The Stress-Responsive Gene GADD45G Is a Functional Tumor Suppressor, with Its Response to Environmental Stresses Frequently Disrupted Epigenetically in Multiple Tumors

Epigenetic inactivation of tumor suppressor genes (TSG) is frequently associated with tumor pathogenesis (1). The major mechanism of this epigenetic inactivation is through hypermethylation of promoter CpG islands, which leads to the binding of transcription repressors, compressed chromatin, and transcription silencing (1). Increasing number of TSGs has been documented with epigenetic inactivation in tumors, such as p16^INK4a, hMLH1, VHL, BRCA1, and RASSF1A (2). Furthermore, promoter hypermethylation can be used as a biological marker for the identification of novel candidate TSGs and tumor diagnosis (3). Various methylation-based strategies, including methylation-sensitive representational difference analysis (MS-RDA; ref. 4), restriction landmark genome scanning (5), arbitrarily primed PCR (6), and CpG island microarray (7), have been developed and proven to be useful for identifying hypermethylated sequences. MS-RDA has been successfully used to identify silenced genes in tumors, imprinted regions on mouse chromosome 2, and differentially methylated regions after viral infections (reviewed in ref. 4).

GADD45 is a family of proteins involved in DNA damage response and cell growth arrest. GADD45A was initially identified as a gene rapidly induced by DNA-damaging agents, such as methylmethane sulfonate, UV radiation, hydroxyurea, and ionizing radiation (8). It is a classic TP53-regulated gene, and GADD45A-null mice exhibit a phenotype similar to TP53-null mice (9). GADD45B (MyD118) was firstly identified as a myeloid differentiation responsive gene, activated in M1 myeloid leukemia cells by interleukin-6 after induction of terminal differentiation (10). GADD45G (GRP17/CR6) was identified as an interleukin-2-induced immediate-early gene (11–13). These family proteins share 55% to 58% amino acid homology and mediate the activation of the p38/c-Jun NH₂-terminal kinase pathway via MTK1/MEKK4 in response to various environmental stresses (11). They can also inhibit cell proliferation at different stages, including G₁-S and G₂-M checkpoints, and induce cell apoptosis (11, 13–15). GADD45B is down-regulated in hepatocellular carcinoma through methylation (16), and GADD45G is also down-regulated in anaplastic thyroid cancer and pituitary adenoma (17, 18).

To identify aberrantly methylated genes, we used a novel approach by combining MS-RDA with demethylation treatment in cell lines of nasopharyngeal carcinoma, a prevalent tumor in our locality of Southern China and Southeast Asia. We identified GADD45G as a target gene for epigenetic inactivation in nasopharyngeal carcinoma as well as multiple other types of carcinomas and lymphomas. We further showed that exogenous expression of GADD45G strongly suppressed tumor cell growth and colony formation, indicating that GADD45G can act as a functional TSG.

Cell lines and primary tumors. In total, 75 cell lines of carcinomas and lymphomas were used, including nasopharyngeal (C666-1, CNE-1, CNE-2, HK1, NPC-BM1, HNE1, HNE2, HNE3, HONE-1, 5-8F, and HKM1), breast (MCF-7, T47D, ZR-75-1, and MDA-MB231), esophageal (EC1, EC18, EC109, HKESC-1, HKESC-2, HKESC-3, SLMT-1, KYSE30, KYSE70, KYSE140, KYSE150, KYSE180, KYSE270, KYSE410, KYSE450, KYSE510, and KYSE520), colorectal (HCT116, HT-29, HCT15, SW48, and LoVo), cervical (HeLa, CaSki, C33A, and SiHa), lung (A549, H1299, H2126, H1395, and H292), hepatocellular (HepG2, huH1, huH4, huH6, and huH7), gastric (KatoIII), and laryngeal carcinoma (HEp-2), non-Hodgkin's lymphoma (BJAB, CA46, Rael, Namalwa, Raji, AG876, OCI-Ly1, Ly3, Ly7, Ly8, Ly13.2, Ly17, and Ly18), Hodgkin's lymphoma (L428, L540, L591, HD-LM2, HD-MY-Z, and KM-H2), and leukemia (HL-60, THP-1, and K562). Four immortalized normal epithelial cell lines (NP69, NE1, NE3, and RHEK-1; refs. 19, 20) with many features of normal epithelial cells were also used. HCT116 DNMT1^−/− (1KO), HCT116 DNMT3B^−/− (3BKO), and HCT116 DNMT1^−/−DNMT3B^−/− (DKO) cells (gifts of Bert Vogelstein, Johns Hopkins, Baltimore, MD) were grown with either 0.4 mg/mL geneticin or 0.05 mg/mL hygromycin (21). DNA and RNA of primary carcinomas and lymphomas have been described (19, 22–25). Cell lines were treated with 5-aza-2′-deoxycytidine (Sigma, St. Louis, MO) as previously (19). For the treatment combining 5-aza-2′-deoxycytidine and trichostatin A (Cayman Chemical Co., Ann Arbor, MI), cells were treated with 5-aza-2′-deoxycytidine for 3 days and subsequently with trichostatin A (100 ng/mL) for 24 hours.

