Abstract
Tissue inhibitor of metalloproteinase-3 (TIMP-3) antagonizes matrix metalloproteinase activity and can suppress tumor growth, angiogenesis, invasion, and metastasis. Loss of TIMP-3 has been related to the acquisition of tumorigenesis. Herein, we show that TIMP-3 is silenced in association with aberrant promoter-region methylation in cell lines derived from human cancers. TIMP-3 expression was restored after 5-aza-2′deoxycytidine-mediated demethylation of the TIMP-3 proximal promoter region. Genomic bisulfite sequencing revealed that TIMP-3 silencing was related to the overall density of methylation and that discrete regions within the TIMP-3 CpG island may be important for the silencing of this gene. Aberrant methylation of TIMP-3 occurred in primary cancers of the kidney, brain, colon, breast, and lung, but not in any of 41 normal tissue samples. The most frequent TIMP-3 methylation was found in renal cancers, which originate in the tissue that normally expresses the highest TIMP-3 levels. This methylation correlated with a lack of detectable TIMP-3 protein in these tumors. Together, these data show that methylation-associated inactivation of TIMP-3 is frequent in many human tumors.
Introduction
The TIMPs 4 are a family of molecules that inhibit the proteolytic activity of the MMPs. Tumors frequently show an increase in MMP expression and/or a decrease in TIMP expression leading to a net increase in proteolytic activity, which may be necessary for tissue remodeling during tumor growth, angiogenesis, invasion, and metastasis (1) . However, tissue-specific patterns of expression and differences within regions of primary tumors illustrate the complexity of the balance between the TIMPs and the MMPs, and it is simplistic to view these alterations as a straightforward stochastic relationship in the tumor as a whole (2) . Despite this complexity, in a variety of experimental cancers, TIMP expression can suppress primary tumor growth, angiogenesis, invasion (in vitro and in vivo), and metastasis, whereas depletion of TIMPs can increase primary tumor growth, invasion, and metastasis (1 , 3, 4, 5, 6) .
TIMP-3 is a secreted 24-kDa protein that, unlike other TIMP family members, binds to the extracellular matrix. TIMP-3 is most highly expressed in normal kidney and brain (1) . TIMP-3 overexpression in tumor cells has been shown to induce apoptosis (6 , 7) , as well as to suppress primary tumor growth and angiogenesis (8, 9, 10) , indicating that TIMP-3 may suppress even the earliest aspects of tumor development. Furthermore, decreased TIMP-3 expression at the invasive front of human colon carcinomas suggests that a regional loss of TIMP-3 may facilitate tumor invasion and metastasis (11) . Neoplastic variants of the mouse JB6 model of tumor progression lack TIMP-3 expression, whereas the preneoplastic variants retain TIMP-3 expression. In these neoplastic variants, TIMP-3 expression was restored by treatment with the demethylating agent, 5-azacytidine, indicating that the lack of TIMP-3 expression in these cells may be associated with aberrant promoter region methylation (12) . The role of methylation of TIMP-3 in human cancer has not been examined.
In this study, we examined whether aberrant 5′ CpG island methylation may be associated with the loss of TIMP-3 in human cancers. We show that the loss of TIMP-3 expression is associated with dense methylation of the 5′ CpG island in cell lines from many common human cancers and can be restored in colon cancer cell lines after 5Aza-dC-induced demethylation. Furthermore, in studies of primary cancers, we show that the methylation-associated silencing of TIMP-3 is tumor-specific, is associated with lack of TIMP-3 protein expression, and is particularly frequent in renal cancer, in which 78% have aberrant TIMP-3 methylation, with associated lack of protein expression.
Materials and Methods
MSP.
Genomic DNA was isolated from cell lines and primary tissues, as described (13) . Approximately 1 μg of DNA was bisulfite-modified and subjected to MSP, as described (14) . MSP primer sequences that specifically recognized unmethylated TIMP-3 sequence (-91 bp to +25 bp; Fig. 1b ⇓ ) were 5′-TTTTGTTTTGTTATTTTTTGTTTTTGGTTTT-3′ (upstream) and 5′-CCCCCAAAAACCCCACCTCA-3′ (downstream) and the methylated TIMP-3 sequence were 5′-CGTTTCGTTATTTTTTGTTTTCGGTTTC-3′ (upstream) and 5′-CCGAAAACCCCGCCTCG-3′ (downstream). TIMP-3 MSP reactions were performed at a 59°C annealing temperature. Each PCR reaction (15 μl) was loaded onto 6% nondenaturing polyacrylamide gels, ethidium bromide-stained, and visualized under UV light.
