Methylation of a CpG island in a promoter region causes silencing of its downstream gene, and many tumor-suppressor genes, such as RB1, CDKN2A (p16, INK4A), VHL, MLH1, CDH1 (E-cadherin) and BRCA1, are reported to be inactivated by aberrant methylation of their promoter CpG islands in various cancers.1 In colorectal cancers, silencing of CDKN2A,2MLH1,3PTEN,4HIC1 (hypermethylated in cancer 1)5 and SFRP16 have been reported. Especially, HIC1 and SFRP1 genes demonstrated the presence of a novel class of tumor-suppressor genes.7, 8HIC1 inactivation promotes tumorigenesis by activating the stress-controlling protein SIRT1 and thereby attenuating p53 function.9SFRP1 inactivation promotes colon carcinogenesis by enhancing WNT signaling even in the presence of downstream mutations of APC or CTNNB1 (β-catenin).8 These genes tend to be inactivated solely or mainly by promoter methylation, and the inactivation occurs in early stages of carcinogenesis. Their inactivation can confer a growth advantage to a cell in cooperation with alterations of other genes in the signal transduction pathways, such as TP53 (p53), APC and CTNNB1, and are considered to provide a “milieu,” where a cell with a mutation of a tumor-related gene can gain a clear growth advantage.10
Therefore, it is important to clarify more genes that are silenced in colorectal and other cancers. To make a genome-wide screening for genes silenced by their promoter methylation, various technologies have been developed.11 Among these technologies, the cDNA microarray analysis of cells before and after treatment of cells with a demethylating agent (5-aza-2′-deoxycytidine; 5-aza-dC) (chemical genomic screening) is technically simple and powerful in identifying silenced genes with normally abundant expression.6 This technique was successfully applied to identify the SFRP1 tumor-suppressor gene8 and hypermethylation of cyclin A1 in head and neck cancers.12
In this study, we performed a chemical genomic screening using HCT116 human colon cancer cells, and identified UCHL1, whose product has both ubiquitin hydrolase activity and ubiquitin ligase activity, as a gene silenced in human colorectal cancers and also in human ovarian cancers.
Material and methods
Cell lines, tumor samples and DNA/RNA extraction
Normal human colon epithelial cells (CRL-1790 and CRL-1831) and human colorectal cancer cell lines (HCT116, DLD-1, LoVo, SW480, Caco-2, HT-29, LS174T, LS180, T84 and WiDr) were purchased from the American Type Culture Collection (Manassas, VA). Two additional colorectal cancer cell lines (SW837 and CC1) were utilized. SW837 was gifted by the Japanese Collection of Research Bioresources (Tokyo, Japan), and CC1 was obtained through the Aichi Cancer Center (Aichi, Japan). Ovarian cancer cell lines were obtained as reported,13 and human ovarian surface epithelial (HOSE) cells immortalized by the papilloma virus E6 and E7 were gifted by Dr. Tsao.14
Colorectal cancer samples were obtained from patients who underwent colectomy or polypectomy at the Koyo Hospital with informed consents. Ovarian cancer samples and endometrial cyst samples were obtained from patients who underwent ovariectomy at the University of Tokyo Hospital with informed consents. All samples were kept frozen until DNA/RNA extraction. DNA was extracted by a standard phenol/chloroform extraction and ethanol precipitation, and total RNA was isolated using ISOGEN (Nippon Gene, Tokyo, Japan).
Cells were seeded at a density of 3 × 105 cells/10 cm dish on Day 0. On Day 1, the medium was changed to one containing 5-aza-dC (Sigma-Aldrich, St Louis, MO) at a concentration specified. 5-Aza-dC was freshly dissolved in PBS, and filtered through a 0.2-μm filter. The cells were harvested on Day 4.
Oligonucleotide microarray analysis
Ten micrograms of total RNA was used to prepare a probe, and hybridized to GeneChip Human Genome U133A arrays, which contained probe sets for 18,400 transcripts, including 14,500 known genes (Affymetrix, Santa Clara, CA). The arrays were scanned with a GeneArray scanner (Hewlett-Packard, Palo Alto, CA), and the scanned images were processed using an Affymetrix GeneChip Analysis Suite (Version 4.0.1) to obtain the “signal intensity” for each probe set. The values were scaled so that the average signal intensity of all probe sets would be 500.
