNA of microdissected tissue was extracted by proteinase K digestion followed by standard phenol-chloroform extraction (26).

CADM1 methylation-specific PCR. Methylation of the CADM1 promoter region was determined by Methylation-Specific PCR (MSP) on sodium-bisulfite–modified genomic DNA (19). In short, DNA was modified using the EZ DNA Methylation kit (Zymo Research), according to the manufacturer's instructions. MSPs targeting 3 regions within the CADM1 promoter (M1, −696 to −582; M5, −415 to −258; and M9, −61 to +39 relative to the ATG) were performed using primer sets described by Overmeer and colleagues (19). Detection of PCR products was performed by reverse line blot hybridizations to probes specific for methylated and unmethylated DNA. All samples were tested in duplicate for all three regions, and only samples that showed methylation of a certain region in both experiments were considered positive for that region.

CGH microarrays. Genomic DNA of CIN2/3 lesions was hybridized on CGH BAC microarrays (5K) produced at the Microarray facility of the VU Medical Center4

against a reference pool of DNA from 10 healthy male individuals. The use of a male reference pool enabled us to determine whether the hybridization had succeeded, based on the expected gain of chromosome X and loss of chromosome Y in our female test samples. In total, 4632 BAC clones, including the 1Mb resolution Sanger BAC clone set5 and a subset of clones from the Children's Hospital Oakland Research Institute, with equal distribution across the genome were spotted in triplicate. Hybridizations were essentially performed as described by Snijders and colleagues (27). Both prehybridization and hybridization were performed in a hybridization station (HybStation12; Perkin-Elmer Life Sciences). Hybridized arrays were scanned using a G2505B scanner (Agilent). Spots were quantified using ImaGene 5.6.1 software (BioDiscovery Ltd) with default settings for the flagging of bad quality spots.

Genomic DNA of CIN1 lesions was hybridized on CGH oligo microarrays (44K) produced by Agilent following the manufacturer's protocol against the same normal reference pool as used for the CIN2/3 lesions. Quantification of these arrays was done using Feature Extraction software version 9.5.1 (Agilent).

The entire data set described here is available from the Gene Expression Omnibus6

through series accession number GSE11573.

Array CGH analysis. For CIN2/3 lesions, BAC clones were positioned along the genome according to the May 2004 freeze. After exclusion of clones with one or more flagged spots, the average of the triplicate spots was calculated for each BAC clone. Log2 ratios were normalized per spotted subarray by subtraction of the median value of all BAC clones spotted within that subarray. For CIN1 lesions, arrays were also normalized by median subtraction. Segmentation and subsequent calling of gained and lost regions was done using CGHCall, a recently described automated calling algorithm. Segments with a probability score of ≥0.5 were considered gained or lost (28).

Reduction of data set into chromosomal regions. We used the CGH regions algorithm to reduce our data set to chromosomal regions, accepting maximally 0.1% information loss (threshold, 0.001; ref. 29). It was shown by Van de Wiel and colleagues (29) that the use of regions instead of single BAC clones improves the effectiveness of subsequent statistical analyses and facilitates interpretation of the results.

Clustering analysis. Unsupervised hierarchical clustering of CIN2/3 lesions together with SCCs was performed. Weighted Clustering of Called aCGH data (WECCA) was applied to the region data (defined as described above; ref. 30). To reflect differences in the regions' coverage of the genome, we chose a region's weight equal to the number of clones it contains. The weight of a region determines its relative influence on the clustering. WECCA builds a dendrogram with the concordance similarity and total linkage. The latter results in compact, well-separated clusters. Finally, to determine the regions that are most distinct between clusters, we calculated (per region) the maximum pairwise symmetrized Kullback Leibler divergence. Regions with the highest divergence are most distinct between clusters.

Statistical analysis. The number of altered features was compared between CIN1 and CIN2/3 lesions and CIN2/3 lesions and carcinomas, respectively, using the Wilcoxon rank test. The number of samples showing dense CADM1 promoter methylation was compared between the clusters using the χ2 test. This test was also applied to determine whether an association between cluster assignment and histologic grading was present. Two-sided P values of <0.05 were considered statistically significant.

Chromosomal profiling identifies frequent gains and losses. Chromosomal profiling was performed on a well-defined set of 46 CIN2/3 lesions, all of which harbored a transforming hrHPV infection. A transforming infection was reflected by the presence of hrHPV DNA as detected by GP5+/6+ PCR and a diffuse immunostaining pattern for p16INK4a in the dysplastic areas, with both cytoplasmic and nuclear involvement (Fig. 1).

