Involvement of RUNX3 in TGF-β-dependent induction of endogenous p21 expression.
To determine whether RUNX3 is a critical downstream effector in TGF-β-mediated growth inhibition, we initially employed DNA microarray technology to screen for genes whose transcription is critically regulated by TGF-β1. The cell line used for this analysis was gastric cancer-derived SNU-16, whose growth is partially inhibited by TGF-β1, unlike most other gastric tumor lines (31
). To understand the mechanism by which TGF-β-inhibited cell growth, SNU-16 mRNAs were harvested at five different time points following TGF-β1 stimulation and screened with the Human 10K Digital Genomics cDNA array as previously described (16
). The result revealed that various genes encoding cyclins and cyclin-dependent kinase (CDK) inhibitors were induced by TGF-β1 treatment. Among the CDK inhibitors, p21
was the most dramatically up-regulated gene in the array, and its mRNA levels peaked 2 h after stimulation (Fig. ). In contrast, relatively little change was observed in the expression of other CDK inhibitors such as p27kip1
, and p19
(Fig. ). TGF-β1 stimulation of p21 protein levels was confirmed by Western blotting (Fig. ).
FIG. 1. TGF-β1-dependent induction of p21 expression in SNU16 cells and its requirement for RUNX3. (A) SNU16 cells were stimulated with TGF-β1 (2 ng/ml) for various periods, and their gene expression profile was examined by microarray analysis. (more ...)
To examine whether RUNX3 is involved in the TGF-β-dependent induction of p21 expression, SNU16 cells were transfected with an antisense RUNX3 construct and the stably transfected cell line SNU16-AS was isolated. Analysis by reverse transcription-PCR using two RUNX3-specific primers revealed greatly reduced endogenous RUNX3 transcript levels in SNU16-AS cells (Fig. ). The transfected antisense RUNX3 gene was confirmed to be transcribed by reverse transcription-PCR using primers specific for the plasmid vector and antisense RUNX3 (Fig. ). Northern blot analysis of SNU16 with or without TGF-β1 treatment showed the TGF-β treatment elevated endogenous p21 mRNA levels (Fig. , compare lanes 1 and 2, and E), thus confirming the microarray data (Fig. ). However, TGF-β1 induced a lower level of p21 mRNA when RUNX3 expression was inhibited by antisense RUNX3 (Fig. , compare lanes 2 and 4, and E). This result suggests that RUNX3 participates in TGF-β-dependent p21 induction.
The induction of p21 expression was examined further by using stable cell lines expressing wild-type RUNX3, RUNX3-R122C, and antisense RUNX3. Interestingly, SNU16 cell lines that express RUNX3 at high levels could not be obtained, whereas many RUNX3-R122C-overexpressing clones were isolated without much difficulty. We assumed that high levels of RUNX3 expression alone would be sufficient to induce p21 gene expression and inhibit growth. Therefore, stably transfected clones expressing RUNX3 and RUNX3-R122C at low levels were chosen for the present study; these were designated SNU16-RX3 and SNU16-R122C, respectively. First, we examined the effect of exogenous expression of RUNX3 on TGF-β-dependent p21 induction. The SNU16 and SNU16-RX3 cell lines were treated with increasing concentrations of TGF-β, and the mRNA level of p21 was measured by Northern blotting. As shown in Fig. , mRNA levels of p21 were increased by TGF-β treatment, while those of endogenous and exogenous RUNX3 were unchanged. Notably, SNU16-RX3 cells showed a higher level of p21 induction than SNU16 cells at low doses of TGF-β1 (Fig. , compare lanes 3 and 4). Next, we compared SNU16, SNU16-RX3, SNU16-R122C, and SNU16-AS cells. In the absence of TGF-β1, all these cell lines expressed p21 at similarly low levels (Fig. , lanes 1, 4, 7, and 10). However, when the cells were treated with low doses of TGF-β1, the induction profiles of the three cell lines differed significantly. SNU16-RX3 cells required only 0.03 ng/ml TGF-β1 to attain high p21 expression levels, while the control SNU16 cells needed 10 times more TGF-β1 (i.e., 0.3 ng/ml) to generate the same p21 mRNA levels (Fig. , compare lanes 3 and 5). Thus, the SNU16-RX3 cell line is markedly more sensitive to TGF-β1 than its parent line and SNU16-AS cells inhibited the induction of p21 expression at the same concentration of TGF-β1 (Fig. , lanes 10 to 12). Densitometric analysis of the band intensities in this Northern analysis is shown in Fig. . Since the vectors used for the SNU16-RX3 and SNU16-R122C cell lines are different (see Materials and Methods), the sizes of the RUNX3 and RUNX3-R122C transcripts could be different due to the different 3′ noncoding regions provided by the vectors (Fig. , lanes 4 to 6 and 7 to 9). It is unclear why several species of transcript are seen in the SNU16-AS line (Fig. , lanes 10 to 12). The transcripts are not due to multiple integrations of the transgene, because two other independent clones showed the same species of transcripts (data not shown). We think that these transcripts may be caused by inappropriate transcriptional termination.
