Cervical precancers are strongly associated with human papillomavirus (HPV) infection. About 5% of untreated cases may progress to the invasive stage of squamous cell carcinoma of cervix after a long latency period. A high percentage of infected women clear the infection, probably by immunological mechanisms. The progression of HPV-infected lesions to cervical cancer and the development of subsequent metastases requires additional modification of cellular genes that allow the affected cell to escape immune responses and facilitate cell invasion and metastasis.1 The molecular basis for this tumor progression is not defined completely.
Tissue inhibitors of metalloproteinases (TIMPs) are well known for their ability to inhibit the activity of matrix metalloproteinases (MMPs). MMPs are associated with extracellular matrix turnover that plays an active role in tumor invasion and metastasis. The net MMP activity is the result of the balance between the levels of the activated enzyme and TIMPs. The overproduction of MMPs and their unrestrained activity are associated with malignant conversion of tumor cells.2 Thus, the inhibitory activity of TIMPs seems to be important for inhibition of malignant progression. In fact, it has been demonstrated in model systems that tumor invasion and metastasis can be inhibited by upregulation of TIMP expression in tumor cells.3 TIMPs, however, are multifunctional proteins with diverse biological functions different from their MMP-inhibitory activities. To date, 4 different TIMPs are identified: TIMP-1, TIMP-2, TIMP-3 and TIMP-4. The products of all these genes inhibit MMPs, but they differ in tissue-specific expression, in abilities to inhibit various MMPs, to interact with pro-MMPs and to influence tumor angiogenesis. In addition, they may have pro-apoptotic or anti-apoptotic effects.4 The inhibitory effects of TIMPs on the primary tumor growth were also demonstrated by the overexpression of TIMPs genes in tumor cells. TIMPs may promote the growth of different cells including normal keratinocytes and fibroblasts.3 These multiple functions of TIMPs observed in animal model systems indicate that these proteins may modulate even early stages of tumor development and raise questions about their physiological role and possible contribution to the human cancer development and progression.
Contradictory results were obtained when TIMPs expression was studied in primary human tumors. On the one hand, the increased expression of both TIMPs and MMPs is often associated with paradoxical adverse prognosis in many solid tumors.3, 5, 6 On the other hand, the increased MMP-2 activity and the lack of TIMP-2 expression correlate with poor prognosis in soft tissue sarcomas.7 In addition, the methylation-associated silencing of TIMP-3 is often observed in many tumor types.8, 9
It has been demonstrated that the high transcription levels of both MMP-2 and MMP-9 and their inhibitors TIMP-1 and TIMP-2 are typical features of the invasive cancer cells of cervix and the surrounding stromal cells and are rarely detected in cervical intraepithelial neoplasias (CIN) by in situ RT-PCR and hybridization.10, 11 At the same time, Nuovo et al.11 demonstrated that the increased depth of invasion was associated with a significant increase in the ratio of MMP mRNA to TIMP mRNA due to the decrease in the number of tumor and stromal cells expressing TIMPs. It has been suggested that the imbalance between TIMPs and MMPs determines the aggressiveness of cervical cancer.11 Mechanism of suppression of TIMPs expression in cervical carcinomas remains unknown.
Aberrant methylation of the promoter region of the tumor suppressors and certain other genes that are functionally important in neoplastic process is strongly implicated in the transcriptional inactivation of these genes in different tumor types.12 We have examined whether methylation of the promoter region is responsible for the decreased expression of TIMP-2 gene in cervical carcinomas. We have demonstrated that CpG island located in TIMP-2 5′ regulatory region is frequently methylated in invasive cervical carcinomas and cervical cancer cell lines. TIMP-2 expression is effectively restored after5-aza-C and 5-aza-dC treatment of the cell lines with intensive methylation of the 5′ portion of the TIMP-2 CpG island upstream to the transcription start site. Our finding indicates that TIMP-2 is another target of aberrant methylation during tumorigeneses. This is the first description of aberrant hypermethylation of the TIMP-2 gene in human cancer.
CIN, cervical intraepithelial neoplasia; 5-aza-C-5-azacytidine, 5-aza-dC-5-aza-2′-deoxycytidine; HPV, human papillomavirus; MMP, matrix metalloproteinase; MSP, methylation-specific PCR; RT-PCR, reverse transcription PCR; SCC, squamous cell carcinoma; TIMP, tissue inhibitor of metalloproteinase.