Methylation-specific PCR and bisulfite sequencing
Five hundred nanograms of DNA, digested with BamHI restriction enzyme, was treated with sodium bisulfite as previously reported,15 and was dissolved in 20 μl of TE buffer (10 mM Tris-HCl and 1 mM EDTA, pH 8.0). Methylation-specific PCR (MSP) was performed using 0.5 μl of the solution and primers specific to the sequence produced from methylated (M) or unmethylated (U) DNA. The primer sequences and PCR conditions were: (M), 5′-TCGTATTTATTTGGTCGCGATC-3′ (sense), 5′-CTATAAAACGCCGACCAAACG-3′ (antisense), 32 cycles with annealing at 62°C; (U), 5′-GGTTTGTATTTATTTGGTTGTGATT-3′ (sense), 5′-CAACTATAAAACACCAACCAAACA-3′ (antisense), 34 cycles with annealing at 60°C. As controls, we prepared fully methylated DNA by methylating DNA of cancer cells with SssI-methylase (New England Biolabs, Beverly, MA) and fully unmethylated DNA by twice amplifying DNA of normal cells with phi29 DNA polymerase (GenomiPhi DNA Amplification Kit; Amersham Biosciences, Piscataway, NJ).16 A number of PCR cycles that will yield a minimal visible band was determined using these fully methylated DNA (for M primers) and fully unmethylated DNA (for U primers), and 4 more cycles were added for actual analysis of test samples.
A 386-bp fragment that spanned from the 5′ promoter region to exon 1 (from nt. 28,626 to nt. 29,011; GenBank accession number AC095043) was amplified by PCR using 0.5 μl of the bisulfite-treated DNA and primers 5′-AGTGAGATTGTAAGGTTTGG-3′ (sense) and 5′-CACTCACTTTATTCAACATCTAA-3′ (antisense) with annealing at 58°C. The PCR product was cloned into pGEM-T Easy TA Vector (Promega, Madison, WI). For each sample, more than 10 clones were cycle-sequenced using T7 primer, 5′-TAATACGACTCACTATAGGG-3′, and an Applied Biosystems 310 sequencer. Proportional amplification of methylated and unmethylated DNA molecules was confirmed by observing their equal amplification in a control sample that contained equal amounts of fully methylated DNA and fully unmethylated DNA.
cDNA was synthesized from 3 μg of total RNA treated with DNaseI (Ambion, Austin, TX) using the random hexamer (Promega) and Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA). Real-time PCR analysis was performed using an iCycler iQ detection system (Bio-Rad Laboratories, Hercules, CA) with 200 nM of primers and 10× SYBR Green I nucleic acid gel stain (BioWhittaker Molecular Applications, Rockland, ME). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used for normalization of mRNA expression levels. Primer sequences were as follows: UCHL1, 5′-GGATGGCCACCTCTATGAAC-3′ (sense), 5′-AGACCTTGGCAGCGTCCT-3′ (antisense) (annealing at 55°C); GAPDH, 5′-AGGTGAAGGTCGGAGTCAACG-3′ (sense), 5′-AGGGGTCATTGATGGCAACA-3′ (antisense) (55°C). To quantify the number of molecules of a specific gene in a sample, a standard curve was generated using samples that contained 10 to 108 copies of the gene.
Total protein lysate was prepared with RIPA buffer (20 mM Tris-HCl (pH 7.6), 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% NP-40) in the presence of a protease inhibitor cocktail (Sigma-Aldrich). The protein lysate (5 μg/lane) was separated by 12% SDS-PAGE and transferred to a PVDF membrane (Immobilon, Millipore, Bedford, MA). After blocking, UCHL1 was detected by anti-UCHL1 primary antibody (1:500 dilution, Biogenesis, England). After washing, the membrane was incubated with the secondary antibody (1:5,000 dilution) affinity-purified rabbit antibody to IgG conjugated with peroxidase (Chemicon International, Temecula, CA). The specific complexes were visualized using ECL (Amersham Biosciences).
Identification of genes upregulated by 5-aza-dC in HCT116 colon cancer cell line
GeneChip analysis was performed using HCT116 human colon cancer cells before and after 5-aza-dC treatment. Among the 14,500 known genes, we identified 7 genes that were upregulated (≥8-fold) by 5-aza-dC treatment and expressed at high levels (signal intensity ≥ 1,000) after 5-aza-dC treatment (Table 1). Genes located on the X chromosome and cancer testis antigens, such as MAGE and GAGE, were excluded. Among the 7 genes, 5 genes had CpG islands in their putative promoter regions. UCHL1 was known to be involved in regulation of cellular ubiquitination,17, 18 and since disturbance in ubiquitination is deeply involved in human cancers,19 we decided to focus our analysis on the UCHL1 gene.
Table I. Genes Whose Expression Increased IN 5-aza-dC-Treated HCT116
Signal log ratio
5′ CpG island
Signal log ratio was expressed as the log2 ratio of the change in expression level between nontreated and 5-aza-dC-treated cells.
