¹Department of Viral Oncology, Institute for Virus Research, Kyoto University, Shogoin, Sakyo-ku Kyoto 606-8507, Japan

²Laboratory of Pathology, Aichi Cancer Center Research Institute, 1-1 Kanokoden, Chikusa-ku, Nagoya 464-8681, Japan

Correspondence to: Y Ito, Institute of Molecular and Cell Biology, National University of Singapore, 30 Medical Drive, 117609, Singapore; E-mail: itoy@imcb.nus.edu.sg

^aCurrent addresses: Institute of Molecular and Cell Biology, National University of Singapore, 30 Medical Drive, 117609, Singapore

^bGladstone Institute of Virology and Immunology, Department of Medicine, University of California, San Francisco, California, CA 94141, USA

We reported recently that the silencing of RUNX3 is causally related to gastric cancer in humans. Here we report that in three of four cell lines derived from N-methyl-N-nitrosourea-induced mouse glandular stomach carcinomas, Runx3 is silenced due to hypermethylation of CpG islands in the promoter region, as we also observed for human gastric cancer cells. Although two of the sites we tested in the promoter of the fourth line were not methylated, in all four cases the silencing of Runx3 could be reversed by treatment of the cells with 5'-azacytidine and trichostatin A. Interestingly, the exogenous expression of RUNX3 in cell lines that do not express the endogenous gene caused an inhibition of growth in soft agar, suggesting that anchorage-independent growth could be used as an assay of RUNX3 activity in vitro. These observations suggest that the mouse system described here may be useful as a model for the study of human gastric carcinogenesis.

Oncogene (2002) 21, 8351-8355. doi:10.1038/sj.onc.1206037

Runx3; tumor suppressor; methylation; gastric cancer

Gastric cancer is the second most frequent cause of cancer-related deaths worldwide. Pathological and histological studies have highlighted the heterogeneous nature of tumors, and numerous cytogenetic alterations associated with tumors have been found on several chromosomes, including the short arm of chromosome 1. Molecular analysis revealed mutations in such genes as those encoding E-cadherin (Guilford et al., 1998), p53 (Kim et al., 1991; Tamura et al., 1991) and the transforming growth factor- (TGF-) receptor (Park et al., 1994). However, these mutations have been found only in a limited number of cases.

Members of the RUNX gene family are mammalian homologs of the Drosophila genes runt and lozenge, and proteins in the family play important roles in development and oncogenesis. RUNX1 anomalies are responsible for about 30% of the cases of human acute leukemia, and Runx2 collaborates with c-myc in inducing mouse T-lymphoma. Recently, we reported on the importance of RUNX3 as a tumor suppressor in gastric carcinogenesis (Li et al., 2002). Available evidence suggests a possible involvement of RUNX3 at very early stages of carcinogenesis as well as in tumor progression.

Gastric cancers were induced in C3H mice by providing N-methyl-N-nitrosourea (MNU) in drinking water (Tatematsu et al., 1993), and four cell lines derived from these mouse glandular stomach carcinomas were isolated (Ichinose et al., 1998). In this study, we found that these mouse gastric cancer cells exhibit properties remarkably similar to those of human gastric cancer cells.

To investigate whether this mouse system can serve as a model useful for the study of human gastric carcinogenesis, we first examined the expression of Runx3 in these cell lines. Runx3 mRNA was not detectable in three of these lines, MGT-40E16 (E16), MGT-40E25 (E25) and MGT-93 (93U). Expression of Runx3 in the fourth line, MGT-40E1 (E1), was barely discernible (Figure 1a). Interestingly, the related genes Runx1, Runx2, and Cbfb/Pebpb2 were all expressed in these cell lines, suggesting that Runx3 is the sole member of the Runx family to be inactivated in gastric cancer cells.