MATERIALS AND METHODS
The study population consisted of 36 patients diagnosed with squamous cell carcinoma of uterine cervix and treated in the department of radiosurgery of Russian Cancer Research Center. The surgical material consisted of 36 specimens of invasive squamous cell carcinoma of I, II and III FIGO stages and of morphologically normal cervical tumor-adjacent tissues or leukocytes from peripheral blood of the same patients. All specimens were collected according to the regulations of the NN Blochin Cancer Research Center. All tumors were HPV16- or HPV18-positive according to PCR analysis.
The human cervical carcinoma cells HeLa, SiHa and Caski (American Type Culture Collection, Rockville, MD) were maintained in DMEM supplemented with 10% FCS and 2 mM glutamine.
Treatment with demethylating agents
Cells were seeded at routine cell densities and treated with 10 μM 5-aza-C or 5 μM 5-aza-dC (Sigma, St. Louis, MO) for 3–4 days. The medium was changed daily. Cells were used on Day 3 after treatment for RNA isolation.
DNA and RNA preparations
DNA and RNA were isolated from frozen tissues or cells homogenized in the 4.5 M solution of guanidine thiocyanate by centrifugation through cesium chloride cushion as described previously.13
Bisulfite-based cytosine methylation analysis
The reaction of bisulfite conversion was carried out according to Olek et al.14 with minor modifications. DNA (50–200 ng) was digested with RsaI or EcoRI restriction enzymes, denatured in 0.3 M NaOH for 5 min at 95°C and 2× volume of 2% low melting agarose was immediately added to the hot solution. Then the aliquots of the mixture (10 μl) were transferred into cold mineral oil to form agarose beads with denatured DNA. Agarose beads were incubated with 3.1 M sodium bisulfite and 0.5 mM hydroquinone solution for 16 hr at 50°C under mineral oil, and then washed 6 times for 15 min with TE buffer pH 8.0. The DNA samples were desulfonated by addition of NaOH to the final concentration 0.2 M followed by incubation of agarose beads 2 times at 25°C for 15 min. Afterwards the agarose beads were washed with TE buffer pH 8.0 for 15 min and 3 times for 15 min with bi-distilled H2O and stored at −20°C. The DNA aliquots (5–20 ng) were used for PCR after melting of the agarose beads.
TIMP-2 primers that correspond to the upper strand of bisulfite-modified DNA and do not contain CpG dinucleotides in the original sequences were used in the first round of PCR: sense 5′GGT TTT TGT TTT AAA GGA TAT TTT TTG3′ (1788) and antisense 5′ACT CCT TAC CTA CAT CTA CAT TAC A3′ (2654, GenBank accession number: U 44381). Here and further the positions of the primers are given in base pairs related to the respective gene sequence in GenBank. The bisulfite converted DNAs were amplified with these primers under the following conditions: 94°C, 2 min; 35 amplification cycles (94°C, 30 sec; 54°C, 30 sec; 72°C, 30 sec) and finally at 72°C, 5 min. For sequencing of the promoter region of TIMP-2 the DNA fragments obtained after the first round of PCR were amplified in the second round of PCR with the primers: sense 5′GGT TTT TGT TTT AAA GGA TAT TTT TTG (1788) and antisense 5′ CAA ACT TTC TCT CCT CTT TAT CTC (2368), under the following conditions 94°C, 2 min; 35 cycles (94°C, 30 sec; 53°C, 30 sec; 72°C, 30 sec) and final extension at 72°C for 5 min. The resulted PCR products were fractionated in low melting agarose (Sigma), purified using QIAEX II kit (Qiagen) and used for automated sequencing on DNA Sequencer (ABI Prism 3100-Avant Genetic Analyzer) or for cloning into pGEM-T Easy vector (Promega, Madison, WI) and then used for sequencing.
The first-round PCR products were used for methylation-specific PCR (MSP) assay also (15). The second round of MSP were carried out with 2 set of primers specific for upper strand of bisulfite-modified DNA containing CpG dinucleotides: (i) sense 5′-TTT GGT GTT TTG GAA GAA TGG GTG (1919) and antisense 5′-CCA ACC CCA ATC CCC ACT ACA (1987) for unmethylated sequence (U), (ii) sense 5′-TTT GGT GTT TTG GAA GAA CGG GCG (1919) and antisense 5′-CGA CCC CGA TCC CCG CTA CG (1987) for methylated sequence (M). The PCR was carried out under the following conditions: 94°C, 3 min; 30 cycles (94°C, 30 sec; 64°C, 20 sec and 72°C, 20 sec for U-primers), 30 cycles (94°C, 30 sec; 67°C, 20 sec and 72°C, 20 sec for M-primers) and 72°C, 5 min. Aliquots of the PCR samples were fractionated in 3% agarose gel and photographed.
