| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Biochemistry and Biophysics |
2(I) Gene Occurs in Both Cancer Cell Lines and Primary Colorectal Cancers1
Departments of Biochemistry [P. K. S., E. M. S., M. J. M., B. D. S.] and Pathology [M. J. M.], Boston University School of Medicine, Boston, Massachusetts; Boston VA Medical Center, Boston, Massachusetts [P. K. S., E. M. S., B. D. S.]; Mallory Institute of Pathology, Boston Medical Center, Boston, Massachusetts [M. J. M.]; and Chonnam University, South Korea [K. K.]
 |
ABSTRACT |
|---|
 Top ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
2(I) gene transcription start site occurs frequently in human cancer cell lines (9 of 10), including breast cancer cell lines (MCF-7 and Hs578T), hepatocellular carcinoma cell lines (SNU387, SNU449, SNU398, and PLC/PRF/5), a fibrosarcoma cell line (HT1080), and colorectal carcinoma cell lines (HCT116, SW480, and SW620). In addition, the collagen gene is more methylated in colorectal cancer tissues compared with normal mucosa. The increased DNA methylation of the collagen gene in cell lines is inversely correlated with collagen mRNA steady-state levels. Most importantly, treatment of fibrosarcoma or breast carcinoma cells with a DNA methyltransferase inhibitor, 5-aza-2'-deoxycytidine, resulted in lower methylation and reactivation of the collagen gene in a dose-responsive manner. This is the first demonstration that the collagen 2(I) gene is methylated in multiple cancer cell lines correlating with loss of collagen expression and also methylated in primary cancer tissues. These data also suggest that methylation-induced repression of collagen transcription may be a frequent occurrence in cancer. |
INTRODUCTION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
1(I) chains and one 2(I) chain. This form of collagen represents the major fibrillar component of the stroma in most solid malignancies, such as colorectal, hepatocellular, and breast carcinomas. The correct extracellular matrix in stroma is necessary for the growth of tumors. These matrix components, in particular collagen type I, can be produced by either the tumor cells or the stromal fibroblast cells. Carcinoma cells that produce type I collagen tend to have low tumorigenic potential (1)
. In benign breast cancer lesions, the type I and III collagen bundles are regularly organized, and there is detectable, though weak, expression of collagen in carcinoma cells (2)
. However, in more aggressive tumors, the stromal fibroblasts or myofibroblasts produce the collagen stroma (2)
. There is evidence that metastatic tumor cells produce factors that stimulate collagen synthesis by the stromal fibroblasts, and the carcinoma cells produce little or no collagen (1)
.
Synthesis of collagen by cells in culture is down-regulated on oncogenic transformation with viruses or with chemical carcinogens (3, 4, 5, 6)
. We have demonstrated previously (5)
that collagen 2(I) is not synthesized in a tumorigenic line, W8, after treatment of the parental liver epithelial-like cell line, K16, with the carcinogen 2-N-(acetoxyacetyl)-aminofluorine. The promoter-5' region of the 2(I) gene, COL1A2, was methylated in the DNA isolated from W8 cells (7)
. Furthermore, reporter constructs containing the COL1A2 promoter (218 bp) and 5' region of the rat and human COL1A2 first exon (54 bp) were inactivated by DNA methylation in transient transfection experiments (8)
and in vitro transcription assays (9)
. In fact, a minimal COL1A2 promoter containing the preinitiation region (-41 to +54) driving expression of the luciferase reporter gene was also inactivated by DNA methylation (9)
. The inhibition of reporter gene expression was attributable to CpG methylation of COL1A2 sequences within the first exon surrounding the transcription start site. Finally, the collagen transcription start site (-1 to +20) contains a low affinity binding site for the RFX3
family (10
, 11)
. The binding affinity of RFX1 is increased if the CpG site at +7 is methylated on the coding strand. This family of closely related proteins, RFX1–4 (12, 13, 14, 15)
, can bind methylated DNA sequences with higher affinity within a sequence-specific, 14-bp consensus sequence. Methylation-dependent binding sites have been located for RFX at the beginning of the human genes for hypoxanthine phosphoribosyl transferase, -galactosidase A, human leukocyte antigens, and the apoferritin H gene (16)
, as well as collagen (10)
, suggesting a role for this protein family in DNA methylation-induced gene repression. Most importantly, the RFX proteins repress collagen gene expression in transient transfection, as well as in vitro transcription assays (11)
.