Expression and methylation of TIMP-3in human cancer cell lines. A, analysis of TIMP-3 expression. RT-PCR expression analysis of TIMP-3 in colon cancer cell lines before and after 5Aza-dC treatment. Expression of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was performed for all samples to ensure the integrity of the reverse transcription (RT) reactions. The presence or absence of reverse transcriptase in RT reactions is indicated by + or −, respectively. B, TIMP-3 5′ CpG island methylation analysis. Sodium bisulfite DNA sequence analysis was performed before and after 5Aza-dC treatment in colon cancer cell lines with different TIMP-3 expression profiles. Each circle indicates a CpG site in the primary DNA sequence, and each line of circles represents analysis of a single cloned allele. ○, unmethylated CpG sites; •, methylated CpG sites; arrow, the transcription start site (18) .
Cell Culture and 5Aza-dC Treatment.
Colon cancer cell lines (RKO, SW48, and HT29) were routinely cultured in RPMI (Life Technologies, Inc.) supplemented with 10% FCS. SW480 cells were routinely cultured in McCoy’s 5A media (Life Technologies, Inc.) with 10% FCS. Cells were treated with a final concentration of 1.0 μm 5Aza-dC (Sigma Chemical Co.) for 3 days, harvested, and subjected to RNA and DNA extraction.
RNA isolation and RT-PCR.
Cytoplasmic RNA was isolated as previously described (15) . Approximately 6 μg was reverse-transcribed (16) and amplified for TIMP-3 RNA using the RT-PCR primers previously described (10) . As a control for cDNA integrity, glyceraldehyde-3-phosphate dehydrogenase expression was also analyzed (17) .
Sodium Bisulfite DNA Sequencing.
Sodium bisulfite-modified DNA (sequence –247 bp to +341 bp, relative to transcription start; Ref. 18 ), was amplified with the following primers: 5′-GGGAGTGGGGTTAGGGTGTAGA-3′ (sense) and 5′-AAACTACTACTCTCCTCTCCAAAATTACC-3′ (antisense). Amplification with these primers was conducted, as described for MSP analysis (14) , using an annealing temperature of 56°C. PCR products were cloned into the TA vector pCR2.1-TOPO and transformed into bacteria, as per the manufacturer’s instructions (Invitrogen). Plasmid DNA from isolated clones containing modified TIMP-3 sequence was purified using Wizard Plus Minipreps (Promega) and subjected to automated DNA sequence analysis (ABI automated sequencing).
Analysis of TIMP-3 Protein Expression in Normal and Malignant Kidney.
Immunoperoxidase staining of TIMP-3 protein in primary kidney tissue was performed, as described (19) , using a rabbit polyclonal antibody to human TIMP-3 (Chemicon) at a 1:1000 dilution.
Results
To determine whether TIMP-3 may be silenced in human cancers in association with aberrant 5′ CpG island methylation, we first used MSP to analyze the methylation status in and around the transcription start site of the human TIMP-3 gene in human tumor cell lines. We found that the TIMP-3 CpG island was methylated in many lung, breast, colon, and kidney tumor cell lines (data summarized in Table 1 ⇓ ). The incidence of methylation was lowest in breast cancer cell lines (29%) and highest in colon cancer cell lines (63%).