Thus, our results demonstrate that TGF-β-induced p21
expression is mediated by RUNX3. Interestingly, in the absence of TGF-β1, low exogenous RUNX3
expression did not affect p21
expression, and the effect of RUNX3
expression on p21
expression was only observed in the presence of TGF-β1. It is known that TGF-β-activated SMADs activate the p21
) and that RUNX3 physically interacts and functionally cooperates with TGF-β-activated SMADs (8
). These observations suggest that p21
promoter activation by TGF-β1 requires cooperation between TGF-β1-activated SMADs and RUNX3. This possibility was examined as described below.
It is important to note that, unlike wild-type RUNX3 expression, RUNX3-R122C, which lacks the tumor suppressor activity, failed to increase TGF-β-dependent p21 induction. This result suggests that the role RUNX3 plays in TGF-β-dependent p21 induction is related to the tumor suppressor activity of RUNX3.
Involvement of RUNX3 in TGF-β-dependent cell growth inhibition.
We examined whether the increased sensitivity of p21 to TGF-β1 due to exogenous RUNX3 is also associated with enhanced cell growth inhibition. When SNU16-RX3, SNU16-R122C, and SNU16-AS cells were treated with very low concentrations of TGF-β1 (0.03 ng/ml) and counted, only SNU16-RX3 cells showed dramatically reduced growth in the presence of 0.03 ng/ml TGF-β1 (Fig. ). In contrast, the growth of SNU16, SNU16-R122C, and SNU16-AS cells was only marginally affected by TGF-β1 at the same concentration (Fig. ).
To examine whether high levels of RUNX3
expression alone may be sufficient to induce p21
gene expression and inhibit growth, we established RUNX3
-inducible cell lines using the MKN28 human gastric cancer cell line, which lacks RUNX3
expression and sensitivity to TGF-β1 (18
). We generated the MKN28-Lac-RX3 and MKN28-Lac-R122C lines that, respectively, express RUNX3
upon IPTG treatment. IPTG-induced RUNX3
expression coincided with elevated endogenous p21
expression (Fig. ), but RUNX3
expression failed to induce p21
(Fig. ). Densitometric analysis of the band intensities in Fig. is shown in Fig. . In agreement with these observations, IPTG treatment inhibited the growth of MKN28-Lac-RX3 but not MKN28-Lac-R122C (Fig. ). Taking these findings together, we conclude that RUNX3, but not RUNX3-R122C, is capable of inducing p21
expression and inhibiting cell growth.
The data presented in Fig. and suggest that RUNX3 participates in TGF-β1-dependent cell growth suppression by inducing p21 expression. These observations also explain why we could not obtain cells expressing RUNX3 at high levels.
RUNX3 binds to and activates the p21 promoter.