The promoter region of the mouse Runx3 (accession number AF169246), like the human RUNX3, has CpG islands (Figure 2b and Li et al., 2002). We therefore tested by methylation specific PCR (MSP) whether Runx3 silencing is also due to hypermethylation of the promoter region. A fragment from -2312 and -2028 (with respect to the translational start site) could be amplified from DNA prepared from the E1 and E16 cell lines but only with a methylation-specific primer set (Figure 2a). Both methylation- and non-methylation-specific primers weakly amplified DNA from E25 cells. On the other hand, only a non-methylation-specific primer set amplified this fragment from 93U cells. We also examined the region between -312 and +12 by MSP. This fragment could be amplified from E1, E16 and E25 cells using a methylation-specific primer set. Again, the fragment could be amplified from 93U cells only with the non-methylation specific primer set (Figure 2a). To examine the methylation status of the Runx3 CpG islands more directly, we sequenced these amplified DNA fragments from -2312 to -2028. For E1 cell line, methylation of the C in CpG dinucleotides was strongly correlated with the absence of Runx3 expression. For E16 cells, a partial methylation of the CpG islands can be observed (Figure 2b,c).

It has been shown that histone deacetylase is recruited to hypermethylated regions and that it inhibits gene expression (Nan et al., 1998; Jones et al., 1998). To test the possibility that the silencing of Runx3 might be reversed by treatment with methyltransferase inhibitors and histone deacetylase inhibitors, these cell lines were cultured in the presence of 5'-AC or TSA or a combination of the two for 3 days. In the case of E1 and E25, 5'-AC or TSA alone did not significantly reactivate the gene, but a combination of the two inhibitors induced the robust expression of Runx3 (Figure 1b, lanes 4 and 12). Reactivation of Runx3 was also observed in E16 cells treated with a combination of 5'-AC and TSA, but the level of expression was low (Figure 1b, lane 8). Expression of Runx3 was restored to high levels in 93U cells treated with TSA alone (Figure 1b, lane 14). A single treatment with 5'-AC did not change the level of expression, and the addition of 5'-AC to TSA only slightly increased it (Figure 1b, lane 16). In contrast, the expression of Runx1 and Runx2 were unchanged (data not shown). Overall, these results strongly support the notion that the silencing of Runx3 in the E1, E16, and E25 cell lines is due to DNA methylation and subsequent histone deacetylation of the promoter region. The silencing of Runx3 in 93U cells is likely caused by histone deacetylation but not DNA methylation.

We then examined whether exogenously expressed RUNX3 could inhibit the growth of mouse gastric cancer cell lines cultured in monolayers. Growth curves of the parental E1 cell line as well as of a clone containing only the vector and three RUNX3-expressing clones, cl.1, cl.2 and cl.3, are shown in Figure 3a. Figure 3b shows that the cl.1 clone exhibits the highest level of RUNX3 expression and that the cl.2 and cl.3 clones exhibit lower levels. Interestingly, the cl.1 line, which expresses the highest level of RUNX3, grows more slowly than the other cell lines, whereas cl.2 and cl.3 display intermediate levels of growth inhibition.

We next examined whether RUNX3 could also reduce the tumorigenicity of mouse gastric cancer cells. Unfortunately, none of the four cell lines that we studied formed tumors in nude mice as reported earlier (Ichinose et al., 1998). Since the growth of cancer cells in nude mice often correlates with anchorage-independent growth in vitro, we examined the growth of mouse gastric cancer cells in soft agar (Figure 3c,d). Interestingly, the exogenous expression of RUNX3 caused a marked reduction in both colony size and number, as compared with control cells. Again, the cells expressing the highest levels of RUNX3 were inhibited most strongly (Figure 3c,d). These results confirm that RUNX3 similarly inhibits the growth of mouse and human gastric cancer cells.