Hypermethylation of cytosine-rich regions in promoters has been associated with the transcriptional inactivation of several genes (17 , 18) . Although there is an overall genomic hypomethylation, abnormal hypermethylation of genes has been detected frequently in cancers and associated with inactivation of tumor suppressor genes (19 , 20) . Therefore, it is important to determine whether the RFX binding site within the collagen gene is indeed methylated in cancer cells and/or tumor tissue.
This study addresses whether the COL1A2 gene is methylated in cancer cell lines and tumors at the site where RFX binds. The methylation status of both DNA strands of the COL1A2 gene has been examined in genomic DNA isolated from multiple cancer cell lines, as well as pairs of colorectal carcinomas and patient-matched normal colon tissue. The COL1A2 mRNA levels in cell lines have been measured by quantitative PCR and correlated to DNA methylation status. In addition, cells were treated with the demethylating agent, aza-dC, to decrease methylation and reactivate collagen gene transcription.
|
MATERIALS AND METHODS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Aza-dC Treatment.
Cells were plated at 200,000 cells/100-mm dish and treated with different doses (50–1000 nM) of aza-dC daily for 3 days before harvesting.
Tissue Specimens.
Matched sets of colorectal carcinomas and normal control mucosa were obtained from the Cooperative Human Tissue Network, and DNA was extracted as described earlier (21)
. Normal mucosa samples were taken at 5–10 cm from the tumor margins as matched controls to minimize possible sample differences attributable to genetic variation. Solid areas of colorectal carcinoma were carefully separated from both nontumorous elements and areas resembling residual adenomatous tissue. All samples were snap frozen immediately and stored at -80°C until extraction. Frozen tissue was cut into slices using a cryostat machine. Several slices were placed in an Eppendorf tube for DNA extraction and subsequent bisulfite modification.
The six colorectal carcinomas analyzed for COL1A2 gene methylation included three Dukes’ stage B cancers (showing invasion into or through the muscle of the bowel wall but without positive lymph nodes), one Dukes’ C cancer (with positive lymph nodes) and two stage D cancers (with liver metastases). One Dukes’ D cancer was well differentiated, whereas the remainders of cancers were moderately differentiated, including one with a mucinous phenotype. These six cancers, characterized previously for K-ras mutations (21 , 22) and p53 loss of heterozygosity (23) , were chosen to represent a mixture of cases both positive and negative for these oncogenetic changes.
Genomic DNA Isolation.
Tissue DNA was isolated by incubation of tissue sections in Tris-EDTA-sodium chloride buffer [10 mM Tris-HCL (pH 8.0), 1 mM EDTA (pH 8.0), and 10 mM NaCl] containing 0.2 mg/ml proteinase K and 0.5% sodium dodecyl sulfate at 55°C for 3 h or at 37°C overnight under moderate agitation. The resulting digested tissue was spun down briefly and incubated at 95°C for 8 min to inactivate proteinase K. Phenol was added to the lysate and mixed gently for 2 min by inverting the tube. The extract was spun down at 12,000 x g for 1 min, and the top aqueous phase was transferred to a clean tube. After repeating the phenol extraction, the DNA-containing aqueous phase was extracted two times with a chloroform and isoamyl alcohol mixture (24:1, volume for volume). The DNA-containing top phase was removed, and two volumes of 100% ethanol and 1/10 volume of 3M sodium acetate were added to precipitate DNA for 1 h in a -20°C freezer before spinning down at 12,000 x g for 15 min. The supernatant was poured off, and the white DNA pellet was washed with 1 ml of 70% ethanol, spun down, and air dried. The DNA pellet was dissolved in 100 µl of Tris-EDTA buffer (pH 8.0). DNA amounts were calculated by reading absorbance at 260 nm in a Gilford (Oberlin, OH) spectrophotometer.
