In this study, we found a poor correlation between CN alterations and changes in gene expression. Only a small percentage of genes located in CNAs was de-regulated (15.6%), which was a minor difference (0.8%) from that of the subset of genes with non-CN alterations (14.8%). In the subset of CN-altered and deregulated genes, the way of deregulation was not necessarily the same as the CN alteration, i.e.
, amplified genes were not always upregulated, instead they were often downregulated, and some deleted genes were found to be upregulated. This analysis was essentially based on the comparison of copy-number alterations and global gene expression investigated through genomic technologies and the characterization of certain regions by FISH and qPCR. The healthy normal epithelium of the cervix might not be the best control to measure the level of gene expression in cell lines. However, it is difficult to select an appropriate control, because these cell lines have been maintained in culture for many years. Although they do not fully represent the complexity of a tumor, they usually retain their genetic properties 
. The use of primary cultures of normal cervical epithelium may be a better control, but there are not any available commercially. On the other hand, similar global results have been observed between cervical carcinomas and the cell lines reported in this study, when the same set of healthy normal controls were used (data not shown).
The results of this paper suggest that most of the CNAs in the evaluated cell lines, identified formerly with the 100 K microarray as succession of altered SNPs, are not continuously altered regions that include the full DNA segment defined by the altered SNPs. Rather, they appear to be composed of small or partial deletions or gains, where the altered SNPs are located, alternating with long stretches of normal DNA. On the other hand, the MRRs are more likely to be completely CN altered, because the proportion of deregulated genes was increased up to 3 fold, particularly in those having a high density or more than 500 SNPs. In those MRRs, the proportion of deregulated genes did not change with the number of SNPs per gene, further supporting they are entirely CN altered. In fact, the 2 MRRs having more than 500 SNPs (5-1 and 5-4), located at 5p, were confirmed as being completely amplified by FISH. In contrast, the finding that in MRRs composed of less than 500 SNPs, the percentage of deregulated genes increased with the number of SNPs/gene, strongly suggests those regions are CN affected in a discontinuous manner. However, it cannot be ruled out that some MRRs having less than 500 SNPs are completely CN altered, such as MRR 5-5 located at 5p, which had 346 SNPs and was confirmed fully amplified by FISH.
In previous studies, Chr 5p has often been found to be amplified in CC and cell lines derived from them, and many genes located in the region have been involved in the tumorigenic process 
. Nevertheless, a rather poor correlation has been found between amplification and gene expression in this region, both in cell lines (22%) 
and invasive tumors (18.9%) 
. Although in the present study, a higher percentage of genes was deregulated (33.5%), a large proportion of the investigated genes was not deregulated. Furthermore, not all the deregulated genes were upregulated, 9 of them were downregulated. Because the entire 5p arm was demonstrated to be amplified by FISH (), it is clear that not all amplified genes are upregulated, instead, some of them might be repressed, possibly by epigenetic mechanisms. The clustering of deregulated genes suggests that, besides the amplification of the segment, the location of genes within the same region of chromatin, perhaps at a loop level, may influence gene expression 
The upregulated genes at MRR 5-1 may have a role in the carcinogenic process. They include BRD9
, which participate in DNA repair and cell-cycle regulation, SDHA
involved in mitochondrial oxidative phosphorylation 
, and TRIO
, which promotes the exchange of GDP by GTP and could play a role in coordinating cell-matrix and cytoskeletal rearrangements necessary for cell migration and cell growth. In fact, TRIO
has been associated with progression of bladder cancer 
and soft tissue sarcomas. In the latter, a clear correlation has been demonstrated between amplification and gene upregulation 
. Another upregulated gene located in this region was CEP72
, which regulates the localization of key centrosome proteins involved in spindle formation 
. This gene has been found frequently amplified in non-small-cell lung cancers 
, which encodes the catalytic subunit of the telomerase complex hTERT and have been found to be amplified or upregulated in more than 90% of squamous cell cervical carcinomas and 40% of CIN III lesions 
, had a fold change of 1.45, just below the selected cut-off (). The concordance between the upregulated genes found at 5p, in this and previous reports, was close to 60% 
. Most of the remaining genes that were shown to be upregulated in other studies had a fold change higher than 1.5 in this study but did not pass the delta score (; see Materials and Methods
). However, additional genes linked to cancer processes were found deregulated in this study including AHRR
, and MRPS30
involved in apoptosis, CDH6
in cell adhesion and CEP72
in the cell cycle.
