Our previous studies suggest that lack of RUNX3 function is causally related to the genesis and progression of human gastric cancer, but potential roles of other members of the RUNX family genes have not yet been reported. We examined the expression of 3 Runt-related (RUNX) genes, RUNX1, RUNX2 and CBFB, in gastric cancer cell lines and primary gastric cancer specimens and compared them to those of RUNX3 reported earlier in conjunction with clinicopathologic factors. Expression of RUNX family genes in 9 gastric cancer cell lines, 56 primary gastric cancer specimens and surrounding normal gastric mucosa were estimated by Northern blot analysis, quantitative RT-PCR and in situ hybridization. Northern blot analysis in gastric cancer cell lines showed downregulation of RUNX1 and RUNX3 in 67% and 78% of the cell lines tested, respectively. The ratio of the average RUNX mRNA/β-actin mRNA ratio (×103) for RUNX1 was 48.0 ± 21.1 vs. 21.4 ± 8.1; RUNX2, 1.1 ± 0.3 vs. 1.0 ± 0.2; RUNX3, 9.2 ± 6.3 vs. 3.1 ± 1.3 and CBFB, 42.0 ± 19.4 vs. 21.0 ± 8.4 (normal vs. tumor, respectively, average ±SD). The basal RUNX2 expression was very weak, and there was no significant change in gastric cancers. Both RUNX1 and RUNX3 showed remarkable downregulation in 62% and 69%, respectively, of surgically resected specimens compared to surrounding mucosa analyzed by quantitative RT-PCR (p < 0.01). Furthermore, CBFB, the gene encoding the cofactor of RUNX1, -2, -3, was also downregulated in significant fraction (32%, p < 0.05). The percentage of downregulation of RUNX1, RUNX3 and CBFB increased as the cancer stage progressed. Tricostatin A and 5′-azacitidin reactivate RUNX3 expression, but they could not reactivate expression of RUNX1 and CBFB in gastric cancer cells, suggesting that the downregulation was due to mechanisms other than methylation of the promoter region. These findings suggest that RUNX1 and CBFB in addition to RUNX3 play some roles in gastric cancers and that roles of RUNX gene family in gastric cancer are more widespread and complex than previously realized.
The 3 mammalian Runt-related (RUNX) genes, RUNX1, -2, -3, encode a set of closely related DNA binding proteins homologous to runt, one of the Drosophila pair-rule gene products. Duplication of this family appears to have been an early event in evolution, preceding the emergence of vertebrates.1, 2, 3, 4, 5 The 3 RUNX proteins bind DNA as heterodimeric complexes with a common partner protein, core binding factor β (CBFβ) or polyomavirus enhancer binding protein 2β (PEBP2β), which is an essential partner of RUNX proteins as a transcription factor, and they function together to regulate downstream genes, which are important in development and differentiation as well as in cancer.
Recent studies indicate that these RUNX family genes are involved in many kinds of human cancers. RUNX1 is essential for definitive hematopoiesis,6, 7 but at the same time, RUNX1 is expressed in a variety of myeloid and lymphoid lineages and the CBF/PEBP2-binding sites are present in many hematopoietic cell-specific target genes, suggesting important roles at subsequent stages of development. RUNX1 is a frequent target of chromosome translocations as well as mutations in myeloid and lymphoid leukemia. Chromosomal translocations result in truncation and fusion of RUNX1 to heterologous proteins,8, 9 which, in most cases, inhibit normal RUNX1 function, perturb lineage differentiation and predispose to leukemia.10, 11RUNX2 is essential for bone formation.12, 13, 14, 15 Oncogenic activity of the gene has been demonstrated in the mouse system in which Runx2 is shown to function as a dominant oncogene in T-cell lymphoma.16, 17 CBFβ/PEBP2β is involved in the inv(16) acute myelogenous leukemia.18, 19, 20, 21
Recently, we reported the causal relationship between the loss of RUNX3 expression and gastric cancer.22 However, other members of the RUNX family proteins are also expressed in the gastric epithelial cells, and they share the binding sites. Therefore, it is important to study possible roles of other members of the RUNX family in growth and differentiation of stomach epithelial cells and formation of gastric cancer. In our study, we examined the expression of RUNX1, -2, -3 and CBFβ/PEBPB2 in normal gastric mucosa and gastric cancers.