DNA was extracted from cells by methods described above for tissue, except that the incubation with proteinase K-containing buffer was performed for 20 h at 50°C with shaking (80 rpm). The samples were phenol/chloroform extracted and alcohol precipitated before dissolving in Tris-EDTA buffer (pH 8.0).
Bisulfite Modification of Genomic DNA.
DNA, isolated from cells or tissue, was modified by bisulfite treatment to analyze methylation of collagen at the +7 CpG site as described previously (11
, 24
, 25)
. The bisulfite modification causes unmethylated Cs to be converted to uracil. Methylated Cs are resistant to deamination. PCR was used to amplify the modified DNA replacing uracil residues with thymine. Primers were designed using a converted sequence (Cs to Ts) in a region that did not contain any possible CpG methylation sites. Separate primers were designed to amplify the coding strand (-65 to +151) or template strand (-71 to +144). (Table 1)
. The PCR product was separated on 2% low melting agarose gel and purified using the Qiaquick Qiagen gel extraction kit (Qiagen, Valencia, CA).
|
, and their relationship to the collagen gene is diagrammed in Fig. 1
. The reaction mixture contained gel-purified PCR fragments, primers, 1 x Taq polymerase buffer, and 1 µCi radiolabeled [32P] dCTP or dTTP to identify methylated or unmethylated residues, respectively, for coding strand and [32P] dGTP or dATP to identify methylated or unmethylated residues, respectively, for template strand (Promega, Madison, WI). After denaturation at 95°C for 3 min, the primers were annealed at 40°C for 2 min (template strand) or 50°C for 2 min (coding strand) and extended at 72°C for 1 min with Taq polymerase (1 unit; Promega). The reaction was stopped by the addition of 10 µl of stop buffer. The reaction products were heated at 95°C for 2 min before loading onto a 15% acrylamide 7 M urea gel. The radioactivity of the specific bands was quantified by a Packard Instant Imager (Perkin-Elmer) and pictured by autoradiography. The percentage of methylation was calculated by dividing the amount of radiolabeled dCTP or dGTP incorporated into the primer by the total radioactivity (Fig. 2)
.
|
|
. Sequencing was performed by a DNA sequencing facility at Boston University School of Medicine.
RNA Isolation and Real-time PCR.
Total RNA was isolated from confluent cells in a p100 dish with RNeasy columns and RNase-free DNase according to the manufacturer’s protocol (Qiagen). Next, cDNA was produced using 500 ng of RNA, reverse transcriptase, and random primers using a first-strand synthesis kit (Invitrogen, Carlsbad, CA). Equal aliquots (2 µl) of cDNA were amplified according to manufacturer’s TaqMan universal (50 µl) PCR master mix protocol using real-time PCR ABI Prism 7700 (PE Applied Biosystems, Foster City, CA; Refs. 26, 27, 28
). A primer set and TaqMan probe used for collagen 2(I) RT-PCR was designed with Primer Express software (Perkin-Elmer). The data were normalized using RT-PCR 18S rRNA primers (Perkin-Elmer). The collagen primers for these experiments are found in Table 2
. The PCR conditions were as follows: (a) stage 1, 50°C for 2 min; (b) stage 2, 95°C for 10 min; and (c) stage 3, 95°C for 15 s followed by amplification at 60°C for 1 min. Stage 3 was repeated for 40 cycles. For most experiments, data were analyzed by the 2[-Delta Delta C(T)] method (29)
using Sequence Detector version 1.7 software (PE Applied Biosystems). For several assays, a standard curve was prepared by serial 10-fold dilutions of a collagen cDNA plasmid [pGGH18; from 1 µg (150 nM) to 0.1 pg (0.15 nM)]. The curve was linear over 7 logs with a 0.998 correlation coefficient. The range of samples spanned 6 logs.