Stress treatments. Heat shock was done as previously (19), except for an incubation at 42°C for 1 hour with recovery at 37°C for 2 hours. For UV treatment, medium was removed and the flask was turned upside down to face the light source in a UV cross-linker 500 (Amersham Biosciences, Piscataway, NJ). Cells were irradiated for a dose of 70 J/m². After irradiation, fresh medium was added, and the cells were recovered at 37°C for 1 hour and then harvested.

Methylation-sensitive representational difference analysis. MS-RDA was done as previously (26). We used the genomic DNA of a nasopharyngeal carcinoma cell line CNE-1 (the Driver) and CNE-1 treated with 50 μmol/L 5-aza-2′-deoxycytidine (the Tester) and did two cycles of competitive hybridization. PCR products from the second hybridization were cloned into the pCR2.1-Topo vector (Invitrogen, Carlsbad, CA), sequenced, and analyzed using the BLASTN algorithm (National Center for Biotechnology Information).

5′-Rapid Amplification of cDNA Ends. We determined the GADD45G transcription start site using 5′-Rapid Amplification of cDNA Ends version 2.0 (Invitrogen). Briefly, the first-strand cDNA was synthesized from placenta RNA using primer GADD45GR2 (Table 1). Homopolymeric tails were then added to the 3′ ends with terminal deoxynucleotidyl transferase. PCR was done using Abridged Anchor Primer and a second gene-specific primer GADD45GR1 (Table 1). The Rapid Amplification of cDNA Ends product was enriched by reamplifying with the Abridged Universal Amplification Primer and GADD45GR1, cloned, and sequenced.

Semiquantitative reverse transcription-PCR analysis. Reverse transcription-PCR (RT-PCR) was done as previously (22, 27). Primers are shown in Table 1.

Bisulfite treatment and promoter methylation analysis. Bisulfate modification of DNA, methylation-specific PCR (MSP), and bisulfite genomic sequencing (BGS) were carried out as previously (27, 28). GADD45B promoter and two regions of the GADD45G promoter were analyzed by MSP (Table 1). MSP primers were tested previously for not amplifying any unbisulfited DNA.

Analyses of tumor suppressor functions. The full-length cDNA of GADD45G was PCR cloned from human trachea RNA (BD Clontech, Palo Alto, CA), sequence verified, and then subcloned into both pcDNA3.1(+) and pcDNA3.1(−) vectors (Invitrogen) to generate sense plasmid pcDNA3.1(+)GADD45G and antisense plasmid pcDNA3.1(−)GADD45G. For colony formation assay using monolayer culture, cells (2 × 10⁵ per well) were plated in a 12-well plate and transfected with 2 μg sense, antisense, or vector plasmid using LipofectAMINE 2000 (Invitrogen). Cells were stripped off, plated in a six-well plate 48 hours post-transfection, and selected for 2 to 3 weeks with 0.4 to 0.5 mg/mL G418. Surviving colonies (≥50 cells per colony) were counted after staining with Giemsa. For colony formation assay using soft agar culture, cells were transfected as above and suspended in RPMI 1640 containing 0.35% agar, 10% fetal bovine serum, and G418 and then layered on RPMI 1640 containing 0.5% agar, 10% fetal bovine serum, and G418 in a six-well plate 48 hours post-transfection. Colonies were counted and photographed 16 days post-transfection. Total RNA from transfected cells was extracted, treated with DNase I, and analyzed by RT-PCR to confirm the ectopic expression of GADD45G. All experiments were done in triplicate wells and repeated thrice.

Cell proliferation assay. CA46 cells (1 × 10⁶ per well in 12-well plate) were transfected with empty vector or pcDNA3.1(+)GADD45G and plated in a 12-well plate (7.5 × 10⁴ per well) 12 hours post-transfection. Cells were counted at indicated time points using a hemocytometer based on trypan blue exclusion. The experiment was done in triplicate wells and repeated thrice.