promoter contains five putative RUNX-binding sites, namely, three perfect matches to the consensus binding site called sites a to c and two imperfect matches called sites d and e (Fig. ) (21
). When we performed a ChIP assay, we found that the DNA fragment of the p21
promoter was immunoprecipitated by an anti-RUNX3 antibody (Fig. ). Thus, endogenous RUNX3 binds to the p21
promoter in vivo. We then examined whether RUNX3 must bind to the p21
promoter to induce p21
expression by subcloning the 2.3-kb-long DNA fragment of the p21
promoter upstream of the luciferase gene in the pGL3Basic vector (Promega) and performing a transient luciferase expression assay in the 293 cell line. Cotransfection with increasing amounts of the RUNX3
expression plasmid resulted in a dose-dependent increase in reporter activity (Fig. ). Mutation of each RUNX-binding site in the p21
promoter significantly decreased its responsiveness to RUNX3, except for site e. Mutation of all three perfectly matching RUNX-binding sites did not further reduce the responsiveness to RUNX3. This result indicates that sites a to d in the p21
promoter are mainly responsible for RUNX3-mediated p21
expression and that all four sites are required for full responsiveness. Site e also appears to be functional since weak responsiveness to RUNX3 remained even after sites a to d were all mutated. Mutations in both sites d and e abolished responsiveness to RUNX3; however, the basal activity of the reporter gene was also decreased. The mutation of all five RUNX-binding sites completely abrogated responsiveness to RUNX3 and even suppressed the basal activity of the reporter gene. One possible explanation is that the mutation of sites d and e may inactivate other essential transcription factor-binding sites. Further analysis is required to understand the role sites d and e play. Taken together, our results demonstrate that RUNX3 must bind the p21
promoter in order to stimulate its expression.
RUNX3 induces p21 expression in cooperation with SMADs.
Although TGF-β1 induces endogenous p21
expression and exogenous expression of RUNX3
activates the p21
promoter, we did not detect an increase in RUNX3
mRNA levels after TGF-β1 stimulation (Fig. ). Since p21
expression can be induced by TGF-β-activated SMADs (3
) and the physical interaction of RUNX3 with SMADs has been reported (8
), we speculated that the ability of RUNX3 to transactivate the p21
promoter may be turned on by TGF-β-activated SMADs. We tested this by cotransfecting MKN28 cells with the pGL3-p21 luciferase reporter plasmid and plasmids expressing SMAD3
in the presence or absence of the plasmid expressing RUNX3
. SMAD3 and SMAD4 could transactivate the p21
promoter by about twofold on their own (Fig. ), which is similar to what has been reported previously (25
on its own also transactivated p21
promoter activity by about threefold (Fig. ). However, when RUNX3
was cotransfected with SMAD3 and SMAD4, the transactivation activity was increased by up to 10-fold (Fig. ). Similar observations were made when the luciferase assay was performed with 293 cells (Fig. ). Thus, RUNX3 transactivates p21
expression in collaboration with SMADs. In contrast, RUNX3-R122C only weakly transactivated the p21
promoter and poorly cooperated with SMADs (Fig. ). These observations further substantiate the above findings, namely, that RUNX3
, but not RUNX3
, increases TGF-β1-dependent p21
expression (Fig. ). Collectively, these results indicate that RUNX3 induces p21
expression in collaboration with TGF-β1-specific SMADs. This explains why the SNU16-RX3 cell line showed much higher TGF-β1-induced p21
expression levels compared to its parental cell line and why SNU16-R122C cells failed to show any TGF-β1-induced p21
expression (Fig. ).
Recently, it was reported that RUNX1 and RUNX2 suppress p21
expression in cooperation with mSin3A (21
) and histone deacetylase 6 (39
), respectively. However, the activation of p21
expression by RUNX3 in cooperation with SMADs is not surprising because RUNX family members function as both transcriptional activators and repressors, depending on the partner proteins involved.
Recently, Seoane and others reported that Forkhead transcription factor FoxO, together with SMAD3 and SMAD4, binds to the promoter of the p21
gene and activates its transcription when HaCaT cells are treated with TGF-β (33
). It is known that Akt in the phosphoinositol 3 (PI3) kinase pathway phosphorylates FoxO to exclude it from the nucleus. Therefore, these results suggest that FoxO is a node for integration of TGF-β and PI3/Akt pathways. In their study, they also showed that in human keratinocyte and glioblastoma cells, TGF-β did not alter the level of phospho-Akt or the subcellular localization of FoxO, suggesting that involvement of FoxO in activation of the p21
gene varies from cell type to cell type. Although we did not see a significant effect of FoxO on the overall p21
gene expression in human gastric cancer cell lines, it would be worth examining further whether any other signals would coregulate p21
gene expression together with SMADs in stomach epithelial cells.
Differential DNA-binding affinity of wild-type RUNX3 and the R122C mutant.