|
|
RESULTS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Genomic DNA extracted from 10 cancer cell lines and two normal cell lines was modified by bisulfite technique and analyzed using MS-SNuPE assays as we have reported previously (11)
. Bisulfite treatment converts Cs to uracil in single-stranded DNA under conditions that do not alter 5-methyl-cytosine. After bisulfite modification, the collagen promoter and first exon regions were amplified by PCR using primers for each strand (Table 1)
schematically represented in Fig. 1
. To analyze methylation within the collagen RFX-binding site, specific primers (+7 primers; Fig. 1
and Table 1
) were annealed to a sequence adjacent to the +7 CpG site followed by MS-SNuPE with radiolabeled nucleotides. A control primer (control primer; Fig. 1
and Table 1
) at a site that cannot be methylated within the collagen promoter was designed to test the efficiency of bisulfite conversion. The control primer single nucleotide extension indicated that between 1 and 2% of the Cs in the promoter were left unconverted in these experiments (Fig. 2A
, bottom panels). This was considered background and subtracted from the calculated percentage of methylation.
In the first experiments, we examined the methylation status of DNA isolated from a well-characterized human fibrosarcoma cell line (HT1080) that does not synthesize collagen fibrils. HT1080 was compared with the human fibroblast cell line analyzed previously (IMR-90; Ref. 11
). Fig. 2A
contains a typical autoradiogram of the radiolabeled primers beneath a graph containing the calculated percentage of methylation with SD calculated from eight separate experiments. It is clear that the +7 site is highly methylated in the fibrosarcoma cells, especially on the coding strand (80.2%; Fig. 2A
). The template strand of the fibrosarcoma cells had less methylation (61.6%) and contained more variability between different extractions.
Next, DNA was isolated from several cell lines derived from hepatocellular, breast, and colorectal carcinomas. The majority (8 of 10) of the carcinoma cell lines had >50% methylation at the +7 site in COL1A2. The hepatocarcinoma cell lines contained the most variable methylation of the collagen gene (Fig. 2B)
, and the coding strand was more methylated than the template strand except in one cell line (PLC/PRF/5). SNU387 had the least amount of methylation and grew the slowest with a population doubling of 
48 h compared with <24 h for the other hepatocarcinoma cell lines. The MCF-10A cell line, derived from fibrocystic mammary tissue, is considered an immortalized differentiated breast epithelial cell line similar to luminal ductal cells (32
, 33)
. This cell line also grows slowly and has low methylation at the +7 collagen site (Fig. 2C)
. On the other hand, the breast carcinoma cell lines, MCF-7 and Hs578T, contained a highly methylated (70–80%) collagen gene (Fig. 2C)
. Finally, the collagen gene in all colorectal carcinoma cells (HCT116, SW480, and SW620) was almost completely methylated on both strands (Fig. 2D)
. IMR-90 methylation on the coding strand was significantly different from all of the cells used in this study except SNU 387.
DNA methylation is often spread in large regions of genes. Therefore, we confirmed the methylation status flanking the +7 collagen site by sequencing purified PCR products from several cell lines (Fig. 3)
. The bisulfite unconverted sequence from +51 to +76 is presented at the top of the three representative sequencing tracings in Fig. 3A
. In the top collagen sequence, all of the C residues are converted to T residues, indicating that there are no methylated sites in this IMR-90 DNA sample. These C residues were scored as 0% methylated in Fig. 3B
, because there was only a baseline peak of C. In the second collagen sequence (PLC/PRF/5), three C residues at CpG sites are partially converted to T, suggesting partial methylation. The first and third C residues in the second sequence were scored as 50% in Fig. 3B
, because the sequence was considered N with equal amounts of C and T. The middle C in the second collagen sequence was scored as >50% methylated (Fig. 3B)
with a C peak clearly larger than a T peak. On the other hand, in the third collagen sequence (HT1080), all these C residues have primarily remained unconverted with baseline T peaks indicating high methylation at these sites. These were all scored as 100% methylated (Fig. 3B)
. Fig. 3B
shows a representation of all of the CpG sites in the coding strand sequenced in this manner. If the +7 CpG site was >50% methylated, the majority of the surrounding CpG sites was also methylated. Again, the hepatocarcinoma cell lines contained more variability in the methylation pattern than the other carcinoma cell lines tested. The sequencing data at the +7 site correlated very well with the MS-SNuPE data. The two cell lines scored as 0% methylation at the +7 site had MS-SNuPE values of 3%. The one scored as <50% had a SNuPE value of 42%. The four cell lines scored as >50% had SNuPE values ranging from 68 to 80%. Finally, the three cell lines scored as 100% had SNuPE values >87%.