Figure 1.

Representative examples of p16INK4a immunostaining results in CIN2/3 lesions and CIN1 lesions. Positive results for CIN2/3 lesions HGCIN24 and HGCIN39 are shown in A and B, respectively. In C and D, a positive CIN1 lesion, LGCIN1, and a negative CIN1 lesion, LGCIN6, are depicted, respectively. All pictures were taken using a ×10 magnification.

Figure 1.

Representative examples of p16INK4a immunostaining results in CIN2/3 lesions and CIN1 lesions. Positive results for CIN2/3 lesions HGCIN24 and HGCIN39 are shown in A and B, respectively. In C and D, a positive CIN1 lesion, LGCIN1, and a negative CIN1 lesion, LGCIN6, are depicted, respectively. All pictures were taken using a ×10 magnification.

Close modal

Genomic profiles of the 46 p16INK4a-positive CIN2/3 lesions were determined using 5K arrays in combination with frozen microdissected material and a highly accurate calling algorithm. On average, 7.2 alterations (range, 0–20) were found per sample, with a mean number of 2.8 gains (range, 0–10) and 4.4 losses (range, 0–15). Only 2 of 46 CIN2/3 lesions showed no alterations. In comparison, previous chromosomal profiling of invasive cervical SCCs using the same platform showed on average 13.3 (range, 6–28) alterations, with a mean number of 5.7 gains (range, 3–9) and 7.6 losses (range, 2–23), which was significantly higher than in the CIN2/3 lesions (P = 0.012; ref. 22). To determine common chromosomal alterations, the frequency of gains and losses per BAC clone was plotted for all CIN2/3 lesions (Fig. 2). Frequent alterations (>20% of cases) included gains on chromosome 1, 3, 7, and 20 and losses on chromosome 4, 11, 16, 17, and 19. Smallest regions of overlap (SRO) were determined for the frequently altered regions by alignment of all samples (Table 1). Frequent alterations in carcinomas also included gains on chromosome 1, 3, and 20. However, no overlap was found when comparing the frequent losses in CIN2/3 lesions to those in SCCs (22).

Figure 2.

Frequency plot of gains and losses for chromosome 1 to 22 as determined by array CGH in 46 CIN2/3 lesions. Percentages of gains (positive axis) and losses (negative axis) are shown for each BAC clone in all CIN2/3 lesions analyzed.

Figure 2.

Frequency plot of gains and losses for chromosome 1 to 22 as determined by array CGH in 46 CIN2/3 lesions. Percentages of gains (positive axis) and losses (negative axis) are shown for each BAC clone in all CIN2/3 lesions analyzed.

Close modal
Table 1.

Percentages of frequently altered regions in CIN2/3 lesions

SROGain/loss% (n)
1p36.11-p35.2 Gain 21.7 (10) 
1p31.3-p21.1 Gain 26.1 (12) 
1q25.3-q32.1 Gain 23.9 (11) 
1q32.2-q44 Gain 23.9 (11) 
3p26.3-p26.1 Gain 21.7 (10) 
3p14.3-p14.2 Gain 21.7 (10) 
3q11.2-q29 Gain 26.1 (12) 
7q31.1-q31.2 Gain 28.3 (13) 
20p13-p11.21 Gain 26.1 (12) 
20q12 Gain 32.6 (15) 
4p16.3-p16.1 Loss 32.6 (15) 
4q31.21 Loss 21.7 (10) 
4q35.2 Loss 26.1 (12) 
11q13.3 Loss 28.3 (13) 
16q24.2-q24.3 Loss 21.7 (10) 
17p13.3 Loss 34.8 (16) 
17q25.3 Loss 41.3 (19) 
19p13.11-q12 Loss 32.6 (15) 
SROGain/loss% (n)
1p36.11-p35.2 Gain 21.7 (10) 
1p31.3-p21.1 Gain 26.1 (12) 
1q25.3-q32.1 Gain 23.9 (11) 
1q32.2-q44 Gain 23.9 (11) 
3p26.3-p26.1 Gain 21.7 (10) 
3p14.3-p14.2 Gain 21.7 (10) 
3q11.2-q29 Gain 26.1 (12) 
7q31.1-q31.2 Gain 28.3 (13) 
20p13-p11.21 Gain 26.1 (12) 
20q12 Gain 32.6 (15) 
4p16.3-p16.1 Loss 32.6 (15) 
4q31.21 Loss 21.7 (10) 
4q35.2 Loss 26.1 (12) 
11q13.3 Loss 28.3 (13) 
16q24.2-q24.3 Loss 21.7 (10) 
17p13.3 Loss 34.8 (16) 
17q25.3 Loss 41.3 (19) 
19p13.11-q12 Loss 32.6 (15) 

NOTE: Regions depicted here are the SRO between samples and are altered in >20% of cases.