The level of mRNA of TIMP-2 gene was determined by reverse transcription combines with RT-PCR. RT was carried out in 20 μl reaction mixture containing 1μg of total cellular RNA, 1× buffer (50 mM Tris, pH 8.3, 61.5 mM KCl, 3 mM MgCl2); 1 mM of each dATP, dTTP, dGTP and dCTP; 20 pmol of random hexamer; 200 U of reverse transcriptase (Superscript II Rnase H−, Gibco BRL, Carlsbad, CA) at 25°C for 10 min and then at 42°C for 50 min. The reaction was stopped by incubation of the reaction mixture at 70°C for 15 min. TIMP-2 specific PCR was carried out with the primers: sense 5′ GGT CTC GCT GGA CGT TGG AG (exon 2) and antisense 5′ GGA GCC GTC ACT TCT CTT G (exon 4, GenBank accession number: NT 010843) under the following conditions: 94°C for 2 min; 26 cycles (94°C, 30 sec; 58°C, 30 sec; 72°C, 40 sec), and finally at 72°C for 3 min. To control the content of the initial total mRNA the expression of GAPDH gene was analyzed using the same cDNA with the primers: sense 5′ACC ACA GTC CAT GCC ATC AC and antisense 5′TCC ACC ACC CTG TTG CTG TA under the following conditions 94°C, 3 min; 23 cycles (94°C, 30 sec; 60°C, 30 sec; 72°C, 1 min,) and 72°C, 10 min.
Methylation pattern of the 5′-CpG island of TIMP-2 gene
To analyze the methylation pattern of the CpG island of TIMP-2 gene and to determine what region is the most frequent target for methylation, the DNAs from 5 primary tumors and 2 cell lines SiHa and Caski were analyzed by bisulfite sequencing (Fig. 1). These samples were chosen for detailed sequence analyses, because they demonstrated methylation at HpaII restriction enzyme sites within the TIMP-2 CpG island by Southern blot analysis (data not shown). We observed methylation at many CpG dinucleotides across CpG island in SiHa and Caski cells (Fig. 1, lanes 2–5) that express TIMP-2 mRNA at hardly detectable levels (Fig. 2a). In contrast, nearly all CpG sites were unmethylated within the examined region in HeLa cells (Fig. 1, lane 1). HeLa cells express TIMP-2 mRNA at well-detectable by RT-PCR level (Fig. 2a). The methylation patterns of both methylated cell lines are different, but they have common features. The discrete region of dense methylation is observed within the 5′-portion of the sequenced region (from −350 bp to −240 bp) that contains putative binding sites for transcription factors AP1, AP2 and Sp1.16 The region flanking the transcription start site (from −100 to +100) displays different degrees of mosaic methylation in different cell lines. Methylation pattern of the population of molecules (Fig. 1, lane 3) was consistent to the pattern of clones (Fig. 1, lanes 4,5). The discrete region of dense methylation (−350 bp to −240 bp) is also clearly detectable in individual molecules.
The primary tumors that were found to be methylated by Southern blot analysis were shown to be methylated at the same region by bisulfite sequencing. They showed complex methylation pattern represented by a mix of methylated, hemimethylated and unmethylated CpG sites (Fig. 1, lanes 7–13). It seems that a more complex methylation pattern in the primary tumors reflects the presence of normal tissues in the samples or methylation of only 1 of 2 alleles. Normal tissues from the same patients were largely unmethylated, but partially methylated CpG dinucleotides were present in 1 of 5 samples in the 5′-portion of the CpG (Fig. 1, lane 6). This indicates that some alleles of TIMP-2 may be methylated in morphologically normal tissues. Leucocytes of peripheral blood were unmethylated also, although we detected methylation of individual CpG sites within this region (Fig.1, lane 14). Although the methylation pattern of tumors was complex, all of them displayed the discrete region of dense methylation, detected earlier in cell lines, and mosaic methylation of region flanking the transcription start site.