|
). At least three samples in duplicate were analyzed for each cell line (bar graph; Fig. 4
). A standard curve using collagen cDNA indicated that IMR-90 produced 15 Fmol of mRNA, whereas the low levels of collagen mRNA in MCF-7 or the colorectal carcinoma cells contained 10-3 aMol range of collagen mRNA. Because these values were so low, 5-fold more RNA was analyzed. HCT116 and SW620 were below the linear range for this assay (data not shown).
|
) with two exceptions, MCF-10A, Hs578T, both of which are breast tissue cell lines. The overall average methylation of the +7 site for both strands of MCF-10A DNA (41%) was lower than the breast carcinoma cell lines [Hs578T (64%) and MCF-7 (74%)] and similar to one of the hepatocarcinoma cell lines, SNU449 (39%). Nevertheless, the collagen mRNA expressed by these cell lines was much different. MCF-10A mRNA levels were much lower than expected based on methylation status. On the other hand, the Hs578T cell line had more collagen expression than expected. This suggests that mechanisms other than methylation also repress collagen genes.
Methylation Status of Collagen Gene in Pairs of Colorectal Cancer and Normal Colon Mucosal Tissues.
To determine whether the promoter for collagen is methylated in cancer tissues, we analyzed pairs of colorectal carcinoma and normal tissues from four patients, as well as two additional colorectal carcinomas. These samples had already been analyzed for Ras (22
, 34)
and p53 mutations (23)
, as well as for protease levels (35, 36, 37, 38)
. Three of four colorectal carcinomas contained more methylated collagen genes than the normal tissue. In one pair, both samples contained the same methylation pattern. The mean percentage of methylation for the collagen gene from normal tissue was 5.5%. Five of six colorectal cancers had higher collagen methylation than mean values from normal tissue (Fig. 5)
. The mean percentage of methylation of the collagen gene in the cancerous tissue was 19%. One sample from a cancer tissue had no methylation.
|
2(I) Expression Is Reactivated by Demethylation Using aza-dC.
. The methylation status of both strands assessed by bisulfite/MS-SNuPE analysis demonstrated that methylation decreased at the +7 site between 25 and 30% in a dose-responsive manner (Fig. 6B)
. This was consistent in three experiments. In more recent experiments using lower density HT1080 cells, the collagen mRNA levels increased 200-fold, and the methylation levels were reduced 70%. When MCF-7 or Hs578T cells were treated with 1 µM aza-dC, collagen expression increased 5-fold. When colon carcinoma cell lines (HCT116 and SW480) were treated with 1 µM aza-dC, collagen mRNA was detected (6.4 pg of HCT116-8 and pg of SW480), whereas untreated cells contained no detectable mRNA. Even the human lung fibroblast collagen mRNA levels increased 3-fold after treatment with 1 µM aza-dC. Therefore, the expression of collagen is related, in part, to the methylation status of the gene surrounding the transcription start site.