Material and methods
Cell culture, gastric cancer specimens and RNA preparation
Gastric cancer cell lines SNU-1, SNU-5, SNU-719 cells were established previously by Park et al.23 KATO-III and GT3TKB were purchased from Riken Cell Bank (Tsukuba, Japan). NUGC-3 cells were purchased from Health Science Research Resources Bank (Osaka, Japan). MKN1, MKN28, MKN45, MKN74 and GT3TKB were maintained at 37°C in a humidified atmosphere of 5% CO2 in high-glucose DMEM (Sigma, St. Louis, MO), whereas NUGC-3 were maintained in RPMI 1640 (Sigma). Both media were supplemented with 10% fetal bovine serum, penicillin and streptomycin. When they reached 80–90% confluence, cells were washed with ice-cold PBS and homogenized immediately in Isogen reagent (Nippon Gene, Osaka, Japan), and total RNA was extracted. mRNA extracted from each cell line was extracted by FAST Track Kit Ver. 2 (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions.
The study population consisted of 56 patients with gastric cancer undergoing surgery at Kyoto Prefectural University of Medicine from 1999 to 2003. Clinical data of these patients are shown in Table I. Clinical samples were washed with ice-cold PBS and homogenized immediately in Isogen reagent (Nippon Gene), and total RNA was extracted and stored at −80°C until use. Ethics approval exists and written informed consent was obtained from each patient prior to tissue acquisition.
Table I. Clinicopathologic Characteristics of 56 Patients with Gastric Cancer
Clinical stage according to Japanese Gastric Cancer Classification. t classification: t1, mucosa to submucosa; t2, muscularis propria to subserosa; t3, serosa-exposed; t4, serosa-infiltrating.
66.3 ± 10.5
Depth of invasion
Lymph node metastasis
Northern blot analysis
Northern blot was performed as we previously described.24, 25, 26 In brief, poly (A)+ RNA from each cell line was extracted by FAST Track Kit Ver. 2 (Invitrogen) according to the manufacturer's instructions, then fractionated on 1% agarose/2.2 M formaldehyde gels. Probes were labeled with 32P by random priming. Each blot was then hybridized with the probe for the selected gene and β-actin. Signals were analyzed with a BAS 2000 image analyzer followed by calculation of degree of downregulation compared to control.
In situ hybridization
To detect RUNX family gene expression in human gastric cancer specimens, in situ hybridization on paraffin-embedded sections was performed as described previously22 using sense and antisense DIG-labeled probes consisting of RUNX3 nucleotide 550 to 848, RUNX1 nucleotide 998 to 1300, RUNX2 nucleotide 761 to 1081, CBFβ nucleotide 322 to 599 (L20298), respectively.
RNA probe was synthesized with T7 RNA polymerase, using a digoxigenin (DIG) RNA Labeling Kit (Boehringer Mannheim, Indianapolis, IN). After proteinase K digestion (18 μg/ml), the sections were post-fixed with 4% (w/v) paraformaldehyde in phosphate-buffered saline for 10 min and treated with 0.1 M triethanolamine-HCl (pH 8.0) for 1 min. After acetylation for 10 min, the sections were dehydrated, air-dried and then incubated overnight at 50°C in hybridization buffer composed of 50% formamide, 10 mM Tris-HCl (pH 7.5), 1 mg/ml yeast tRNA (Sigma), 1× Denhalt's solution (Sigma), 10% PEG6000, 600 mM NaCl, 0.25% SDS, 1 mM EDTA and 0.2 μg/ml probe. After hybridization, the sections were washed at 45°C for 1 hr in 50% formamide and 2× SSC and digested with 20 μg/ml RNase (Sigma) in 10 mM Tris-HCl (pH 8.0) and 500 mM NaCl at 37°C for 30 min. Hybridized DIG-labeled probes were visualized with a Nucleic Acid Detection Kit (Boehringer Mannheim).