To further characterize the anti-oncogenic function of RUNX3, a human gastric cancer cell line, MKN74, which does not express endogenous RUNX3, was transfected with a series of constructs expressing truncated versions of RUNX3 and stable transfectants were isolated. The activity of a construct bearing a point mutation, RUNX3(R122C), which was identified in a gastric cancer patient, was also examined. All transfectants expressed comparable levels of RUNX3 and its derivatives (Figure 4a). Cells transfected with full length as well as with RUNX3 truncation derivatives that retain the transactivation domain gave rise to a significantly reduced number of G418-resistant foci. In contrast, cells transfected with RUNX3(R122C) as well as with RUNX3 truncation derivatives lacking the transactivation domain produced foci similar to those produced by control cells containing only the vector (Figure 4b). We also transfected RUNX3 and RUNX3(R122C) expression plasmids into mouse E1 cells. A dramatic reduction in the number of hygromycin-resistant colonies was observed in cells transfected with full-length RUNX3, but not with RUNX3(R122C), similar to the result of human cells shown in Figure 4b (data not shown). Because of this inhibitory effect, it was difficult to obtain clones expressing exogenous RUNX3, whereas it was easy to isolate full length RUNX3 harboring the R122C mutation or truncated derivatives lacking transactivation domain (Figure 4d). The RUNX3 (A372T) mutation has also been found in a gastric cancer patient but the biological significance of the A to T transversion in a non-conserved region of the protein is not clear. The possibility that this transversion represents a single nucleotide polymorphism cannot be excluded. Expression of the RUNX3(A372T) protein reduced the number of foci to the same extent as did the full-length RUNX3 (data not shown), suggesting that the A372T transversion is unlikely to be related to gastric carcinogenesis.

We also tested the tumorigenicity of one of the RUNX3 truncation derivatives, 1-234, which contains the first 234 amino acids and completely lacks the transactivation domain, by using a nude mouse assay. As shown in Figure 4c, both the cl.1 and cl.2 clones derived from the MKN74 cell line, which express full-length RUNX3, displayed significantly reduced levels of tumorigenicity in nude mice. In contrast, tumors grown in the presence of the 1-234 protein did not show a significant reduction in size. These results suggest that RUNX3 inhibits the growth of gastric cancer cells through its transactivation function.

It is noteworthy that exogenously expressed RUNX3 inhibits the growth of mouse gastric cancer cells cultured as a monolayer. The reduction in the number of foci produced by cells transfected with the RUNX3 expression plasmid is reminiscent of the effect caused by the expression of the retinoblastoma (RB) tumor suppressor or p53. RB and p53 are negative cell cycle regulators and well-established tumor suppressor genes. Recently, the RB gene was implicated as critical in cell differentiation (Amy et al., 1998; Thomas et al., 2001). Induction of the cell differentiation program must be coordinated with cell cycle regulation, and RB appears to have a central role in the coordination of these processes. RUNX genes are also developmental regulators. Our observation that RUNX3 inhibits cell growth could be interpreted to mean that this gene is also a normal regulator of the cell cycle. We reported previously that TGF- is a major growth regulator in gastric epithelial cells (Li et al., 2002). If the inhibition of cellular proliferation is mediated by the transactivational activity of RUNX3, potential targets of RUNX3 could include inhibitors of cell cycle regulators. It is important to examine whether Smad/RUNX3 complexes can activate any of the known inhibitors of cell cycle regulators. In fact, RUNX1, a relative of RUNX3, has been shown to activate p14ARF gene which is a negative regulator of cell cycle (Linggi et al., 2002).

The mechanism by which MNU induces gastric carcinomas remains unknown. It appears that ras and p53 genes do not play important roles in mouse gastric carcinogenesis induced by MNU (Furihata et al., 1997). Our observations may suggest that Runx3 is one of the major targets of the carcinogen directly or indirectly. It is also noteworthy that DNA methylation is observed in the promoter region of only Runx3 in mouse gastric cancer cells, although Runx1 and Runx2 have CpG islands near their promoters. It therefore appears that RUNX3 is uniquely associated with gastric cancer, underscoring the importance of studying the mechanism of selective DNA methylation of RUNX3 in gastric carcinogenesis. Evidence for gene silencing following hypermethylation, however, has been controversial in cancer studies, since promoter hypermethylation is not uniformly associated with the loss of expression of tumor suppressor genes (Fearon, 2000). Further studies are required to clarify this issue.