|
|
DISCUSSION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
3
4)
. Previously, we demonstrated that the collagen 2(I) transcription start site has a binding site for RFX (10
, 11)
. This protein family represses collagen transcription and binds with higher affinity when the C at +7 is methylated on the coding strand or mutated from C to T (9
, 11)
. The present study establishes that the +7 site, as well as several CpG sites in the surrounding area, become methylated in cancer cell lines and tumors. In cancer, the pattern of methylation is altered with overall genome-wide hypomethylation and regional hypermethylation (39
, 40)
. Recent investigations of the hypermethylated regions have focused on so-called "CpG islands," which are short regions (0.5–4 kb) with high numbers (>100) of methylated CpG sites that are normally not methylated. These regions are usually within promoters of genes. There are multiple examples of genes that are aberrantly methylated in cancer, such as cell cycle-related (RB1, p16INK4A), DNA repair (hMLH1, MGMT), adhesion molecule (E cadherin), and extracellular matrix (thrombospondin, tissue inhibitor of metalloproteinase 3) genes (for review, see Ref. 19
). Methylation of genes in cancer is found in virtually every type of neoplasm. It should be noted that the region surrounding the COL1A2 transcription start site is not a typical CpG island, according to the usual definition. Therefore, this promoter may become methylated more easily than a protected CpG island promoter. We have demonstrated previously that the COL1A2 gene is methylated in the promoter 5' region in a chemically transformed tumorigenic cell line, W8 (7)
. Others have demonstrated that SV40-transformed human fibroblasts contained a methylated COL1A2 gene (41)
. This study is the first demonstration that COL1A2 methylation occurs in cells from cancers and in primary cancer tissues. Overall, the methylation status of COL1A2 at the transcription start site was inversely proportional to collagen gene expression. It has become increasingly apparent that aberrant promoter methylation is associated with loss of gene expression. There are several possible mechanisms for methylation-mediated gene repression. The methyl groups on DNA could interfere with binding of specific transcription factors or enhance binding of a repressor and/or methyl-CpG-binding proteins. In this study, we have examined a binding site for the RFX protein family. RFX1 was the first protein to be recognized as a methyl-DNA-binding protein, called MDBP (42) . Later, it was reported to be a sequence-specific binding factor called RFX1 (15) . Several members in this family (RFX1–3) are sequence-specific transcription factors that maintain a higher binding affinity to a sequence if the middle C is either methylated or converted to a T (14) . The protein can both activate and repress transcription, depending on the context (30) . RFX1 represses collagen transcription by 50% when it is overexpressed in transient transcription assays (11) . When the +7 site is mutated from C to T, mimicking the methylated state, this single mutation caused almost complete loss of transcriptional activity in transfection and in vitro transcription assays (10) . This suggests that RFX1 is an active repressor, especially if this +7 site is methylated or mutated. There is another sequence-specific yet methylation-responsive repressor called Kaiso (43 , 44) . This protein contains Kruppel zinc finger domains (Zn) and a protein domain (POZ) that interacts with p120 catenin. Kaiso requires either methylated sequences or Ts within its specific binding sites for high affinity binding and has been demonstrated to be a transcriptional repressor of an extracellular matrix-degrading enzyme, matrilysin (45) .
Repressor DNA-binding proteins act through interactions with corepressor complexes that contain multiple proteins, including DMNTs, e.g., another POZ/Zn finger protein, RP58, mediates sequence-specific transcription repression and interacts directly with DMNT3a (46 , 47) . Hormone receptors, such as estrogen receptor or retinoic acid receptors, bind to corepressor complexes when there is no ligand (48 , 49) . Retinoblastoma protein, a cell cycle regulator, represses E2F mainly by recruiting chromatin-remodeling factors (histone deacetylases), including DNMT1 (50 , 51) . The chimeric protein promyelocytic leukemia-retinoic acid receptor that promotes leukemia induces silencing by recruiting DMNTs to target promoters and causes hypermethylation of promoters. Finally, there is a family of proteins that binds methyl-CpG sequences and shares a homologous methylated DNA-binding domain. Four of these proteins (MBD1, MBD2, MBD3, and MeCP2) have been implicated in methylation-dependent repression through interactions with corepressor complexes (for reviews, see Refs. 17 , 52 , and 53 ). The DMNTs in some cases can act cooperatively to repress gene transcription. The corepressor complexes alter chromatin structure (53) through deacetylation of histones by histone deacetylases (54) . The repressor complexes generated by DNA methylation compact chromatin, which inhibits transcription.
In addition, direct interference of a methyl group with a transcription activator has also been described for several transcription factors (55
, 56)
. Our initial gel shift experiments were performed to investigate this possibility (9)
. We demonstrated that when the region between -41 and +54 of the mouse genome was methylated in vitro, there was a decrease in binding of TATA-binding protein in addition to the increased binding of another protein. However, there are no conserved methylation sites near TATA box in the mouse genome so that this could not be interference by a methylated base. Therefore, as shown in our model, we hypothesize that RFX proteins bind to the +7 CpG-methylated site and interfere with preinitiation complex formation (Fig. 7)
. Most likely, RFX recruits a corepressor complex similar to Kaiso, retinoblastoma protein, or RP58 that spreads methylation and alters chromatin structure.