Screening for GADD45G mutations. Single-stranded cDNA was synthesized with SuperScript II reverse transcriptase (Invitrogen) using total RNA from cell lines or cell lines treated with 5-aza-2′-deoxycytidine. RT-PCR was used to amplify the full-length coding sequence with Accume Pfx DNA polymerase (Invitrogen) and primers GADD45GF and GADD45GR2. PCR products were electrophoresed, purified, and sequenced in both directions.

Statistical analysis. Statistical significance of the assays was analyzed by t test. P < 0.05 was taken as statistically significant.

Identification of the CpG island of GADD45G as a hypermethylated target. By combining MS-RDA with pharmacologic demethylation, DNA fragments hypermethylated in CNE-1 but demethylated after 5-aza-2′-deoxycytidine treatment were identified, cloned, and analyzed by bioinformatics. All but 2 of the 31 unique fragments identified have typical CpG islands.⁸

Frequent epigenetic silencing of GADD45G but not GADD45B in various cell lines. We then examined the expression of GADD45G, together with other family members GADD45A and GADD45B, in a series of cell lines to determine whether there are widespread epigenetic inactivation of these genes in tumors. We analyzed 11 nasopharyngeal carcinoma and 56 other cell lines using semiquantitative RT-PCR. We found that GADD45G expression was remarkably reduced or silenced in 8 of 11 nasopharyngeal carcinoma, 11 of 13 non-Hodgkin's lymphoma, 3 of 6 Hodgkin's lymphoma, and 13 of 37 other cell lines. In contrast, the expression levels of GADD45A and GADD45B remained high in virtually all cell lines, with GADD45B the highest (Fig. 1C). Furthermore, along with GADD45A and GADD45B, GADD45G was readily detected in all 30 normal adult and fetal tissues, including normal peripheral blood mononuclear cells (Fig. 1B).

Before we could assess the epigenetic alterations of GADD45G, we determined the accurate location of the promoter and the transcription start site using 5′-Rapid Amplification of cDNA Ends. We obtained a strong PCR band of ∼260 nucleotides (data not shown). Sequence analysis of the product (AY845250) showed that the identified transcription start site matched exactly the published cDNA 5′ sequence (NM_006705) in National Center for Biotechnology Information database. The GADD45G promoter is located within a typical CpG island (19), containing the core promoter, exons 1 to 3, and part of exon 4 (Fig. 1A). CpG islands are frequently silenced by methylation in tumors as an alternative epigenetic mechanism to inactivate TSG functions (1). We suspected that the silencing of GADD45G might also be mediated through epigenetic regulation. Therefore, the methylation status of two regions in the GADD45G promoter was analyzed by MSP in a total of 75 cell lines (Fig. 1C; Table 2). The MSP results of region 2 were identical to that of region 1 (data not shown). In total, GADD45G methylation was detected in 85% (11 of 13) non-Hodgkin's lymphoma, 50% (3 of 6) Hodgkin's lymphoma, 73% (8 of 11) nasopharyngeal carcinoma, 29% (5 of 17) esophageal carcinoma, 50% (2 of 4) cervical carcinoma, and 40% (2 of 5) lung carcinoma and other tumor cell lines. All the cell lines with hypermethylation had either reduced or silenced expression depending on the extent of methylation level. No methylation was found in the four immortalized normal epithelial cell lines, which had virtually normal phenotypes and expressed this gene (Fig. 2A). In contrast, no methylation of the GADD45B promoter was detected, which correlated with its broad expression in cell lines (Fig. 1C).

To further examine GADD45G methylation in more detail, we analyzed 12 cell lines, 2 normal tissues, and 5 tumors using the high-resolution BGS analysis (Fig. 2B). We analyzed 33 CpG sites spanning the core promoter and part of exon 1 in a 355-bp region. The BGS results strongly correlated with the MSP analysis. Methylated CpG sites were not found or only scattered in cell lines expressing GADD45G. In contrast, densely methylated CpG sites were detected in all silenced cell lines (Fig. 2B). Our data indicate that epigenetic silencing of GADD45G is involved in the pathogenesis of a few tumors.

