The present study describes genome wide LOH and copy number analyses of 10 cervical cancer cell lines using SNP and CGH arrays to identify common regions of amplification and deletion. Also, the sensitivity to detect CNA by the 2 platforms was compared. Furthermore, integrated copy number and gene expression analyses were performed to investigate an association between CNA and gene expression.
CNA retrieved from array CGH and genotype information analyzed from SNP arrays revealed very complex large-scale changes. Chromosome arms that were found to be gained and/or amplified in most of the cell lines by array CGH in this study were 5p, 8p, 8q, 9p, 9q, 17q and 20q. Especially gains of 5p, 8q and 20q have been frequently reported in cervical cancer [6, 7, 10, 21–25]. Also the 1q and 3q arms, often found to be gained [7, 10, 21, 24–26], were overrepresented in 5 of the 10 cell lines. Gains of 17q and chromosome 9 are less consistently reported [6, 22]. Gain of 8q24, present in 8 cell lines, included c-Myc in the minimal region of overlap. Amplification of the c-Myc oncogene is a frequent event in cervical tumors, found in 25% to more than half of the tumours [27, 28]. Chromosome arms affected by loss in half, or more, of the cell lines were 4p, 8p, 11q and 13q of which especially 4p and 13q have been reported regularly in cervical cancer [7, 10, 22, 25, 26, 29]. Six cell lines showed loss of 11q25 that includes OPCML in the minimal region of overlap. OPCML is a tumour suppressor gene in epithelial ovarian cancer . Although the genotype data have to be interpreted with some caution because of the use of cell lines, many of the frequent changes that we observed were in concordance with previous studies.
The SNP-LOH array data showed that most chromosome arms were affected by LOH in the cell lines. More than 5 overlapping regions of LOH were evident for chromosome region 5q, 6p, 6q, 8p, 10q, 11q, 14q, 18p, 18q and 20p. LOH of the majority of these chromosomes, specifically 5q, 6p, 8p, 10q, 11q and 18, has been reported previously as frequent events in cervical cancer [9, 31, 32]. Whereas LOH of 8p resulted in copy number loss, LOH of 6p did not result in CNA for most cases, indicative of mitotic recombination events.
A comparison between CGH and SNP array revealed that the overall concordance in detection of the same areas with CNA was on average 91% for gain and 94% for loss (Table 1). Thus both platforms are comparable in the detection of particular regions with gain or loss. Additionally, the use of high-density SNP arrays showed that cervical cancer cell lines harbor multiple regions of LOH which are not detected by methods which screen for CNA such as array CGH. Accordingly, comparison of array CGH and SNP array in genome-wide analysis shows that detection of CNA by SNP array is reliable and has the added advantage of providing genotype information. However, LOH was not consistently detected by SNP array because many copy number changes did not result in true reduction to homozygosity since most of the cervical cancer cell lines are multiploid (Figure 2C and Table 2). As expected, the areas with gain and amplification showed less LOH than regions with deletion. The overlap between copy-reduction regions and LOH was 10% higher for array CGH than for SNP array (Table 2), which suggests that array CGH is slightly better in assigning the correct regions for CNA. Although some regions of LOH are missed, thereby particularly underestimating the amount of LOH in areas with gain and amplification, an estimation was made for the distribution of LOH in regions with loss, gain and copy-neutral regions. We found that copy-neutral and copy-gain regions accounted for around 75% of all LOH identified (17% represented copy-gain LOH) while LOH associated with copy-reduction occurred at a rate of about 25% (Table 3). These percentages are similar to a recently published paper from Calhoun et al. that described only 32% of LOH events to be associated with copy-reduction . Further improvement in the data from the SNP arrays may be possible using higher density SNP arrays, thereby increasing resolution and the ability to assess subtle copy number changes.
Genome-wide, we did not find a correlation between copy-number changes and expression. Lack of an association may be due to the different histological types from which the cell lines originate (Squamous carcinoma, Adenocarcinomas and Adenosquamous carcinomas) with complex genetic aberrations. Besides increased DNA dosage which may result in overexpression, transcription is influenced by other factors such as the presence and quantity of transcription and repressive factors; CpG methylation; histone methylation and acetylation status, and possibly microRNA's . Thus, amplification or deletion on its own does not have to be a prerequisite for a change in gene expression. Amplification of 5p, however, correlated with a significant higher gene expression for 22% of the genes, including SKP2, TRIO, OSMR, RPL37 and PDCD6 (Table 4). The latter genes were previously reported in cancer, either associated with increased expression or with growth stimulatory properties [34–38].
Array CGH findings of CNA at 5p, 8p and 20q were confirmed by FISH analysis. Gains and amplifications at 5p, as found by array CGH, were validated by BACs covering candidate oncogenes, SKP2, TRIO, ANKH, GDNF and hTERT. SKP2 is important for cell cycle progression and its inhibition decreases proliferation of tumour cells . Besides gain of the region covering the SKP2 gene, we found a significant increase in expression between normal cervical epithelial cells, tumour cells without amplification and tumour cells with amplification. Correlation between amplification of chromosome 5p and increased expression of SKP2 was previously shown in HPV-immortalized keratinocytes . Upregulated expression of SKP2, ANKH and TRIO, as determined by qRT-PCR in cell lines with 5p amplification, confirmed gene expression data from Affymetrix focus arrays. TRIO, which has a putative role in cell cycle regulation, was found to be amplified and highly expressed in bladder cancer  that associated with proliferation and invasive phenotype. ANKH, which is involved in tissue calcification , was found to be amplified in bladder cancer  and small cell lung cancer cell lines . Thus, upregulation of expression through amplification may be the case for TRIO, ANKH and SKP2 in some neoplasms, including cervical cancer, as indicated by our data. This may hold for hTERT and PRKAA1 as well since it was reported that amplification at 5p in cervical cancer correlated with increased gene and/or protein expression [43, 44]. The telomeric region at 8p includes a candidate tumour suppressor gene, TUSC3 [45, 46]. However, physical loss of this region did not correlate with a decrease in gene expression of TUSC3, as determined by qRT-PCR. All cell lines showed gains or amplifications of 20q, a chromosome arm which is often amplified in cancer, including cervical cancer [10, 22, 24, 47]. At 20q increased copy numbers, detected by array CGH, were confirmed by FISH. Candidate oncogenes, ZNF-217, CYP24A1, MYBL2 and AIB2, encompassing the region analyzed by FISH, were previously reported to be amplified in other types of cancer as well [48–50]. However, no significant difference in gene expression was found between tumours with gain and tumours with amplification of 20q. Also in oesophageal adenocarcinoma, no correlation was detected between amplification of ZNF-217/CYP24A1 at 20q and increased mRNA expression .