Ubiquitin carboxyl-terminal esterase L1
Perforin 1 (pore forming protein)
Leucine rich repeat neuronal 4
H2B histone family, member S
Dehydrogenase/reductase (SDR family) member 2
Silencing of the UCHL1 gene in HCT116 colon cancer cell line
The UCHL1 gene has a large CpG island covering its reported promoter region20 and exon 1 (Fig. 1a). The methylation status of its promoter CpG island was examined by MSP in normal human colon epithelial cells (CRL-1790 and CRL-1831), HCT116 colon cancer cell line and 5-aza-dC-treated HCT116 cells (Fig. 1b). Only unmethylated DNA molecules were detected in normal human colon epithelial cells, whereas only methylated DNA molecules were detected in HCT116 cells. When HCT116 cells were treated with 5-aza-dC, increasing amounts of unmethylated DNA molecules were detected with increasing doses of 5-aza-dC. Methylation status in CRL-1790 and HCT116 cells (untreated and 5-aza-dC treated) was confirmed by bisulfite sequencing (Fig. 1c). In the CRL-1790 colon epithelial cells, all the DNA molecules were unmethylated, although some scattered methylated CpG sites were present. On the other hand, in HCT116 cells, the entire CpG island was densely methylated. After the 5-aza-dC treatment, appearance of unmethylated DNA molecules was observed.
The reexpression of mRNA (Fig. 1d) and protein (Fig. 1e) in accordance with the demethylation was analyzed by quantitative RT-PCR and immunoblotting. Both mRNA expression and protein expression were restored by treatment of HCT116 cells with increasing doses of 5-aza-dC. These data showed that methylation of the promoter CpG island caused the loss of UCHL1 mRNA and protein expression in HCT116 cells. The reproducibility of these data was confirmed by 3 independent treatments of HCT116 cells with 5-aza-dC.
UCHL1 silencing in colorectal cancer cell lines and primary colorectal cancers
In addition to HCT116 cells, silencing of UCHL1 was analyzed in 11 other colorectal cancer cell lines. MSP analysis of all 12 colorectal cancer cell lines showed that 8 cell lines, including HCT116, had only methylated DNA molecules, whereas LS174T, LS180 and T84 cells had both methylated and unmethylated DNA molecules, and that Caco-2 had only unmethylated DNA molecules (Fig. 2a). To confirm these results, bisulfite sequencing was performed for Caco-2 (only unmethylated DNA), DLD-1 and SW480 (only methylated DNA) (Fig. 2b). In DLD-1 and SW480, the entire CpG island was densely methylated. In Caco-2, most DNA molecules were unmethylated, with only limited methylation of CpG sites, and the region was considered to be unmethylated as a CpG island. Accordingly, the UCHL1 mRNA and protein were detected in Caco-2 cells (Figs. 2c and 2d), but not in any of the other 11 colorectal cancer cell lines. The fact that cell lines with methylated DNA molecules did not express UCHL1 mRNA lends further support that its silencing occurred by methylation of its promoter CpG island.
Methylation of the UCHL1 promoter CpG island was analyzed by MSP in 17 primary colorectal cancers (Fig. 2e). Methylated DNA molecules were detected in 8 of the 17 samples (47%). Notably, methylated DNA molecules were identified in the noncancerous tissue in Case 7. Further, UCHL1 methylation was confirmed by bisulfite sequencing in 3 representative cancers (Cases 2, 4 and 12) (Fig. 2f). DNA molecules with methylation of most CpG sites (densely methylated DNA molecules) were detected in all 3 cancers, and confirmed the presence of methylation that can repress gene transcription. However, the proportion of densely methylated DNA molecules was variable in clinical samples, which was considered to be due to the fact that only subsets of tumor cells had UCHL1 methylation and that the normal cells with unmethylated UCHL1 were contaminated.
UCHL3, a member of the ubiquitin C-terminal hydrolase family, has a similar activity to UCHL1 and is ubiquitously expressed in various tissues.21, 22 Therefore, we analyzed methylation of the UCHL3 promoter CpG island, but it was completely unmethylated in all the 12 colorectal cancer cell lines (data not shown).
UCHL1 promoter methylation in ovarian cancer cell lines and primary ovarian cancers
It is known that UCHL1 is expressed at high levels in ovarian tissue.18 Therefore, we analyzed UCHL1 silencing in 13 ovarian cancer cell lines and 17 primary ovarian cancers. While the UCHL1 promoter CpG island was not methylated in HOSE cells, 6 (OV-90, MCAS, RMUG-L, RMG-I, RTSG and TYK-nu) of 13 ovarian cancer cell lines had only methylated DNA molecules (Fig. 3a). The methylation status of UCHL1 promoter CpG island and the expression levels of mRNA and protein had good correlation (Figs. 3b and 3c).