Carcinoma-like chromosomal signatures in a subset of CIN2/3 lesions. To directly relate the chromosomal profiles detected in CIN2/3 lesions to those previously found in cervical SCCs, unsupervised hierarchical clustering was performed on the combined data of CIN2/3 lesions and cervical SCCs (22). This resulted in two distinct clusters, one of which contained only CIN2/3 lesions with a relatively low number of alterations (Fig. 3, cluster 1). The other cluster was formed by all SCCs together with 13 CIN2/3 lesions (Fig. 3, cluster 2). The fact that genomic profiles of this subset of CIN2/3 lesions are similar to those of invasive SCCs, suggests that these lesions are more advanced than the ones in cluster 1. The p16INK4a staining pattern was similar in lesions from cluster 1 compared with cluster 2. As an example, CIN2/3 lesion HGCIN24, shown in Fig. 1A, belongs to cluster 1, whereas CIN2/3 lesion HGCIN39 in Fig. 1B belongs to cluster 2. CIN2 and CIN3 lesions represent separate histologic grades of cervical premalignant disease, of which CIN3 lesions are considered the closest precursor lesions to invasive cervical cancer. Therefore, we tested whether histologic grading was associated with our clustering results. Cluster 2 contained 4 of the 15 CIN2 lesions (27%) and 9 of the 31 CIN3 lesions (29%), which was not statistically significantly different (P = 0.87). This indicates that p16INK4a-positive CIN2 and CIN3 lesions cannot be distinguished based on chromosomal alterations.

Figure 3.

Result of unsupervised hierarchical clustering of 46 CIN2/3 lesions together with 10 SCCs (gray). In the heatmap, cases are ordered along the X-axis and genomic loci along the Y-axis marked by chromosome number. Red, losses; green, gains.

Figure 3.

Result of unsupervised hierarchical clustering of 46 CIN2/3 lesions together with 10 SCCs (gray). In the heatmap, cases are ordered along the X-axis and genomic loci along the Y-axis marked by chromosome number. Red, losses; green, gains.

Close modal

To determine which chromosomal regions were most distinct between the two clusters and may therefore be linked to malignant transformation, the maximum pair wise symmetrized Kullback Leibler divergence between the clusters was calculated for each chromosomal region included in the clustering analysis. As is shown in Fig. 4A, 1p36.32-31.2, 1p12-q44, 3q11.2-29, and 20p13-q13.33 are most distinct between clusters. These regions are more frequently gained in cluster 2 (containing SCCs and CIN2/3 lesions) compared with the CIN2/3 lesions in the other cluster (Fig. 3). To ensure that this difference is not largely based on genomic profiles of the SCCs, we performed the same analysis on CIN2/3 lesions in cluster 2 and CIN2/3 lesions in cluster 1 (excluding the carcinomas). The most striking differences were again located at chromosomes 1p36.32-q44, 3q11.2-29, and 20p13-q13.33 (Fig. 4B), indicating that gains of these regions might be related to malignant transformation. Overall, genomic profiles of CIN2/3 lesions with a gain of at least one of these regions contained higher numbers of alterations. More specifically, the mean number of gains (excluding gains of 1, 3q, and 20) was significantly higher in these CIN2/3 lesions compared with CIN2/3 lesions without any of these gains (3.1 versus 0.84, respectively; P = 0.003). The number of losses did not differ between these two groups of CIN2/3 lesions. To further validate our findings that certain chromosomal alterations are specific for high-grade lesions with invasive potential, we subsequently analyzed six hrHPV-positive CIN1 lesions. Five of these lesions were p16INK4a-immunonegative, indicating the absence of a transforming hrHPV infection in these lesions (Fig. 1C). In one lesion, clear p16INK4a staining was observed in the lower one third of the epithelium, suggesting that the hrHPV infection has already become transforming (Fig. 1D). Chromosomal profiling using arrays that have a higher resolution than those used for the CIN2/3 lesions showed that virtually no chromosomal alterations were present in any of the CIN1 lesions. The average percentage of features with alterations as determined by the CGHCall algorithm was significantly lower in CIN1 lesions compared with CIN2/3 lesions (0.97% versus 15.58%, respectively; P = 0.003). When including only CIN2/3 lesions from cluster 2, the difference was even more significant (0.97% versus 40.1%, respectively; P < 0.001). Together, these results indicate that genomic profiling enables further classification of histologically similar lesions.