|
. This cell line is an established cell line that has been characterized previously for its collagen types. HT1080 produces type IV collagen and normally does not produce fibrillar collagens (57)
. On the other hand, the type I collagen promoter is active in this cell line, suggesting that the required transcription factors are available (57
, 58)
. Real-time PCR was able to measure a small amount of collagen mRNA (0.15 aM) in these cells compared with IMR-90 cells (14 fM). One way to test whether DNA methylation of a gene mediates transcription repression is to treat cells with aza-dC to induce re-expression of genes (59)
. This drug incorporates into the genomic DNA, leading to covalent trapping of DNA methyl transferases, thus depleting the cells of methyltransferase enzyme activity (60)
. Aza-dC induces collagen gene expression in HT1080 cells in a dose-dependent manner (Fig. 6A)
with a 40-fold increase in collagen mRNA observed together with a 25% reduction in the gene methylation (Fig. 6B)
. Thus, collagen gene methylation that represses collagen gene expression can be reversed with aza-dC treatment.
To determine whether methylation and silencing of the collagen gene occurred in cells grown from carcinomas, DNA and RNA from several hepatocellular, breast, and colon carcinoma cell lines were examined. The correlation between methylation and silencing of collagen expression was most pronounced in the series of hepatocellular and colorectal carcinoma cell lines (Fig. 4)
. The colon cancers had the highest amount of methylation with no detectable collagen gene expression. The hepatocellular carcinoma cell lines had intermediate methylation and gene expression. The methylation of the collagen gene and collagen gene silencing did not correlate well in the breast cancer cell lines. The MCF-10A cell line was derived from fibrocystic breast tissue, not tumor tissue. The cells are well differentiated, especially when they are maintained in the epithelial cell media used in this study. However, these cells do not senesce and, therefore, may be considered an immortalized breast cell line isolated from nontumor tissue (32
, 33)
. This cell line contained a low amount of methylation compared with the tumor-derived breast carcinoma cell lines (MCF-7 and Hs578T). However, collagen gene expression has been silenced by other methods possibly related to the "immortalization" procedure through chromatin suppression. The transcription rates may correlate with methylation levels in these cell lines, but the mRNA turnover rates may be altered so that the steady-state mRNA levels do not correlate well with methylation status. In conclusion, 90% of the tumor cells that were examined had considerable methylation in the collagen gene at the transcription start site.
To determine whether the same type of collagen gene methylation that occurred in cancer cell lines was also present in primary human colorectal carcinomas, we have analyzed genomic DNA extracted from six colorectal carcinomas and four patient-matched control mucosa samples (Fig. 5)
. This set of six cancers had been characterized previously for the presence of K-ras mutations (22)
and p53 loss of heterozygosity (23)
and were chosen to represent a mixture of cases both positive and negative for these particular oncogenetic changes. The fact that five of six primary human colorectal carcinomas demonstrated a higher percentage of methylation of the COL1A2 gene than the average percentage of methylation of four matched control tissues supports the concept that methylation of this gene occurs not only in cancer cell lines but also in primary human carcinomas.
CpG island methylation has been described previously as a mechanism for transcriptional inactivation of tumor suppressor genes in subsets of primary colorectal cancers (61 , 62) . It has also been shown that some but not all colorectal cancers demonstrate this CIMP. Subsets of colorectal cancers that are CIMP positive represent 80% of sporadic cancers with microsatellite instability (61) , as well as additional sporadic colorectal cancers that are typically diploid and commonly associated with a proximal location in the colorectum (i.e., right sided). These CIMP-positive colorectal cancers are thought to develop by a pathway independent of the chromosomal instability pathway that is more often associated with left-sided, aneuploid colorectal cancers that show sequential changes in APC, ras, and p53 genes classically associated with colorectal tumorigenesis (63) . Thus, the lack of methylation for the COL1A2 gene in one of the cases studied here (a Dukes’ D cancer shown previously to have both a K-ras mutation and p53 loss of heterozygosity) supports previous data indicating that cancers with K-ras mutations and p53 loss of heterozygosity are less likely to demonstrate a CpG island methylation phenotype (64) . In these cases, methylation of the COL1A2 gene and/or other important cancer-related genes may not be required for malignant transformation or progression, because these cancers develop by the accumulation of mutations that result from chromosomal instability. In other cancers studied, methylation of the COL1A2 gene may occur together with methylation of other important genes, as a means of generating the malignant phenotype. The four cancers studied here that showed the highest percentage of methylation of the COL1A2 gene were all right-sided colorectal cancers, in keeping with previous reports that colorectal cancers positive for CpG island methylation are most likely to be proximal in location.