In our present study, we counted 1,000 cells of the tumor or adjacent noncancerous epithelial cells to calculate the percentage of stained cells; cell staining in less than 30% of cells was the criterion for downregulation of RUNX family gene expression.
Real-time quantitative RT-PCR
cDNA was produced from total RNA by using a Superscript preamplification system (BRL, Bethesda, MD) and following the procedures suggested by the manufacturer. RNA was heated to 70°C for 10 min in 14 μl of duethylpyrocarbonate-treated water containing 0.5 μg oligo (dT). Synthesis buffer (10×), 2 μl 10 mM dNTP mix, 2 μl 0.1 M DTT, and reverse transcriptase (Superscript RT; 200U/μL) were added to the sample. The resulting reaction mixture was incubated at 42°C for 50 min, and reaction was terminated by incubating the mixture at 90°C for 5 min.
Quantitative PCR was performed using real-time Taqman TM technology and analyzed on a Model 5700 Sequence Detector (Applied Biosystems, Foster City, CA) as described previously.26
RUNX3 RT-PCR primers are 5′-AAGCACAGCCATCAGGATTCA-3′ and 5′-TGGACATGCTTGCGGATATAAG-3′. Hybridization probes, which bind to PCR products, were labeled with a reporter dye, FAM on the 5′ nucleotide, and a quenching dye, TAMRA on the 3′ nucleotide. Sequences of hybridization probes are RUNX3: 5′-(FAM) CATCTGGAACTTCTCCTGGTCTCTCAGC (TAMRA)-3′. Other sets of primers and hybridization probes for RUNX1, -2, CBFβ, β-actin RNA were purchased from Applied Biosystems.
Fifty-microliter reactions contained: 1.25 units Amp-Taq DNA polymerase, 1× PCR reaction buffer, 180 ng of each primer, 200 mM dNTP, 400 mM dNTP, 100 nM Taqman probe and 0.5 U Amplirase (Applied Biosystems). The Ct value corresponding to the cycle number at which the fluorescence emission monitored in real time reaches a threshold of 10 standard deviations above the mean baseline emission from cycle 1 to 40 was measured serial dilutions of control cDNA, analyzed for each target. These target genes served as standard curves from which to determine the rate of changes of Ct value. Cycling parameters were 2 min at 50°C, 10 min at 95°C followed by 40 cycles of 15 sec at 95°C and 1 min at 60°C.
To minimize the errors arising from the variation in the amount of starting RNA among samples, amplification of β-actin mRNA was performed as an internal reference against which other RNA values can be normalized. Normalized results were expressed as the ratio of copies of each gene to copies of the β-actin gene.
Reactivation of RUNX family genes by treatment with TSA and 5′-AC
The expression change of RUNX1, RUNX2, RUNX3 and CBFB after treatment with TSA and 5′-AC are examined in 10 gastric cancer cell lines as described previously.28 cDNAs, synthesized using the Superscript Preamplification System (Gibco BRL, Gaithersburg, MD), of RUNX1, RUNX2, RUNX3, CBFB and of β-actin as a control were amplified by RT−PCR.
Reactivation of RUNX3 expression by 5′-AC and TSA treatment.
Total RNA was prepared from gastric cancer cells after cultured in the presence of 1.5 M 5′-AC (Sigma) for 36–48 hr, followed by the addition of TSA to 330 nM (Sigma) or ethanol for another 24 hr. The levels of the RUNX family genes and β-actin transcripts were measured by conventional RT−PCR or quantitative RT-PCR.
Statistical methods for analysis
Statistical analysis was performed using the NAP system programmed by Aoki (Version 4.0). The first objective of the statistical analysis was to examine the difference of RUNX expression between gastric cancer specimens and surrounding mucosa with unpaired t-test. Results with p-values of less than 0.05 were considered statistically significant. The clinicopathologic factors in various groups of patients with RUNX family positive or negative were compared by means of either χ2 test or Mann-Whitney U test. Different groups (e.g., surrounding mucosa and tumors of stages I, II, III and IV) were compared using the nonparametric Wilcoxon rank sum test. Results with p-values of less than 0.05 were considered statistically significant.