Figure 4.

The maximum pair wise symmetrized Kullback Leibler divergence, a measure for the distance between clusters, is shown for all chromosomal regions between (A) all samples in cluster 1 compared with cluster 2 and (B) only CIN2/3 lesions in cluster 1 compared with CIN2/3 lesions in cluster 2.

Figure 4.

The maximum pair wise symmetrized Kullback Leibler divergence, a measure for the distance between clusters, is shown for all chromosomal regions between (A) all samples in cluster 1 compared with cluster 2 and (B) only CIN2/3 lesions in cluster 1 compared with CIN2/3 lesions in cluster 2.

Close modal

CADM1 promoter methylation. To validate that the CIN2/3 lesions clustering together with the SCCs were more advanced compared with the CIN2/3 lesions in the other cluster, CADM1 promoter methylation for 3 different regions within the CADM1 promoter was determined in all CIN2/3 lesions. Dense methylation (i.e., methylation at two or more sites) was previously shown to be associated with gene silencing, a feature that has been linked to late stages of HPV-mediated transformation (19, 21). As is shown in Fig. 5A, dense methylation was found in 17 CIN2/3 lesions (37%), whereas 19 CIN2/3 lesions were completely unmethylated. In comparison to this, 8 of the 10 cervical SCCs showed dense methylation of the CADM1 promoter. These results are concordant with previous data, showing dense methylation of the CADM1 promoter in 30% of CIN2/3 lesions and 83% of SCCs (19). We found significantly more dense CADM1 methylation in CIN2/3 lesions with genomic profiles similar to SCCs (cluster 2), compared with the CIN2/3 lesions in cluster 1 (21% versus 77%, respectively; P = 0.004; Fig. 5B).

Figure 5.

A, MSP results in all CIN2/3 lesions for 3 regions (M1, M5, and M9) within the CADM1 promoter. Black boxes, methylation. The presence of dense (≥2 regions) methylation is indicated by black boxes in the bottom row. In the last column, percentages of methylation are shown for the separate regions as well as for dense methylation. B, dense CADM1 promoter methylation (≥2 regions) is shown in relation to the unsupervised hierarchical clustering results. Black boxes, dense methylation.

Figure 5.

A, MSP results in all CIN2/3 lesions for 3 regions (M1, M5, and M9) within the CADM1 promoter. Black boxes, methylation. The presence of dense (≥2 regions) methylation is indicated by black boxes in the bottom row. In the last column, percentages of methylation are shown for the separate regions as well as for dense methylation. B, dense CADM1 promoter methylation (≥2 regions) is shown in relation to the unsupervised hierarchical clustering results. Black boxes, dense methylation.

Close modal

In this study, we performed array CGH analysis of histologically similar CIN2/3 lesions showing p16INK4a overexpression, to determine whether their potential heterogeneity in terms of duration of existence, and most likely, short-term progression risk is reflected in their genomic signature. The obtained chromosomal profiles were related to genomic profiles of invasive cervical SCCs and CADM1 promoter methylation.

To date, only a limited number of studies investigating chromosomal profiles of CIN lesions have been performed, using classic comparative genomic hybridization (CGH; refs. 3136). Whereas the focus of previous studies was primarily to compare CIN lesions of different histologic grades, the present study aimed at the detection of potential genetic differences within a group of well-defined histologically similar p16INK4a-expressing CIN2/3 lesions, which should be considered the immediate precursor lesions of cervical cancer. Because for ethical reasons all CIN2/3 lesions are radically treated, it is not possible to determine the natural clinical course of the lesions included in this study. CADM1 promoter methylation was therefore included in this study as an independent surrogate marker for CIN2/3 lesions with invasive potential.

Unsupervised clustering analysis identified a subset of CIN2/3 lesions with chromosomal profiles closely resembling those of invasive SCCs. Dense CADM1 promoter methylation, a marker of lesions with invasive potential, was significantly more frequent in this subset compared with the other CIN2/3 lesions. Together, these results strongly suggest that the p16INK4a-positive high-grade cervical lesions analyzed comprise both early and more advanced lesions, probably reflecting the age of the lesion. The notion that long existing lesions are characterized by higher numbers of nonrandom chromosomal alterations is supported by the fact that late passages of HPV-immortalized keratinocytes showed higher numbers of alterations compared with early passages of the same cell lines (37). We previously showed that HPV-induced transformation of keratinocytes in vitro accurately reflects cervical carcinogenesis both genotypically and phenotypically (38).