Activation of GADD45G expression by pharmacologic and genetic demethylation. To test whether methylation is directly responsible for silencing GADD45G, we treated five cell lines (Rael, Namalwa, Raji, L428, and EC109) with hypermethylated and silenced promoter with the demethylating agent 5-aza-2′-deoxycytidine. 5-Aza-2′-deoxycytidine restored GADD45G expression in all of them although with different efficiency (Fig. 3A). No significant increase of either GADD45A or GADD45B expression was detected (data not shown). Concomitantly, after 5-aza-2′-deoxycytidine treatment, unmethylated GADD45G alleles were increased as determined by both MSP and BGS (Figs. 2B and 3A). The Rael cell line showed marginal restoration of GADD45G after 5-aza-2′-deoxycytidine treatment. However, further treatment with 5-aza-2′-deoxycytidine combined with histone deacetylase inhibitor, trichostatin A, resulted in synergistic activation of GADD45G to greater levels (data not shown), suggesting that histone deacetylation is also involved in repressing GADD45G in this cell line. GADD45G could also be activated to similar extent as the 5-aza-2′-deoxycytidine–treated in HCT116, which has a minority of methylated alleles, by genetic biallelic disruption of both DNMT1 and DNMT3B (DKO) but not DNMT1 or DNMT3B alone (ref. 21; data not shown). In contrast, neither GADD45A nor GADD45B expression was affected in the DKO cell line (data not shown). These results indicate that the maintenance of GADD45G promoter methylation is mediated by DNMT1 and DNMT3B together, like other known typical TSGs (30).⁹

GADD45G methylation is tumor specific. We further investigated the presence of GADD45G methylation in a large collection of primary tumors, including various lymphomas and carcinomas and normal tissues (Fig. 2A and B; Table 2). GADD45G methylation was detected in multiple tumor types with different frequencies, more frequently in lymphomas than carcinomas and in carcinoma cell lines than primary carcinomas (73% versus 16% in nasopharyngeal carcinoma and 29% versus 11% in esophageal carcinoma; Table 2), indicating that some cell lines may have acquired GADD45G methylation during the establishment or maintenance process. Similar phenomenon has been reported for some TSGs in other tumors (31, 32). None of the 39 normal tissues (12 peripheral blood mononuclear cells, 3 lymph nodes, 10 nasopharynx, 7 normal breast tissues, and 7 normal esophageal epithelia) had methylation as assessed by either BGS or MSP. Therefore, hypermethylation of the GADD45G promoter is tumor specific.

Genetic inactivation of GADD45G is very rare. Because TSGs can also be inactivated genetically by various mutations, including deletions, and GADD45G is located in a frequently deleted locus, we tested this possibility for GADD45G and did not detect its deletion in any cell line. We further used direct sequencing to screen for any possible point mutation in 25 cell lines, including expressing cell lines and silenced cell lines treated with 5-aza-2′-deoxycytidine. Only a single sequence change G344A (G112E) in exon 3 was identified in one cell line AG876 (Figs. 1D and 3B), in which the promoter was also methylated. This sequence substitution might be a rare mutation or a polymorphism. Therefore, the disruption of GADD45G functions in tumor cell lines is solely through epigenetic rather than genetic mechanism.

Promoter hypermethylation disrupts the stress response of GADD45G. Because GADD45 family proteins are involved in cellular responses to environmental stresses, we inspected the GADD45G promoter for potential stress-responsive elements. Multiple heat shock factor binding sites (6 in the core promoter, with a total of 26 sites in a region from −1,500 to +150) are present in the promoter (Fig. 1A), indicating that it is likely stress responsive. It has been reported that this promoter is inducible by γ-ray, H₂O₂, and UV irradiation (11). We further tested whether it is responsive to heat shock and whether its stress response would be affected by promoter methylation. Stress treatments (heat shock and UV irradiation) of normal and tumor cell lines with an unmethylated or only partially methylated promoter resulted in increased GADD45G expression. However, this response was decreased or abolished in cell lines with methylated promoter and was inversely correlated with the promoter methylation levels (Fig. 3C). Under our conditions of stress treatment (heat shock at 42°C for 1 hour with 2-hour recovery and 70 J/m² UVC treatment with 1-hour recovery), we only observed marginal up-regulation of GADD45A but not GADD45B RNA. It has been shown that different GADD45 genes respond to stress stimuli differently (i.e., GADD45A is inducible by p53, whereas GADD45B and GADD45G are not, and interleukin-12 can induce GADD45B but not GADD45G; ref. 15). Interestingly, in treated normal (NE3) and some tumor cell lines (EC109 and Rael), two new isoforms of GADD45B were induced after heat shock (Fig. 3C). These isoforms have extended exon 1 of different lengths, resulting in the translation of a shorter isoform (15 amino acids less) from an alternate ATG code more downstream (accession nos. AY615270 and AY615271). The functions of these new GADD45B isoforms need further study.