RUNX family gene expression in gastric cancer cell lines
Northern blot and in situ hybridization analyses showed that RUNX1 is expressed in gastric epithelial cells. Therefore, we examined gastric cancer-derived cell lines for the expression of RUNX1, RUNX2, RUNX3 and CBFB and observed that downregulation of RUNX1 and RUNX3 in 6 (67%) and 7 (78%) of the 9 cell lines tested, respectively. Expression of CBFB was observed in most gastric cancer cells except for SNU-719 and GT3TKB cells as shown in Figure 1. Of them, 4 lines exhibited downregulation of both genes, 2 lines exhibited only RUNX3 and 1 line exhibited only RUNX1 downregulated. Unlike other RUNX family genes, RUNX2 was expressed only at a very low level in control and gastric cancer cells, suggesting that RUNX2 does not play significant roles in gastric epithelial cells (Fig. 1).
RUNX family gene expression in primary gastric cancer specimens analyzed by quantitative RT-PCR
To standardize the amount of RNA in each sample, β-actin mRNA was used as an internal control. The relative value for the level of expression of RUNX1, -2, -3 and CBFB was determined as the RUNX1, -2, -3 and CBFB mRNA/β-actin mRNA ratio. The plot for RUNX1, -2, -3 and CBFB mRNA/β-actin mRNA (×107) is shown in Figure 2. The average RUNX mRNA/β-actin mRNA ratio (×103) in normal gastric mucosa was: RUNX1, 48.0 ± 21.1; RUNX2, 1.1 ± 0.3; RUNX3, 9.2 ± 6.3; CBFB, 42 ± 19.4 (average ± SD). In gastric cancer specimens, the average RUNX mRNA/β-actin mRNA ratio (×103) was: RUNX1, 21.4 ± 8.1; RUNX2, 1.0 ± 0.2; RUNX3, 3.1 ± 1.3; CBFB, 21.0 ± 8.4 (average ± SD).
The results showed a significant difference between tumor and normal gastric mucosa (RUNX1, RUNX3, 62% and 69%, respectively, p < 0.01, unpaired t-test). CBFB expression in primary gastric cancers decreased relative to normal gastric mucosa in 18 of 56 cases (33%, p < 0.05, unpaired t-test). In contrast, the RUNX2 mRNA/β-actin mRNA ratio was significantly low in all cases, and there was no significant difference between gastric cancer and normal mucosa. In normal gastric mucosa, basal expression levels of RUNX1 and CBFB are about 5-fold higher than RUNX3.
Overall, 56 specimens were examined, and the results of quantitative RT-PCR are summarized in Table II. Of 22 stage I cases, 11 (50%) did not express RUNX1 significantly. In the stage II, III and IV cases, 6 of 12 (50%), 11 of 14 (79%) and 7 of 8 (88%) showed reduction of RUNX1 expression. In summary, RUNX1 expression was reduced in 50% of earlier stage carcinomas, and this level increased to nearly 90%. On average, 62% of the clinical specimens showed reduced or no expression of RUNX1 in cancer tissue compared to surrounding normal mucosa. Likewise, on average, 69% of the clinical specimens showed reduced or no expression of RUNX3 in cancer tissue compared to surrounding normal mucosa. CBFB was also significantly downregulated in 32% of clinical samples, but this incidence and degree were lower than either RUNX1 and RUNX3 (Table II). The percentage of downregulation of each gene (RUNX1, 3 and CBFB) increased as the cancer stage progressed (p < 0.01–0.05, stage I and II vs. III and IV examined by the nonparametric Wilcoxon rank sum test). There was no significant difference between diffuse- and intestinal-type gastric cancers in RUNX family gene expression.