In this study, we found a median of five alterations per CIN2/3 lesion. Fewer alterations were detected in premalignant cervical lesions by Heselmeyer and colleagues (32), Kirchhoff and colleagues (33, 34), and Umayahara and colleagues (36). However, these studies were performed by classic CGH while using formalin-fixed paraffin-embedded material, which often results in DNA of poor quality, and microdissection was not always performed.

CIN2/3 lesions resembling SCCs were characterized by gains of chromosome 1, 3q, and 20, indicating that these alterations may be associated with malignant transformation. The fact that CIN1 lesions did not, even using a platform with a higher resolution, reveal any marked chromosomal alterations further supports the idea that these alterations are specific for a subset of high-grade lesions harboring a transforming hrHPV infection. In a study of Alazawi and colleagues (31), comparing CIN lesions of different grades, gains of 1q, 3q, and 20q, were more frequent in high-grade lesions compared with low-grade lesions of the cervix. Based on the frequency of these alterations, in CIN lesions studied by Alazawi and colleagues (31) and us, gains of chromosome 20q are likely to precede gains of chromosomes 1 and 3q.

Low-level gains of chromosome 20q were previously shown to be associated with hrHPV E7–induced immortalization of human epithelial cells (39, 40). An interesting gene located at 20q12, the SRO in this study, is TOP1, which controls and alters the topological state of DNA during transcription. In cancer, TOP1 is thought to mediate illegitimate recombination leading to genomic instability (41). Using multiplex ligation–mediated probe amplification, we previously found increased copy numbers for this gene in 15 of 15 cervical carcinomas (22). Frequent copy number gains of TOP1 were also found in colorectal cancer using the same technique (42).

Gains of chromosome 1q are frequently observed in cervical cancer as well as a number of other solid tumors, including breast and bladder carcinomas (43, 44). A number of studies, using fluorescence in situ hybridization and/or karyotyping, investigated both numerical and structural abnormalities of chromosome 1 in cervical carcinomas and its precursor lesions (4548). These studies showed that genomic alterations of chromosome 1 are restricted to high-grade dysplasia and carcinoma of the cervix.

A gain of chromosome 3q is among the most frequently described chromosomal alterations in cervical cancer and has previously been implied to define progression from high-grade dysplasia toward invasive carcinoma of the cervix (32). Extra copies of hTR located at 3q can be detected in up to 76% of high-grade premalignant lesions, and genomic amplification of hTR detected in Pap smears has a predictive value for the development of cervical cancer (49, 50). We have shown that gains of 3q are more specific for the squamous cell histotype (22) and identified a number of novel candidate oncogenes, including DTX3L, PIK3R4, SLC25A36, and ATP2C1 (51).

Although the overall signature of chromosomal gains was quite similar between CIN2/3 lesions and SCCs, a number of interesting differences were observed. These included a narrow gain at chromosome 7q31.1-31.2 in CIN2/3 lesions. Within this region, three known (RefSeq) genes are located, FOXP2, MDFIC, and TFEC, all of which are transcription factors. Interestingly, MDFIC is a member of a protein family characterized by a specific cysteine-rich COOH-terminal domain, which is involved in transcriptional regulation of viral genomes (52). We were surprised to notice the lack of overlap between lost chromosomal regions in CIN2/3 lesions and SCCs. A number of other studies investigating lung, colorectal, and breast cancer postulated the concept of chromosome loss and subsequent reduplication in epithelial carcinogenesis (5356). This concept implies that, although losses detected in premalignant lesions are no longer present in invasive carcinomas, tumor suppressor genes located within these previously lost regions are still silenced through other mechanisms in more advanced stages of disease, such as allelic loss and potentially epigenetic modifications, all of which may remain undetected by array CGH.

In conclusion, this study showed that genomic profiling of p16INK4a-positive CIN2/3 lesions can distinguish histologically similar high-grade CIN lesions into potentially early and more advanced lesions. This distinction, based on gains of 1, 3q, and 20, probably reflects the age of the lesion and may therefore aid in the identification of lesions with invasive potential and the highest short-term risk of progression. Identification of CIN2/3 lesions with high short-term risk of progression to invasive carcinoma is important because they need immediate treatment in contrast to early CIN2/3 lesions. Identification of biomarkers based on the chromosomal alterations found in this study can therefore raise the efficiency of the current cervical screening programs by reducing the adverse effects associated with treatment of early CIN2/3 lesions with potentially a low short-term progression risk.

No potential conflicts of interest were disclosed.