What is the significance of decreased collagen synthesis in cancer? We have demonstrated previously that the collagen 2(I) gene is repressed and methylated in a tumorigenic line, W8, after treatment of the parental liver epithelial-like cell line, K16, with the carcinogen 2-N-(acetoxyacetyl)-aminofluorine (5)
. These cells produce only 1(I) trimer collagen. W8 cells grow in soft agar, which is considered a hallmark of the transformed phenotype. In fact, s.c. injection of W8 cells in irradiated rats produces tumors. When these cells were sequentially grown in cell culture from tumors, the collagen synthesis decreased with increased tumor potential (65)
. Other investigators (4
, 6) have demonstrated that decreased collagen production by tumorigenic cells correlates with faster cell growth, less adhesion to substratum, and increased tumorigenicity of the transformed cells.
Is collagen a candidate tumor suppressor gene? To directly assess the role of 2(I) chains, we expressed the 2(I) chains in W8 cells (66)
. The 2(I)-expressing W8 cells adhere more firmly to the substratum, maintain slower growth kinetics, and form fewer colonies in soft agar than the mock-transfected W8 cells. Similar experiments (67)
using stable transfectants of 2(I) chains in a vKRas-transformed mouse fibroblast cell line produced cells that displayed a partial restoration of type I collagen production. These cells maintained a flatter morphology with increased adherence to the substratum, a reduced ability to clone in soft agar, slower growth kinetics, and suppression of tumorigenicity in nude mice. This provides important evidence that decreased gene expression of 2(I) chains contributes to the transformed phenotype and that increased 2(I) chains can suppress the transformed phenotype. Therefore, collagen 2(I) gene may be a candidate tumor suppressor gene and play an important role in tumorigenesis.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
1 Supported in part by Veterans Administration merit review program, NIH, NHLBI P01-HL56386, and R01-HL68094. 
2 To whom requests for reprints should be addressed. Phone: (617) 638-4159; Fax: (617) 638-5339; E-mail: smith{at}biochem.bumc.bu.edu
3 The abbreviations used are: RFX, regulatory factor for X box; aza-dC, 5-aza-2'-deoxycytidine; Cs, cytosines; MS-SNuPE, methylation sensitive single nucleotide primer extension; PLC, phospholipase C; RT-PCR, reverse transcription-PCR; CIMP, CpG island mutator phenotype; DMNT, DNA methyltransferase.
Received 10/ 8/02. Accepted 2/ 3/03.
|
REFERENCES |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  Y. Xu, J. A. Harton, and B. D. Smith CIITA Mediates Interferon-{gamma} Repression of Collagen Transcription through Phosphorylation-dependent Interactions with Co-repressor Molecules J. Biol. Chem., January 18, 2008; 283(3): 1243 - 1256. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. Muthusamy, S. Duraisamy, C. M. Bradbury, C. Hobbs, D. P. Curley, B. Nelson, and M. Bosenberg Epigenetic Silencing of Novel Tumor Suppressors in Malignant Melanoma Cancer Res., December 1, 2006; 66(23): 11187 - 11193. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. Ibanez de Caceres, E. Dulaimi, A. M. Hoffman, T. Al-Saleem, R. G. Uzzo, and P. Cairns Identification of novel target genes by an epigenetic reactivation screen of renal cancer. Cancer Res., May 15, 2006; 66(10): 5021 - 5028. [Abstract] [Full Text] [PDF]
|
|
, unmethylated site; 
, <50% methylated site; 
, 50% methylated site; 
, >50% methylated site; 
, fully methylated site.