Table II. Frequent Loss of RUNX Family Gene Expression in Primary Gastric Cancer Specimens
Expression (no. of cases with loss of gene expression)
RUNX family gene expression in primary gastric cancers analyzed by in situ hybridization
In situ hybridization analysis showed that expression of RUNX1 and CBFB was also greatly reduced in clinical specimens (Figs. 3 and 4). The incidence of their downregulation estimated here was similar to that of quantitative RT-PCR. Figure 3a shows a whole-mount view of a surgically resected specimen of cancer tissue surrounded by normal mucosa. RUNX1 expression can be clearly detected within the normal gastric epithelial cell layer but only weakly in the cancer tissue (Fig. 3b). The pattern is more clearly visible at higher magnification. The corresponding regions in Figure 3c–e are shown in Figure 3f–h, clearly demonstrating that RUNX1 expression is mainly restricted in the normal epithelial cells and not evident in the surrounding mesenchymal cells. Figure 3e shows cancer cells with irregularly shaped atypical tubular structures lined by pleomorphic cells and enlarged irregular-shaped vesicular nuclei. Figure 3c–h demonstrate that cancer tissue does not significantly express RUNX1. The result of a negative control is shown when the antisense probe is replaced by a sense probe (data not shown).
Likewise, CBFB expression was reduced in cancer tissues. As revealed by quantitative RT-PCR, the incidence of CBFB expression in primary gastric cancers decreased in comparison to that in normal gastric mucosa, and the incidence of downregulation was lower than that of RUNX1 and RUNX3. In situ hybridization analysis showed that expression of CBFB was also greatly reduced in clinical specimens (Fig. 4d,e,g,h). Figure 4a shows a whole-mount view of a surgically resected specimen of cancer tissue surrounded by normal mucosa. CBFB expression can be clearly detected within the normal gastric epithelial cell layer but only weakly in the cancer tissue (Fig. 4b). The pattern is more clearly visible at higher magnification. The corresponding region in Figure 4c–e is shown in Figure 4f–h, clearly showing that CBFB expression is mainly restricted to the normal epithelial cells and is not evident in cancerous tissues. Figure 4d,e show cancer cells with irregularly shaped atypical tubular structures lined by pleomorphic cells and enlarged irregular-shaped vesicular nuclei. Figure 4g,h demonstrate that cancer tissue did not significantly express CBFB. These data from in situ hybridization coincides with those of quantitative RT-PCR (Fig. 2).
Reactivation of silenced RUNX3, but not for RUNX1 or CBFB, by TSA and 5′-AC
Reactivation of RUNX3 was observed in MKN28, MKN74 and NUGC3 cells treated with a combination of 5′-AC and TSA (Fig. 5a). Expression of RUNX3 was restored to high levels in these cells treated with the chemicals, but the level of expression was low with TSA alone or 5′-AC alone (Fig. 5a). In contrast, the expression of RUNX1 was not changed by the same treatment with them alone or in combination in gastric cancer cell lines MKN1, MKN45 and KATO-III, whose RUNX1 expression is low.
Typical data are shown in Figure 5b of change of RUNX family gene expression after treatment with these agents in gastric cancer cell lines analyzed by quantitative RT-PCR. The fold increases of the RUNX1, RUNX2, RUNX3 and CBFB mRNA/β-actin mRNA ratio was examined. In MKN74 cells, which have RUNX3 downregulation, RUNX3 expression increased about 300-fold after treatment with a combination of 5′-AC and TSA, but significant change could not be detected in RUNX1, RUNX2 or CBFB. In MKN45 cells, which express RUNX3, RUNX3 expression increased about 2- to 3-fold after treatment with a combination of 5′-AC and TSA. Significant changes could not be detected in RUNX1, RUNX2 and CBFB.
The molecular mechanism of multistep carcinogenesis of gastric cancer is still poorly understood.29, 30 Chronic infection by Helicobacter pylori is commonly associated with all subtypes of gastric cancers, particularly to intestinal-type gastric cancer. Infection results in chronic gastritis and is linked to gastric cancer pathogenesis31, 32, 33 through an intermediate stage involving intestinal metaplasia,34, 35 but the underlying mechanism of gastric carcinogenesis is still obscure. β-catenin mutation is more common in intestinal-type gastric cancers,35 whereas mutation in CDH1 is seen more often in diffuse-type gastric cancers, including familial gastric cancer,37, 38 however, neither mutation is exclusive to either subtype.
Recently, we reported that RUNX3 is a novel tumor suppressor of gastric cancer,22 but the role of other RUNX family genes, RUNX1, RUNX2 and CBFB, has not yet been clarified. RUNX1, RUNX2 and RUNX3 share the highly homologous region, the Runt domain, and bind on the same RUNX-binding site, suggesting that they have a potential to co-regulate target genes. Therefore, we examined the expression of RUNX family genes in gastric cancers and surrounding gastric mucosa in our study.