The results shown in this study are, by and large, remarkably similar to those reported earlier for human gastric cancer cells. Therefore, this mouse gastric cancer system, in combination with the Runx3 knockout mouse that we previously described, represents a mouse model for the study of human gastric carcinogenesis. The usefulness of the system is not restricted to cell lines. We may be able to show the effect of inhibitors on the tumor formation directly in in vivo studies.

Acknowledgements

We thank Motomi Osato for advice on the in vivo tumorigenicity assay.

Amy SY, Heather HS, Sergei GT. (1998). Front. Biosci., 3: 532-547.

Fearon ER. (2000). J. Natl. Cancer Inst., 92: 515-517. MEDLINE

Furihata C, Tatematsu M, Saito M, Ishida S, Nakanishi H, Inada K, Tei H, Hattori IV M, Ito T, Sakaki Y. (1997). Jpn. J. Cancer Res., 88: 363-368.

Guilford P, Hopkins J, Harraway J, McLeod M, McLeod N, Harawira P, Taite H, Scoular R, Miller A, Reeve AE. (1998). Nature, 392: 402-405. Article MEDLINE

Hanai J, Chen LF, Kanno T, Ohtani-Fujita N, Kim WY, Guo WH, Imamura T, Ishidou Y, Fukuchi M, Shi MJ, Stavnezer J, Kawabata M, Miyazono K, Ito Y. (1999). J. Biol. Chem., 274: 31577-31582. Article MEDLINE

Herman JG, Graff JR, Myohanen S, Neskin BD, Baylin SB. (1996). Proc. Natl. Acad. Sci. USA, 93: 9821-9826. Article MEDLINE

Ichinose M, Nakanishi H, Fujino S, Tatematsu M. (1998). Jpn. J. Cancer Res., 89: 516-524.

Jones PL, Veenstra GJ, Wade PA, Vermaak D, Kass SU, Landsberger N, Strouboulis J, Wolffe AP. (1998). Nature Genet., 19: 187-191. Article MEDLINE

Kim JH, Takahashi T, Chiba I, Park JG, Birrer MJ, Roh JK, De Lee H, Kim JP, Minna JD, Gazdar AF. (1991). J. Natl. Cancer Inst., 83: 938-943.

Li QL, Ito K, Sakakura C, Fukamachi H, Inoue K, Chi XZ, Lee KY, Nomura S, Lee CW, Han SB, Kim HM, Kim WJ, Yamamoto H, Yamashita N, Yano T, Ikeda T, Itohara S, Inazawa J, Abe T, Hagiwara A, Yamagishi H, Ooe A, Kaneda A, Sugimura T, Ushijima T, Bae SC, Ito Y. (2002). Cell, 109: 113-124. MEDLINE

Linggi B, Muller-Tidow C, van de Locht L, Hu M, Nip J, Serve H, Berdel WE, van der Reijden B, Quelle DE, Rowley JD, Cleveland J, Jansen JH, Pandolfi PP, Hiebert SW. (2002). Nat Med., 8: 743-750.

Nan X, Ng HH, Johnson CA, Laherty CD, Turner BM, Eisenman RN, Bird A. (1998). Nature, 393: 386-389. Article MEDLINE

Park K, Kim SJ, Bang YJ, Park JG, Kim NK, Roberts AB, Sporn MB. (1994). Proc. Natl. Acad. Sci. USA, 91: 8772-8776. MEDLINE

Tatematsu M, Yamamoto M, Iwata H, Fukami H, Yuasa H, Tezuka N, Masui T, Nakanishi H. (1993). Jpn. J. Cancer Res., 84: 1258-1264.

Tamura G, Kihana T, Nomura K, Terada M, Sugimura T, Hirohashi S. (1991). Cancer Res., 51: 3056-3058.