Summary of TIMP-3 5′ CpG island hypermethylation in cell lines, primary tumors, and normal tissues
We next studied, in colon cancer cell lines, the relationship of TIMP-3 expression to the density of methylation within the 5′ CpG island of the gene. We examined the methylation pattern of the region encompassing the transcription start site by genomic bisulfite sequencing of individual alleles from these cell lines. SW480 cells, which express TIMP-3 at high levels by RT-PCR analysis (Fig. 1A) ⇓ , were largely unmethylated, though there was methylation of individual CpG sites at the very 5′ portion of the region sequenced (Fig. 1B) ⇓ . HT29 cells, by RT-PCR, express TIMP-3 at substantial, but lower, levels than SW480 cells (Fig. 1A) ⇓ and showed a more complex pattern of methylation. Every allele showed a mix of methylated and unmethylated CpG sites, with the region immediately flanking transcription start (-100 to +100) having the fewest individual CpG sites methylated (Fig. 1B) ⇓ . In contrast, nearly every allele in the TIMP-3-negative RKO cells (Fig. 1A) ⇓ harbored dense methylation throughout the entire region, including the region immediately surrounding the transcription start site (Fig. 1B) ⇓ . Similarly, SW48 cells, for which TIMP-3 expression was barely detectable (Fig. 1A) ⇓ , were also very densely methylated throughout the region (Fig. 1B) ⇓ .
To determine whether methylation was responsible for the loss or decrease of TIMP-3 expression, we next treated cells with the demethylating agent 5Aza-dC. 5Aza-dC effectively restored TIMP-3 expression in RKO and SW48 cells, where the gene is densely methylated, but did not appreciably affect TIMP-3 expression in HT29 or SW480 cells, which already expressed TIMP-3 (Fig. 1A) ⇓ . Restored TIMP-3 expression in both RKO and SW48 cells coincided with the appearance of individual alleles that were predominantly unmethylated and with a greater proportion of alleles with unmethylated CpG sites within the region spanning the transcription start site (Fig. 1B) ⇓ .
Having established that TIMP-3 hypermethylation is associated with decreased expression in human cancer cell lines, we next examined whether this change occurs in primary human cancers. We explored this question with MSP analysis using primers flanking the transcription start site (-91 to +25), an area densely methylated in nonexpressing colon cancer cell lines. Forty-one normal tissue samples from the colon, kidney, and brain showed no evidence for methylation of the TIMP-3 CpG island (Table 1 ⇓ and Fig. 2 ⇓ ). By contrast, methylation around transcription start was readily detectable in many primary human cancers of the colon, breast, lung, kidney, and brain (Table 1 ⇓ and Fig. 2 ⇓ ). Thus, methylation of the TIMP-3 CpG island is a tumor-specific event. The existence of unmethylated product in these primary tumor samples may well reflect the presence of normal tissues in these nonmicrodissected samples, as well as a possible heterogeneity within the tumor sample itself for this epigenetic change (Fig. 2) ⇓ .
Methylation analysis of TIMP-3 promoter region CpG island in primary brain and kidney tissue. Lane U, amplified product with primers recognizing unmethylated TIMP-3 sequence. Lane M, amplified product with primers recognizing methylated TIMP-3 sequence. A, primary brain tumors and normal brain. Tumor samples B1-B11 and B14 contain alleles that are distinctly methylated, whereas samples B12, B13, B15, B16, and an example of normal brain (Normal) are completely unmethylated. Peripheral blood lymphocytes (PBL) are also completely unmethylated. B, primary renal cell carcinomas and normal kidney. Normal kidney (N) and cancer (C) from patient-matched samples are shown. Samples 1–6, 9, and 10 show methylation, whereas the matched normal kidney is completely unmethylated. Samples 7 and 8 show no evidence of methylation.
Notably, renal cell carcinoma, which originates from the tissue with the highest normal expression of TIMP-3 (1) , demonstrated the highest incidence of methylation. Twenty-eight of 36 (78%) renal cell carcinomas showed evidence for methylation, whereas matched normal kidney tissue from each patient was completely unmethylated (Fig. 2B ⇓ , summarized in Table 1 ⇓ ). Immunohistochemistry for TIMP-3 in primary kidney tissue revealed abundant expression of this protein in normal kidney, especially in renal tubules (Fig. 3A) ⇓ . TIMP-3 was also readily detected in the cytoplasm and cell periphery of renal cell carcinomas in which the TIMP-3 CpG island was unmethylated. Staining was evident in clear cell carcinomas that had either retained (Fig. 3B) ⇓ or lost (Fig. 3C) ⇓ the cytoplasm after tissue processing. In contrast, TIMP-3 was not detected in three carcinomas that were aberrantly methylated (Fig. 3D) ⇓ . Therefore, as in human cancer cell lines, silencing of TIMP-3 expression in primary human cancers directly correlates with methylation in the region flanking the transcription start site.