In contrast with Chr 5, which shows clearly amplified 5p and unaltered 5q, the overall profile of Chr 3 shows that 3p is generally deleted and 3q amplified in some cell lines. They look like the inverse of one another. Furthermore, the phenomenon of amplification in 3q appears to be quite different from that in 5p. For instance, the level of gain or amplification was lower than in 5p and it did not include the entire 3q arm in all the cell lines; instead, only certain sparse regions were found altered recurrently. In fact, the average log2 ratio of MRRs at 5p was almost 2 fold times higher than those located at 3q (p<0.001, t-test; calculated from Table S1
). The full arm (CaLo and HeLa) or several regions (CaSki and SiHa) of 3q were found to be amplified (), similarly to the findings in previous reports 
. This could explain the lower proportion (13.4%) of deregulated genes found in 3q compared with that in 5p (33.5%). However, even in those cell lines where most of 3q was gained, the proportion of deregulated genes did not rise (CaLo) or increased modestly (HeLa). In addition, in 3q, the proportion of downregulated genes was higher than the proportion of upregulated genes, particularly in 3q26, where 8 of 11 deregulated genes were downregulated, even some of them were recurrently gained. Furthermore, similarly to 5p, deregulated genes seemed to be grouped in clusters in 3q26–29. These findings indicate that an increase in the copy number does not necessarily mean that genes located in those regions will be upregulated. It suggests that, in those entirely amplified regions, epigenetic mechanisms could be involved in gene repression. On the other hand, the increased frequency of downregulated genes with the number of amplified SNPs in the subset of genes located in MRRs, which seems to be not entirely amplified (having less than 500 SNPs; ), supports that partial gene amplification may be a mechanism of gene silencing. This idea has been proposed theoretically 
The 3q26 region has been previously identified as gained or amplified in biopsies or cell lines derived from CC by using CGH or FISH 
. Recognized tumor genes, such as EVI1
, and genes associated previously with CC (TERC
, and PIK3CA
) are located in this region. However, it has not been demonstrated that these genes were upregulated 
, particularly in the same samples where the CN alterations were found. In this study, EVI1
, and LAMP3
were neither found CN altered recurrently nor upregulated in all the cell lines studied. TERC
was found gained in CaLo, CaSki, and HeLa but upregulated only in HeLa (data not shown). However, conclusions with these negative results from the microarrays may be too risky without the validation with different methodologies, like qPCR and qRT-PCR. Interestingly, the gene encoding for tumor necrosis factor (ligand) superfamily member 10 (TNFSF10), a protein that induces apoptosis in transformed and tumor cells, was found to be downregulated in the 4 cell lines, even though the gene was recurrently gained (MRR 3-13; ). This gene is located in the same region as 2 other downregulated (NLGN1
) and 2 upregulated (AADACL1
) genes, which have not been previously associated with CC. However, the protein encoded by ECT2
(epithelial cell transforming sequence two oncogene) is a transforming protein that is a nuclear guanine nucleotide exchange factor (GEF) and regulates RhoB-mediated cell death after DNA damage in cervical cell lines 
. The expression of this gene is elevated with the onset of DNA synthesis and remains elevated during the G2 and M phases 
. Increase in gene dosage by DNA amplification is a common mechanism to achieve overexpression of genes in tumors 
showed the highest fold change of the upregulated genes found at 3q; therefore, it is a good candidate for the amplified driven oncogene in 3q26 for CC.
In 1q the correlation between CN and gene expression was also very poor and similar to the figures seen in 3q. However 3 genes, IQGAP3, CENPF and PARP1, were upregulated more than 3 fold times compared with controls. The protein codify by PARP1 (Poly ADP-ribose polymerase-1) is a DNA binding protein that detects specifically DNA strand breaks generated by different genotoxic agents. Whereas activation of PARP-1 by genotoxic stimuli facilitates DNA repair and cell survival, severe DNA damage triggers different pathways of cell death, including PARP-mediated cell death 
. Cells with BRCA1 loss of function are deficient in DNA double strand break repair thus activating PARPs whose catalytic activity is immediately stimulated by DNA strand-breaks 
. Although in these cell lines the expression of BRCA1 and BRCA2 genes did not change (data not show), PARP1 could help these genes in that DNA reparation pathway.
The overall correlation between the CN alterations and changes in gene expression was about 15% in CC cell lines. This low correlation could be related to several factors. First, most genes located in the CNAs, identified formerly in the cell lines with the 100 K microarray, were not altered in the number of copies. Second, in the genomic segments confirmed entirely amplified, like 5p, the percentage of deregulated genes was over 33%, but not all of them were upregulated. Therefore, it is clear that not all amplified genes are upregulated; instead, some of them may be repressed, possibly by epigenetic mechanisms. Third, deregulated genes were found in clusters, suggesting that, besides the segment amplification, the location in the same chromatin region may influence gene expression. Fourth, the steady rise of downregulated genes with the increase of amplified SNPs in regions CN altered discontinuously suggests that partial gene amplification could be a mechanism of silencing gene expression. Additional genes were identified up- or downregulated at 5p, 3q and 1q that could be involved in cervical carcinogenesis, particularly in apoptosis, including CLPTM1L, AHRR, PDCD6, and DAP in 5p, TNFSF10 and ECT2 in 3q and PARP1 in 1q.