Thomas DM, Carty SA, Piscopo DM, Lee JS, Wang WF, Forrester WC, Hinds PW. (2001). Mol. Cell, 8: 303-316. MEDLINE

Figure 1 (a) The expression of Runx1, Runx2, Runx3 and Pebpb2 in four mouse gastric cancer cell lines (Ichinose et al., 1998) and mouse stomach tissue. cDNAs, synthesized using the Superscript Preamplification System (Gibco), of Runx1, Runx2, Runx3, Pebpb2 and of Gapdh as a control were amplified by RT-PCR. (b) Reactivation of Runx3 expression by 5'-AC and TSA treatment. Total RNA was prepared from E1, E16, E25 and 93U cells after cultured in the presence of 1.5 M 5'-AC (Sigma) for 36-48 h, followed by the addition of TSA to 330 nM (Sigma) or ethanol for another 24 h. The levels of the Runx3 and Gapdh transcripts were measured by RT-PCR. (EtOH, ethanol used as a solvent for TSA; T/A, combination of TSA and 5'-AC)

Figure 2 Methylation status of the Runx3 gene. (a) MSP was carried out according to Herman et al. (1996). (M, methylated; U, unmethylated). Mouse genomic DNA extracted from tails was used as a control (Mouse). (b) Methylation status of C residues between -2312 and -2028 relative to the translation initiation codon (ATG) of the Runx3 gene. Shown are the sequences of the MSP products generated by a methylation-specific primer set for the detection of E1 and E16 cells, and generated by a non-methylation-specific primer set for the detection of 93U cells and control mouse genomic DNA. At the top is the published wild type Runx3 sequence (Runx3(wt)). The MSP product from the E25 DNA sample was insufficient for sequencing. (c) Sequences from E1 and E16 cell lines, and from a control (Mouse), showing the underlined sequences indicated in (b)

Figure 3 Growth inhibition of E1 cells by RUNX3 in monolayer or in soft agar. (a) Growth of parental E1 cells (E1), cells containing vector only (control), and three exogenous RUNX3-expressing E1 clones, cl.1, cl.2 and cl.3. Cell counts were done in triplicate. (b) Detection of the exogenous expression of Flag-tagged full-length RUNX3 protein in E1 cells. Cells were harvested in SDS sample buffer, sonicated and boiled for 5 min. Western blot analysis was performed using the ECL Western blotting system as directed (Amersham). Mouse anti-flag M2 monoclonal antibody (Sigma) was used as the primary antibody for detecting flag-tagged RUNX3 proteins. (C) cl.1, cl.2, cl.3 and control colony formation in soft agar. Cells (3´10³ cells/dish) were plated in semi-solid medium containing 0.3% Bacto Agar (Difco) supplemented with 150 g/ml hygromycin over a 0.6% agar layer. Colonies were scored after 3 weeks incubation at 37°C in 5% CO₂ in air. (d) Colony counts on soft agar. Data were evaluated using Student's t-test. Asterisks indicate significant differences from control (P<0.01)

Figure 4 Inhibition of plating efficiency and tumorigenicity of gastric cancer cells by RUNX3. (a) Detection of Flag-tagged full-length RUNX3 and RUNX3(R122C) proteins, and of a series of C-terminally truncated RUNX3 derivatives, 1-187 aa, 1-234 aa, 1-283 aa, 1-325 aa, 1-373 aa, and 1-410 aa (Hanai et al., 1999), by Western analysis in transiently transfected MKN74 cells. (b) Plating efficiency of MKN74 cells containing vector only (control), expressing full-length RUNX3 or the RUNX3 truncation derivatives shown in (a). MKN74 cells were transfected and G418 resistant colonies were selected for 2 weeks. The results represented are one of three independent experiments. (c) Sizes of tumors formed by control, 1-234, and MKN74 parental cells (MKN74), and cells expressing RUNX3, clone 1 (cl.1) or clone 2 (cl.2) in nude mice 45 days after inoculation. Asterisks indicate significant differences from control (P<0.05). (d) Expression of RUNX3 derivatives in stably transfected cell lines. MKN74 cells stably transfected with the indicated RUNX3 derivatives were analysed by Western blotting. Note that RUNX3 (1-325) and RUNX3 (1-373) were hardly detectable