Immunohistochemistry of TIMP-3 in primary human kidney. A, normal kidney showing expression of TIMP-3 protein (diffuse brown staining in all epithelial cells). B, renal cell carcinoma, which retained an unusual amount of cytoplasm during preparation, is unmethylated at TIMP-3 and expresses the protein. C, renal clear cell carcinoma that is unmethylated at TIMP-3 and expresses the protein. Note the brown staining in residual cytoplasm and at the periphery of most tumor cells. Clear cell renal carcinomas are not easily stained for markers processed in the cytoplasm due to large quantities of glycogen and lipid that dissolve when the tissue is processed, causing a “clear” pattern, giving the tumor its name (24) . D, renal clear cell carcinoma that is methylated at TIMP-3 and does not express the protein. Only counterstain is seen in these tumor cells.
As an inhibitor of MMPs, TIMP-3 may suppress many aspects of tumor progression (1) . To determine when, during malignant progression, the TIMP-3 5′ CpG island may become methylated, we related the incidence of TIMP-3 methylation to the different stages of progression for glioblastoma and renal cell carcinoma. Examination of 77 primary brain tumors and 36 primary kidney tumors revealed that TIMP-3 methylation occurs with similar incidence through all grades (Table 1) ⇓ . Similarly, all stages of tumor development in primary kidney carcinoma also yielded uniform results with methylation occurring in 76% (13 of 17) of stage I, 88% (7 of 8) of stage II, 71% (5 of 7) of stage III, and 75% (3 of 4) of stage IV renal cell carcinomas. These data suggest that this epigenetic change most often occurs early during kidney and brain tumor progression.
Our data also suggest that brain tumors from different patient populations had markedly different frequencies of TIMP-3 methylation. More than 50% of brain tumors from Swiss patients showed TIMP-3 methylation, whereas only 20% of brain tumors from United States patients and none of the brain tumors from Japanese patients showed TIMP-3 methylation. In contrast, ongoing studies have revealed that other genes analyzed in the same tumor series, such as O6-methylguanine-DNA methyltransferase, showed no difference in the frequency of aberrant methylation between these populations (20) . These data suggest that in different patient populations, malignant progression of glioblastoma may proceed through different molecular pathways, some of which involve methylation-associated silencing of the TIMP-3 gene.
Discussion
An imbalance in proteolytic activity during tumor progression may result from alterations in the patterns of expression of both the MMPs and TIMPs. Such increased proteolytic activity may facilitate primary tumor growth, angiogenesis, invasion, and the secondary growth of metastases (1) . Within the family of TIMPs, both TIMP-1 and TIMP-2 may have effects, which while inhibiting metalloproteinase activity may stimulate cellular growth in certain cell systems (2) . Expression of TIMP-3, a secreted protein bound to the extracellular matrix, has been shown to inhibit many aspects of tumor development and metastatic progression, including growth, angiogenesis, and invasion (8, 9, 10, 11) . Decreased TIMP-3 expression has been observed in a variety of tumor cell lines (9 , 10 , 12 , 21) and has been associated with 5′ CpG island methylation in a single murine tumor cell line (12) .
Our data now show that the TIMP-3 5′ CpG island is densely methylated in a subset of cell lines from common subtypes of human cancer. Genomic bisulfite sequencing data indicate that the patterns of methylation for TIMP-3 are complex and that density of CpG island methylation, particularly just 5′ to transcription start, correlates best with TIMP-3 silencing. For instance, SW48, RKO and HT29 cells are methylated through this ∼500-bp region, but express TIMP-3 at different levels. Substantial TIMP-3 expression in HT29 cells corresponds with a greater proportion of less densely methylated alleles and with a higher proportion of unmethylated CpG sites immediately surrounding the transcription start site.