The expression of RUNX1 in gastric mucosa is particularly interesting, since our recent analysis revealed that RUNX1 expression is significantly high in gastric mucosa compared to small and large intestines (data not shown). Therefore, it is possible that RUNX1 plays an important role in the development and differentiation of gastric mucosa, and loss of RUNX1 expression plays some role in gastric carcinogenesis. Previous study indicates that frequent LOH is detected on 21q21-22 where RUNX1 is mapped.39 Sporadic point mutations in RUNX1 have been found in some 5% of acute leukemias.40, 41, 42 Familial cases of loss-of-function of RUNX1 causes newly identified disease, termed Familial Platelet Disorder with propensity to Acute Myelogenous Leukemia (FPD/AML).43 It is not known whether the family members of FPD-AML have higher incidence of gastric cancer. Our preliminary experiments suggest that the mutation in RUNX1 is very rare, if any, in gastric cancers (data not shown).
As previously reported, RUNX3 is repressed through promoter hypermethylation, and inhibitors of histone deacetylase and methyltransferase, tricostatin A and 5′-AC, respectively, could reactivate the gene. However, these reagents could not reactivate the expression of RUNX1 in gastric cancer cells, suggesting that inhibition of RUNX1 expression in gastric cancer is mediated by the mechanism other than the methylation of the promoter region. Although regulatory mechanism of RUNX1 expression has been studied,44, 45 further studies seem to be required for understanding the basis of the observation that we made. The results of our expression analysis and LOH analysis by others suggest that RUNX1 may be also in part involved in gastric carcinogenesis. Further examination would be necessary to clarify the role of RUNX1 in the development of gastric mucosa and gastric carcinogenesis.
CBFB encodes the β subunit of the Runt domain transcription factor. It was originally identified at the breakpoint of inv(16)(p13q22)/t(16;16)(p13;q22) of acute myeloid leukemia M4 type.18 At the molecular level, inv(16)/t(16;16) results in the creation of a novel fusion gene, CBFB (PEBPB2)/MYH11. CBFβ forms a heterodimer with RUNX proteins. Previous studies indicate that CBFβ is required for the functions of RUNX1, RUNX2 and RUNX3 and that CBFβ regulates their transcriptional activity through their DNA binding of RUNX1, RUNX2 and RUNX3 as a transcription factor.46, 47 Although the incidence and degree was lower than those of RUNX1 and RUNX3, CBFB was also downregulated in a significant fraction of gastric cancer specimens, suggesting the possibility that downregulation of CBFB leads to dysregulation of RUNX1 and RUNX3 function and may be involved in gastric carcinogenesis.
It has been shown that while Runx2 knockout mice survive to term but are not viable due to the absence of mature bone development,12, 13, 14, 15 and that Runx2 is a frequent target for proviral insertion in murine leukemia virus (MLV) induced T cell tumors in CD2-MYC transgenic mouse.16, 17 It has also been demonstrated that RUNX2 is ectopically expressed in breast cancer cells and that RUNX2, may provide a component of a mechanism that may explain the osteoblastic phenotype of human breast cancer cells that preferentially metastasize to bone.48 However, the expression of RUNX2 in both gastric mucosa and gastric cancers was very weak and, therefore, it is unlikely that RUNX2 has any role in gastric cancers. Since bone metastases from gastric cancer are sometimes observed in gastric cancer patients with poor prognosis, it would still be necessary to examine whether RUNX2 is involved in such cases. Further studies are necessary to clarify it.
In conclusion, the expression of RUNX3, as well as RUNX1 and CBFB, are downregulated in a significant portion of gastric cancer cases, suggesting that not only RUNX3 but also at least RUNX1 and CBFB may be involved in gastric cancer. Involvement of RUNX family genes in gastric carcinogenesis may be more widespread and complex than previously realized. Further examination is necessary to understand how expression of RUNX family genes is modulated during the transition from cell proliferation to differentiation in normal development and carcinogenesis.