Thus, the genomic bisulfite sequencing demonstrated that the highest density of methylation was observed within the 5′-portion of the CpG island containing putative binding sites for transcription factors AP1, AP2 and Sp1 in all analyzed cancer cells. The 3′-portion of the CpG island encompassing the transcription start site and the first exon was less methylated and different samples had unique methylation pattern of this region. Thus, to analyze the methylation status in TIMP-2 CpG island by MSP assay, primers located in the 5′ region should be used.
Effect of demethylating agents on TIMP-2 expression
Methylation of CpG island of TIMP-2 gene correlates with extremely low level of TIMP-2 mRNA in SiHa and Caski cells. To determine whether methylation was related to the decreased level of TIMP-2 mRNA, cells were treated with the demethylating agent 5-aza-C and 5-aza-dC. In several experiments these agents effectively restored TIMP-2 expression in both cell lines as determined by RT-PCR (Fig. 2a). The slight increase of TIMP-2 transcription was observed under the same conditions in HeLa cells (Fig. 2a). These results suggest that TIMP-2 silencing is related to methylation of CpG island.
Incidence of TIMP-2 methylation in cervical carcinomas
MSP assay was used to determine the frequency of methylation of TIMP-2 in cervical carcinomas. The MSP assay uses PCR primers to differentially distinguish between methylated and unmethylated sites after treatment of DNA samples with sodium bisulfite. The primers used for this assay span the discrete region, that was methylated in all samples analyzed by bisulfite genome sequencing (−350 bp to −240 bp). Figure 2b shows representative examples of MSP products analyzed on agarose gel for methylated and unmethylated samples determined by bisulfite sequencing (lanes 2 and 1,3 respectively) and primary tumors (lanes 4–6). The primary tumors always had both methylated and unmethylated alleles. This observation is consistent with results of bisulfite genome sequencing.
The results of MSP assay demonstrated that methylation of the examined region of the CpG island of TIMP-2 was observed in 17 of 36 primary cervical carcinomas (Table I). This results confirmed the data that were obtained earlier by Southern blot analysis (data not presented). In the majority of normal cervical tissue samples and leucocytes from the same patients the CpG island of TIMP-2 was not methylated. TIMP-2 methylated alleles were found in 4 samples of the normal tissue adjacent to the tumors with methylated status of TIMP-2. This result suggests that this epigenetic alteration occurs frequently and may be revealed early during cervical tumor progression.
Table I. Hypermethylation of TIMP-2 Promoter in Cervical Cancer Cell Lines and Primary Tumors
Values are number of samples, with percentage in parentheses.
Morphologically normal tissues
A significant finding of our study is that aberrant hypermethylation of human TIMP-2 is a frequent event in cervical primary tumors and carcinoma cell lines. The restoration of TIMP-2 expression by 5-aza-dC treatment confirmed a causal association of DNA hypermethylation with TIMP-2 silencing. Our finding suggests that the aberrant methylation of TIMP-2 associated with transcription repression favors the development of cervical carcinomas. This result also indicates a possible tumor suppressor role for TIMP-2. Hypermethylation of TIMP-2 gene in human cancer has not been described up to date.
The aberrant hypermethylation of TIMP-2 was absent in the majority of morphologically normal tissues of cervix and so is tumor-specific. Methylated alleles of this gene, however, were found in 4 of 36 samples of morphologically normal tissues adjacent to the tumors with methylated TIMP-2. These data provide indirect evidence that aberrant methylation of TIMP-2 may be important during early tumor growth and support the suggestion that TIMP-2 is capable to modulate early stages of human tumor development.