GADD45G is a functional tumor suppressor.GADD45G is associated with proliferating cell nuclear antigen and the cyclin-dependent kinase inhibitor p21^WAF1/CIP1 and is involved in the negative control of cell growth (13, 15). GADD45G is also associated with MTK1/MEKK4, which in turn activates the p38/c-Jun NH₂-terminal kinase pathway leading to apoptosis, in response to environmental stresses (11). Transient transfection of GADD45 members could induce apoptosis in tumor cells (11, 13, 15). The frequent epigenetic silencing of GADD45G in tumors and cell lines but not normal tissues suggests that it might be a functional tumor suppressor. To test this hypothesis, we cloned the full-length GADD45G cDNA into expression vectors and transfected it into several carcinoma and lymphoma cell lines, which had either complete methylation and silencing (HepG2, EC109, HEp-2, and CA46) or only weak expression with both methylated and unmethylated alleles (CNE-1 and HK1). The colony formation efficiencies of each transfected cell line were evaluated by monolayer and soft-agar culture with G418 selection. Ectopic expression of GADD45G dramatically reduced the colony formation efficiencies of these cell lines in both monolayer cultures (down to 9-20% of controls in methylated cell lines and 19-41% in hemimethylated cell lines; Fig. 4A) and soft-agar assays (down to 8% of controls in CNE-1 and HepG2; Fig. 4B), suggesting that GADD45G can function as a bona fide TSG. Interestingly, HK1 has the relatively highest endogenous expression and lowest methylation level. The inhibition of colony formation by ectopic GADD45G expression in HK1 was also the least effective. Furthermore, cell proliferation assay was done for CA46 (Fig. 4C). Cells transfected with pcDNA3.1(+)GADD45G grew significantly slower than the mock-transfected cells (P < 0.001), indicating that GADD45G dramatically inhibits not only tumor cell colony formation but also their proliferation. Previously, it has been shown that ectopic GADD45G expression in lung carcinoma, pituitary tumor, and anaplastic thyroid carcinoma–derived cell lines resulted in a substantial inhibition of tumor cell growth (17, 18). As shown in this study, GADD45G also strongly suppressed tumor cell growth and colony formation of other tumor cell lines, including nasopharyngeal, hepatocellular, esophageal, and laryngeal carcinoma and lymphoma cell lines, further indicating that GADD45G can function as a TSG in multiple tumors.

In summary, we identified the GADD45G CpG island as a tumor-specific, hypermethylated target sequence. We also found that GADD45G expression is frequently reduced or silenced in multiple tumors, including lymphomas and carcinomas, like other class II TSGs (33). This silencing is due to the hypermethylation of its promoter, which further impairs its response to environmental stresses, but genetic inactivation of GADD45G is rare. We further showed that GADD45G can act as a functional TSG in multiple tumor cells. During the preparation of this article, Bahar et al. also reported the epigenetic down-regulation of GADD45G in pituitary adenoma (34). This frequent inactivation of GADD45G in various tumors is consistent with its proposed role as a negative regulator of cell proliferation (13, 15) and also shows the importance of GADD45G in preventing the development of multiple tumors. As promoter hypermethylation is pharmacologically reversible using demethylating agents, and GADD45G is rarely mutated in tumors, it is therefore a likely target for new epigenetic anticancer therapeutics.

GenBank accession numbers. The sequences of the 5′-Rapid Amplification of cDNA Ends product of GADD45G (accession no. AY845250) and heat shock–induced GADD45B isoforms (accession nos. AY615270 and AY615271) have been deposited to Genbank.

During the final preparation of this article, Zerbini et al. reported that nuclear factor-κB–mediated repression of GADD45A and GADD45G is essential for cancer cell survival (Zerbini LF, et al. Proc Natl Acad Sci U S A 2004;101:13618–23), thus pointing out another mechanism to inactive GADD45G functions in tumors.

We thank Drs. Bert Vogelstein, Soh-ha Chan, Riccardo Dalla-Favera, Meehard Herlyn, Dolly Huang, Katai Yao, Ya Cao, Thomas Putti, Guiyuan Li, Yixin Zeng, Sen-Tien Tsai, Sai Wah Tsao, Johng S. Rhim, Malini Olivo, Goh Boon Cher, and Lee Soo Chin for some cell lines and samples.