Using 17 primary ovarian cancers, methylation of the UCHL1 promoter was analyzed by MSP. Methylated DNA molecules were detected in 1 of the 17 samples (6%), while they were not detected in 5 endometrial cyst samples (Fig. 3d). This showed that UCHL1 promoter methylation is also present in ovarian cancers, but that its incidence was not as high as that in colorectal cancers.
Silencing of UCHL1, also known as PGP9.5, in colorectal cancers was identified for the first time in this study as a result of chemical genomic screening. There was a good correlation between the methylation status of UCHL1 promoter CpG island and the level of UCHL1 mRNA expression. UCHL1 silencing in cancers was first discovered in pancreatic cancers by chemical genomic screening,23 and then in head and neck cancers.12 Recently, UCHL1 methylation was shown to be associated with a poor prognosis in esophageal squamous cell carcinomas.24 In contrast, in pancreatic, colorectal and lung cancers, overexpression of UCHL1 protein was reported,25, 26, 27 and its overexpression was associated with a poor prognosis in pancreatic and colorectal cancers. However, these studies did not analyze UCHL1 methylation, and further investigation is necessary to clarify whether methylation or overexpression of UCHL1 protein is associated with poor prognosis in pancreatic and colorectal cancers. In our study, among the 12 human colorectal cancer cell lines analyzed, UCHL1 methylation was absent only in Caco-2 cells. There is a possibility that the Caco-2 cell line was established from a subset of tumor cells that developed from a milieu where UCHL1 silencing was lacking, while other colon cancers arise from a milieu where UCHL1 was silenced.
UCHL1, located on chromosome 4p14, was originally identified as a member of a gene family whose products hydrolyze small C-terminal adducts of ubiquitin to generate the ubiquitin monomer. Its activity has been tested using synthetic carboxy-terminal ethyl ester derivatives of ubiquitin, but its physiological substrate has not yet been identified. At the same time, UCHL1 possesses a dimerization-dependent ubiquityl ligase activity,17 and can also elongate ubiquitin half-life in vivo.18 These findings strongly indicated that UCHL1 plays various important roles in maintaining appropriate ubiquitin levels within a cell and thus protein levels that undergo ubiquitin-dependent degradation. In neurons, UCHL1 dysfunction is known to be involved in familial Parkinson's disease.28, 29In vivo analysis of UCHL1-deficient mice suggested that UCHL1 functions as a regulator of apoptosis in neurons30 and also in germinal cells during spermatogenesis.31 However, the functional consequences of UCHL1 silencing in cancer cells have yet to be determined.
Regarding the ubiquitin-proteasome system, to which UCHL1 belongs, its important roles in various cellular processes, such as cell cycle, apoptosis and intracellular signaling and its disturbances in cancer development are well recognized.19, 32 Many deubiquitination enzymes are known to regulate tumor suppressors and other critical proteins involved in cancer development.33 Cellular levels of p53 are controlled by its own ubiquitination and by ubiquitination and deubiquitination of its regulators.34 Cell cycle regulatory proteins, such as p27, are regulated through their ubiquitination and degradation.35 Chromosomal stability may be compromised by haploinsufficiency of the FBXW7/hCDC4 gene, which encodes an ubiquitin ligase.36 The CHFR gene, silenced in 40% of primary colon cancer,37 possesses ubiquitination activity against mitotic kinase Aurora A.38
The ubiquitin-proteasome system is disturbed in cancers,19, 32 and UCHL1, an important member of the system, was silenced in some colorectal and ovarian cancer cell lines and primary tumors, suggesting that UCHL1 silencing could provide a growth advantage to a cell. If UCHL1 methylation was present in all the tumor cells (possibly for Cases 4 and 12; Fig. 2f), the methylation may have provided a milieu for carcinogenesis. If it was present in a subset of tumor cells (Case 2), it could confer further growth advantage to a tumor cell. At the same time, a possibility still remains that methylation of UCHL1 is without biological effects, and occurs merely as a “by-stander,” a consequence of malignant transformation. It is observed that genes with no or low transcription tend to be methylated as by-standers.11, 39 However, UCHL1 has abundant transcription in the colon and ovarian epithelium, and the remaining possibility does not seem high.
In conclusion, it was shown here that UCHL1 is inactivated in human colorectal and ovarian cancers by methylation of its promoter, and the presence of disturbed cellular ubiquitin levels was suggested. These data support the possible functional involvement of UCHL1 silencing in tumorigenesis and warrant further investigation.
The authors are grateful to Dr. M. Abe and Dr. T. J. Stedeford for the critical reading of the manuscript.