Conversely, in the SW48 and RKO cells, which express minimal to no TIMP-3 by RT-PCR, each individual allele is densely methylated as are most individual CpG sites in and around the transcription start site (Fig. 1B) ⇓ . When expression is restored in these cells after 5Aza-dC treatment, RKO and SW48 cells now show patterns of methylation that are very similar to those in HT29 cells. Individual alleles are now nearly completely devoid of CpG methylation, and the overall density of methylation for the CpG sites immediately flanking the transcription start site is decreased (Fig. 1B) ⇓ . These results indicate that expression of the TIMP-3 gene can be supported if critical regions of the CpG island remain free of methylation.
Our data also show that the methylation-associated silencing of TIMP-3 is prevalent in many primary solid tumors, corresponds to loss of TIMP-3 protein, and is tumor-specific. The restriction of this epigenetic change to tumors is most clearly evident in the 36 matched normal/tumor pairs of kidney cancer examined. It is interesting to note that the highest level of TIMP-3 expression in normal tissues exists in the kidney and brain (1) and that cancers, particularly from the kidney, have the highest frequency of TIMP-3 methylation. It is, therefore, conceivable that TIMP-3 expression may be critical for the normal growth of these tissues. Indeed, the high frequency of methylation through all grades of glioblastoma and renal cell carcinoma may provide a mechanism for loss of such control, further supporting a tumor suppressor role for TIMP-3, as recently suggested (22) . Finally, methylation of the TIMP-3 5′ CpG island segregates distinctly with different glioblastoma patient populations, suggesting that subgroups of brain tumors may progress through different malignant pathways. Indeed, studies of genetic changes such as chromosome segment loss and epidermal growth factor-receptor amplification have suggested different routes of progression for glioblastoma (23) .
Tumor-specific methylation of TIMP-3 may be a critical step during malignant progression. It is plausible that loss of TIMP-3 may abrogate normal apoptotic programs, enhance primary tumor growth and angiogenesis, invasiveness, and metastasis and possibly, therefore, contribute to all stages of malignant progression (1) . Abrogation of TIMP-3 function may be more involved in tumorigenesis than other members of the TIMP family, perhaps due to differences in function or location. TIMP-1 and TIMP-2 are soluble inhibitors, whereas TIMP-3 is insoluble and bound to the extracellular matrix. Thus, differential expression of these TIMPs may affect regional balances in MMP/TIMP activity. In fact, TIMP-1 and TIMP-2 expression has been reported to be regionally increased, compared with adjacent normal tissue, in some cancers (2) . However, functional and structural differences among the TIMP family members suggest we should view each TIMP activity individually. Finally, the high percentage of TIMP-3 methylation found in renal cell carcinomas at early stages of tumor development, and the sensitivity of MSP analysis, suggest that detection of this aberrant change may be useful as a novel marker for this disease.
Acknowledgments
We thank Drs. Fray Marshall, Jim Brooks, Stan Hamilton, Gavin P. Robertson, Jean-Pierre Issa, and Nita Ahuja for supplying tissue samples. We also thank Drs. Michael Rountree and Robert Casero for critical reading of this manuscript and helpful discussion.
Footnotes
-
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.
-
↵1 Supported by NIH Grant CA43318 and Nationalfonds Grant N-159-9-1995. J. G. H. and S. B. B. receive research funding and are entitled to sales royalties from ONCOR, which is developing products related to research described in this manuscript. The terms of this arrangement have been reviewed and approved by the Johns Hopkins University in accordance with its conflict of interest policies.
-
↵2 To whom requests for reprints should be addressed, at The Johns Hopkins Oncology Center, 424 North Bond Street, Baltimore, MD 21231. Phone: (410) 955-8506; Fax: (410) 614-9884; E-mail: hermanji{at}welchlink.welch.jhu.edu
-
↵3 Present Address: Lilly Research Labs, Eli Lilly and Company, Indianapolis, IN 46285.
-
↵4 The abbreviations use are: TIMP, tissue inhibitor of metalloproteinase; MMP, matrix metalloproteinase; 5Aza-dC, 5-aza-2′deoxycytidine; MSP, methylation-specific PCR; RT-PCR, reverse transcription-PCR.
- Received October 23, 1998.
- Accepted January 5, 1999.
- ©1999 American Association for Cancer Research.
References
Volume 59, Issue 4