Bisulfite sequencing showed small methylated region of TIMP-2 promoter (−350 to −240, relative to the transcription start site). Methylation of this region seems to be important for gene silencing because it correlates with suppression of transcription in cell lines and observed in all hypermethylated tumors. This region contains putative binding sites for transcription factors, including site for methylation sensitive factor AP-2.17 Methylation of the region flanking the transcription start site (from −100 to +100) seems not to be critical for silencing of the gene expression because this region displays different degrees of mosaic methylation in different cell lines and tumors. TIMP-2 gene is not the unique example of this kind of methylation pattern. Dense methylation of a small region, located from −248 to −178 bp relative to the transcription start site, invariably correlates with the lack of hMLH1 gene expression in colorectal carcinoma cell lines.18 Methylation of discrete region of MGMT CpG island is associated with heterochromatinization of the transcription start site and silencing of gene in multiple myeloma tumor cell line, although a relatively methylation-free region that contained the transcription start site, minimal promoter and minimal enhancer was detected in these cells.19
Nuovo et al.11 demonstrated that a significant increase in the MMP mRNA to TIMP mRNA ratio correlated with increased depth of invasion of cervical carcinomas and presence of microvascular invasion or actual metastases. It has been suggested that the evolution of cervical carcinomas with regard to MMPs and TIMPs expression is multi-step process.20 Early invasion probably requires activation of both MMPs and TIMPs. The next steps of evolution select for cells with imbalance between MMPs and TIMPs expression. The TIMP-2 methylation is observed in the high percentage of cervical carcinomas (47%). Because methylation results in decreasing of TIMP-2 expression, it is likely that the imbalance between MMP and TIMP activities due to the suppression of TIMP-2 may occur in many cervical invasive carcinomas and be may related with the later steps of carcinoma evolution. Further investigations are needed to define the possible value of TIMP-2 methylation as a marker of cervical tumor aggressiveness and poor prognosis.
The high percentage of TIMP-2 methylation associated with the decreased transcription in invasive cervical carcinomas contradicts the described earlier paradoxical correlation between the high level of TIMP-2 expression and the poor prognosis for other types of carcinomas.2, 5, 6 It was demonstrated that TIMP-2 possesses mitogenic activity in cell culture, and has anti-tumoral, anti-apoptotic and anti-angiogenic effects in vivo in animal model systems, where its level is super-physiological.3 At high concentration TIMP-2 is a potent inhibitor for MMP-2 and MMP-9. At low concentration TIMP-2 is an adaptor molecule that is required for MMP-2 activation at the cell surface.22 Thus, MMP activity is tightly regulated by TIMP-2. Taking into the account the dose-dependent regulation of MMP-2 by TIMP-2 and multiple other effects of TIMP-2, it is hard to predict the possible contribution of variations of TIMP-2 expression to any type of cancer. Contradictory results mentioned above may be explained by the prevalence of different TIMP-2 functions in different carcinomas. It is also possible that the imbalance between the expressions of TIMP-2 and MMPs, but not the high level of TIMP-2 expression, plays a certain role in the aggressiveness of other carcinomas.
It was shown that common features of many CpG islands are their proximity to normally methylated Alu repetitive elements and position of cluster of Sp1 elements around of transcription start site within CpG island.23 Graff et al.24 have demonstrated that methylated Alu repeats play a role of “de novo methylation centers,” from which the aberrant methylation may spread to the adjacent CpG island of E-cadherin. B1 and B2 repetitive elements, the rodent equivalent of human family of Alu repeats, may also act as a methylation center that directs de novo methylation of the CpG island of rat α-fetoprotein gene. The Sp1 sites are believed to function as the cis-acting elements that protect from methylation the CpG island of the promoter region in normal tissues.25, 26 Thus, it is suggested that the methylation pattern is a result of the dynamic equilibrium between forces that promote and inhibit spreading of methylation.27 The TIMP-2 CpG island contains 3 Sp1 sites upstream of the transcription start site and the 5′-edge of the CpG island is juxtaposed to the Alu repeats (Fig. 1a). The proximal Alu repeat is methylated in normal breast tissues.23 We did not analyze the methylation status of the Alu repeat located upstream of TIMP-2 gene in normal cervix tissues, but we demonstrated that the high density of methylation was observed at the 5′ edge of TIMP-2 CpG island and the density of methylation decreased toward the transcription start site. Our data provide indirect evidence that the proximal Alu repeat may be an origin of methylation spreading into the adjacent TIMP-2 CpG island in cervical carcinomas. Three Sp1 sites cannot prevent this spreading of methylation. These results are consistent with described earlier methylation of CpG islands of E-cadherin and VLH that occurs despite the presence of Sp1 clusters within their CpG islands in renal and breast cancer cell lines.23 It remains obscure why the protective properties of Sp1-rich region are lost in cancer cells.
In conclusion, our results suggest that TIMP-2 methylation is a significant step in malignant progression of cervical carcinomas. It is likely that decreased TIMP-2 expression contributes to different stages of malignant progression of this cancer. This is the first description of aberrant hypermethylation of TIMP-2 gene in human cancer.
We thank Dr. O.V. Sacharova for providing the clinical information and Dr. A.F. Kisseleva for technical assistance.