Grant support: Centre for Medical Systems Biology in the framework of the Netherlands Genomic Initiative. Grant Koningin Wilhelmina Fonds VU2002-2618 contributed to establishing the BAC array platform.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Raghuram Ambadipudi, Clarissa Kooi, Debbie Claassen-Kramer, Renée Overmeer, and Paul Eijk for excellent technical assistance, and The Mapping Core and Map Finishing groups of the Wellcome Trust Sanger Institute for initial clone supply and verification.

1
Fu YS, Reagan JW. Pathology of the uterine cervix, vagina and vulva. Philadelphia: W.B. Saunders and Co.; 1989. p. 288–335.
2
Pisani P, Bray F, Parkin DM. Estimates of the world-wide prevalence of cancer for 25 sites in the adult population.
Int J Cancer
2002
;
97
:
72
–81.
3
McCredie MR, Sharples KJ, Paul C, et al. Natural history of cervical neoplasia and risk of invasive cancer in women with cervical intraepithelial neoplasia 3: a retrospective cohort study.
Lancet Oncol
2008
;
9
:
425
–34.
4
Sadler L, Saftlas A, Wang W, Exeter M, Whittaker J, McCowan L. Treatment for cervical intraepithelial neoplasia and risk of preterm delivery.
JAMA
2004
;
291
:
2100
–6.
5
Doorbar J. Molecular biology of human papillomavirus infection and cervical cancer.
Clin Sci (Lond)
2006
;
110
:
525
–41.
6
Snijders PJ, Steenbergen RD, Heideman DA, Meijer CJ. HPV-mediated cervical carcinogenesis: concepts and clinical implications.
J Pathol
2006
;
208
:
152
–64.
7
Khleif SN, DeGregori J, Yee CL, et al. Inhibition of cyclin D-CDK4/CDK6 activity is associated with an E2F-mediated induction of cyclin kinase inhibitor activity.
Proc Natl Acad Sci U S A
1996
;
93
:
4350
–4.
8
Gurrola-Diaz CM, Suarez-Rincon AE, Vazquez-Camacho G, et al. P16(INK4a) immunohistochemistry improves the reproducibility of the histological diagnosis of cervical intraepithelial neoplasia in cone biopsies.
Gynecol Oncol
2008
;
111
:
120
–4.
9
Klaes R, Friedrich T, Spitkovsky D, et al. Overexpression of p16(INK4A) as a specific marker for dysplastic and neoplastic epithelial cells of the cervix uteri.
Int J Cancer
2001
;
92
:
276
–84.
10
Lambert AP, Anschau F, Schmitt VM. p16INK4A expression in cervical premalignant and malignant lesions.
Exp Mol Pathol
2006
;
80
:
192
–6.
11
Queiroz C, Silva TC, Alves VA, et al. P16(INK4a) expression as a potential prognostic marker in cervical pre-neoplastic and neoplastic lesions.
Pathol Res Pract
2006
;
202
:
77
–83.
12
Negri G, Vittadello F, Romano F, et al. p16INK4a expression and progression risk of low-grade intraepithelial neoplasia of the cervix uteri.
Virchows Arch
2004
;
445
:
616
–20.
13
Negri G, Bellisano G, Zannoni GF, et al. p16ink4a and HPV L1 immunohistochemistry is helpful for estimating the behavior of low-grade dysplastic lesions of the cervix uteri.
Am J Surg Pathol
2008
;
32
:
1715
–20.
14
Wentzensen N, von Knebel DM. Biomarkers in cervical cancer screening.
Dis Markers
2007
;
23
:
315
–30.
15
Zhang Q, Kuhn L, Denny LA, De SM, Taylor S, Wright TC, Jr. Impact of utilizing p16INK4A immunohistochemistry on estimated performance of three cervical cancer screening tests.
Int J Cancer
2007
;
120
:
351
–6.
16
Wallin KL, Wiklund F, Angstrom T, et al. Type-specific persistence of human papillomavirus DNA before the development of invasive cervical cancer.
N Engl J Med
1999
;
341
:
1633
–8.
17
Winer RL, Kiviat NB, Hughes JP, et al. Development and duration of human papillomavirus lesions, after initial infection.
J Infect Dis
2005
;
191
:
731
–8.
18
Zielinski GD, Snijders PJ, Rozendaal L, et al. HPV presence precedes abnormal cytology in women developing cervical cancer and signals false negative smears.
Br J Cancer
2001
;
85
:
398
–404.
19
Overmeer RM, Henken FE, Snijders PJ, et al. Association between dense CADM1 promoter methylation and reduced protein expression in high-grade CIN and cervical SCC.
J Pathol
2008
;
215
:
388
–97.
20
Snijders PJ, van Duin M, Walboomers JM, et al. Telomerase activity exclusively in cervical carcinomas and a subset of cervical intraepithelial neoplasia grade III lesions: strong association with elevated messenger RNA levels of its catalytic subunit and high-risk human papillomavirus DNA.
Cancer Res
1998
;
58
:
3812
–8.
21
Steenbergen RD, Kramer D, Braakhuis BJ, et al. TSLC1 gene silencing in cervical cancer cell lines and cervical neoplasia.
J Natl Cancer Inst
2004
;
96
:
294
–305.
22
Wilting SM, Snijders PJ, Meijer GA, et al. Increased gene copy numbers at chromosome 20q are frequent in both squamous cell carcinomas and adenocarcinomas of the cervix.
J Pathol
2006
;
209
:
220
–30.
23
Jacobs MV, Snijders PJ, van den Brule AJ, Helmerhorst TJ, Meijer CJ, Walboomers JM. A general primer GP5+/GP6(+)-mediated PCR-enzyme immunoassay method for rapid detection of 14 high-risk and 6 low-risk human papillomavirus genotypes in cervical scrapings.
J Clin Microbiol
1997
;
35
:
791
–5.
24
Klaes R, Benner A, Friedrich T, et al. p16INK4a immunohistochemistry improves interobserver agreement in the diagnosis of cervical intraepithelial neoplasia.
Am J Surg Pathol
2002
;
26
:
1389
–99.
25
McCluggage WG, Walsh MY, Thornton CM, et al. Inter- and intra-observer variation in the histopathological reporting of cervical squamous intraepithelial lesions using a modified Bethesda grading system.
Br J Obstet Gynaecol
1998
;
105
:
206
–10.
26
van Zeeburg HJ, Snijders PJ, Pals G, et al. Generation and molecular characterization of head and neck squamous cell lines of fanconi anemia patients.
Cancer Res
2005
;
65
:
1271
–6.
27
Snijders AM, Nowak N, Segraves R, et al. Assembly of microarrays for genome-wide measurement of DNA copy number.
Nat Genet
2001
;
29
:
263
–4.
28
van de Wiel MA, Kim KI, Vosse SJ, van Wieringen WN, Wilting SM, Ylstra B. CGHcall: calling aberrations for array CGH tumor profiles.
Bioinformatics
2007
;
23
:
892
–4.
29
van de Wiel MA, van Wieringen WN. CGHRegions: Dimension Reduction for Array CGH Data with Minimal Information Loss.
Cancer Informatics
2007
;
2
:
55
–63.
30
van Wieringen WN, van de Wiel MA, Ylstra B. Weighted clustering of called array CGH data.
Biostatistics
2008
;
9
:
484
–500.
31
Alazawi W, Pett M, Strauss S, et al. Genomic imbalances in 70 snap-frozen cervical squamous intraepithelial lesions: associations with lesion grade, state of the HPV16 E2 gene and clinical outcome.
Br J Cancer
2004
;
91
:
2063
–70.
32
Heselmeyer K, Schrock E, du MS, et al. Gain of chromosome 3q defines the transition from severe dysplasia to invasive carcinoma of the uterine cervix.
Proc Natl Acad Sci U S A
1996
;
93
:
479
–84.
33
Kirchhoff M, Rose H, Petersen BL, et al. Comparative genomic hybridization reveals a recurrent pattern of chromosomal aberrations in severe dysplasia/carcinoma in situ of the cervix and in advanced-stage cervical carcinoma.
Genes Chromosomes Cancer
1999
;
24
:
144
–50.
34
Kirchhoff M, Rose H, Petersen BL, et al. Comparative genomic hybridization reveals non-random chromosomal aberrations in early preinvasive cervical lesions.
Cancer Genet Cytogenet
2001
;
129
:
47
–51.
35
Narayan G, Pulido HA, Koul S, et al. Genetic analysis identifies putative tumor suppressor sites at 2q35–36.1 and 2q36.3-q37.1 involved in cervical cancer progression.
Oncogene
2003
;
22
:
3489
–99.
36
Umayahara K, Numa F, Suehiro Y, et al. Comparative genomic hybridization detects genetic alterations during early stages of cervical cancer progression.
Genes Chromosomes Cancer
2002
;
33
:
98
–102.
37
Steenbergen RD, Hermsen MA, Walboomers JM, et al. Non-random allelic losses at 3p, 11p and 13q during HPV-mediated immortalization and concomitant loss of terminal differentiation of human keratinocytes.
Int J Cancer
1998
;
76
:
412
–7.
38
Steenbergen RD, Walboomers JM, Meijer CJ, et al. Transition of human papillomavirus type 16 and 18 transfected human foreskin keratinocytes towards immortality: activation of telomerase and allele losses at 3p, 10p, 11q and/or 18q.
Oncogene
1996
;
13
:
1249
–57.
39
Klingelhutz AJ, Qian Q, Phillips SL, Gourronc FA, Darbro BW, Patil SR. Amplification of the chromosome 20q region is associated with expression of HPV-16 E7 in human airway and anogenital epithelial cells.
Virology
2005
;
340
:
237
–44.
40
Savelieva E, Belair CD, Newton MA, et al. 20q gain associates with immortalization: 20q13.2 amplification correlates with genome instability in human papillomavirus 16 E7 transformed human uroepithelial cells.
Oncogene
1997
;
14
:
551
–60.
41
Larsen AK, Gobert C. DNA topoisomerase I in oncology: Dr Jekyll or Mr Hyde?
Pathol Oncol Res
1999
;
5
:
171
–8.
42
Postma C, Hermsen MA, Coffa J, et al. Chromosomal instability in flat adenomas and carcinomas of the colon.
J Pathol
2005
;
205
:
514
–21.
43
Simon R, Burger H, Brinkschmidt C, Bocker W, Hertle L, Terpe HJ. Chromosomal aberrations associated with invasion in papillary superficial bladder cancer.
J Pathol
1998
;
185
:
345
–51.
44
Tirkkonen M, Tanner M, Karhu R, Kallioniemi A, Isola J, Kallioniemi OP. Molecular cytogenetics of primary breast cancer by CGH.
Genes Chromosomes Cancer
1998
;
21
:
177
–84.
45
Cortes-Gutierrez EI, Davila-Rodriguez MI, Muraira-Rodriguez M, Said-Fernandez S, Cerda-Flores RM. Association between the stages of cervical cancer and chromosome 1 aneusomy.
Cancer Genet Cytogenet
2005
;
159
:
44
–7.
46
Hariu H, Matsuta M. Cervical cytology by means of fluorescence in situ hybridization with a set of chromosome-specific DNA probes.
J Obstet Gynaecol Res
1996
;
22
:
163
–70.
47
Pieber D, Bauer M, Gucer F, Reich O, Pickel H, Purstner P. Numerical chromosomal aberrations of chromosome 1 and 7 in dysplastic cervical smears.
Life Sci
2000
;
67
:
671
–8.
48
Sreekantaiah C, Bhargava MK, Shetty NJ. Chromosome 1 abnormalities in cervical carcinoma.
Cancer
1988
;
62
:
1317
–24.
49
Heselmeyer-Haddad K, Janz V, Castle PE, et al. Detection of genomic amplification of the human telomerase gene (TERC) in cytologic specimens as a genetic test for the diagnosis of cervical dysplasia.
Am J Pathol
2003
;
163
:
1405
–16.
50
Heselmeyer-Haddad K, Sommerfeld K, White NM, et al. Genomic amplification of the human telomerase gene (TERC) in pap smears predicts the development of cervical cancer.
Am J Pathol
2005
;
166
:
1229
–38.
51
Wilting SM, de Wilde J, Meijer CJLM, et al. Integrated genomic and transcriptional profiling identifies chromosomal loci with altered gene expression in cervical cancer.
Genes Chromosomes Cancer
2008
;
47
:
890
–905.
52
Thebault S, Gachon F, Lemasson I, Devaux C, Mesnard JM. Molecular cloning of a novel human I-mfa domain-containing protein that differently regulates human T-cell leukemia virus type I and HIV-1 expression.
J Biol Chem
2000
;
275
:
4848
–57.
53
Drouin V, Viguie F, Debesse B. Near-haploid karyotype in a squamous cell lung carcinoma.
Genes Chromosomes Cancer
1993
;
7
:
209
–12.
54
Dutrillaux B. Pathways of chromosome alteration in human epithelial cancers.
Adv Cancer Res
1995
;
67
:
59
–82.
55
Muleris M, Delattre O, Olschwang S, et al. Cytogenetic and molecular approaches of polyploidization in colorectal adenocarcinomas.
Cancer Genet Cytogenet
1990
;
44
:
107
–18.
56
Varella-Garcia M, Gemmill RM, Rabenhorst SH, et al. Chromosomal duplication accompanies allelic loss in non-small cell lung carcinoma.
Cancer Res
1998
;
58
:
4701
–7.