As shown above, the RUNX3-R122C mutant neither induced p21 expression nor cooperated with SMADs as efficiently as wild-type RUNX3. We reasoned that the mutant RUNX3 protein may be defective in binding to its cognate target DNA sequences. Therefore, we examined whether RUNX3-R122C bind to the consensus RUNX-binding DNA sequence. As shown in Fig. , the RUNX3 protein exhibited higher affinity toward the 32P-labeled DNA probe bearing the RUNX-binding site than the RUNX3-R122C protein. A quantitative EMSA using serial amounts of nuclear protein extract (Fig. ) and densitometric analysis (data not shown) revealed that the RUNX3-R122C mutant has only one-fourth of the DNA-binding affinity of the wild-type molecule. That RUNX3 and RUNX3-R122C were expressed at similar levels in the transfected cells was confirmed by Western blotting (Fig. ).
FIG. 6. Comparison of the DNA-binding affinities of RUNX3 and RUNX3-R122C by EMSA. (A) Similar amounts of nuclear extracts obtained from 293 cells transfected with pCS4-myc, pCS4-myc-RUNX3, and pCS4-myc-RX3-R122C were subjected to EMSA using the 32P-labeled RUNX (more ...) Differential SMAD-binding affinity of wild-type RUNX3 and the R122C mutant.
Since RUNX3-R122C binds to the cognate target DNA sequence, albeit more weakly than wild-type RUNX3, we asked whether there are other defects in RUNX3-R122C that would help explain its profound inability to mediate TGF-β1-induced p21 expression. We thus compared the binding affinities of wild-type RUNX3 and its R122C mutant for SMADs by cotransfecting 293 cells with myc-tagged RUNX3 or RUNX3-R122C and HA-tagged SMAD3 and/or SMAD4. Coimmunoprecipitation analysis revealed that RUNX3-R122C binds more weakly to SMAD3 and SMAD4 than wild-type RUNX3 (Fig. ). The expression levels of the transfected SMAD, RUNX3, and RUNX3-R122C genes were confirmed to be comparable by Western blotting.
FIG. 7. Interaction between RUNX3 and TGF-β-specific SMADs. (A) 293 cells were transfected with a fixed amount of pCS4-myc-RUNX3, pCS4-myc-RX3-R122C, pCS4-HA-SMAD3, and/or pCS4-HA-SMAD4 as indicated. (Upper panel) The physical interactions between RUNX3 (more ...)
These observations demonstrate that the ability of RUNX3 to induce p21 expression is controlled by SMADs which are activated by TGF-β1 and that RUNX3-R122C lacks the ability to activate p21 expression due to the combination of reduced DNA- and SMAD-binding affinities.
Requirement of RUNX3 for p21 expression in vivo.
Our finding that RUNX3
is required for p21
expression in gastric epithelial cell lines prompted us to examine whether RUNX3
expression is also required for p21
expression in mouse and human gastric epithelia. Immunohistochemical staining of the newborn Runx3+/−
mutant stomach showed that Runx3
is expressed in epithelial cells of the glandular stomach (Fig. ) as reported previously (18
). Interestingly, staining of serial sections with an anti-p21 antibody revealed that the cells expressing Runx3
and those expressing p21
overlap (compare Fig. ). Furthermore, p21
expression in Runx3−/−
stomach epithelial cells was significantly lower than that in normal stomach epithelial cells (compare Fig. ). The low level of p21
expression in the Runx3−/−
mouse gastric epithelium is consistent with the hyperplasia observed in these mutant mice.
FIG. 8. RUNX3 and p21 expression patterns in mouse and human gastric epithelium overlap. (A to F). Pyloric areas of stomach tissues were obtained from newborn normal, Runx3+/−, and Runx3−/− mutant mice, and the expression levels (more ...)
Since RUNX3 is frequently inactivated in primary human gastric cancer specimens, we examined whether p21 expression was also suppressed in RUNX3-inactivated gastric cancer tissues. Thus, 16 surgically resected primary gastric cancer specimens were subjected to in situ hybridization analysis. Typical results are shown in Fig. . These data reveal that p21 expression is restricted to the gastric epithelial cells that express RUNX3 (compare Fig. ). In the cancerous regions that do not express RUNX3, p21 expression is also significantly decreased. These results together suggest that RUNX3 expression is required for the in vivo expression of p21.
Targeted inactivation of Runx3 in mice resulted in gastric epithelial hyperplasia due to stimulated proliferation and suppressed apoptosis accompanied by a decreased sensitivity to TGF-β1. However, the role Runx3 plays in the TGF-β signaling pathway has not been established. In this investigation, we studied the molecular mechanism by which RUNX3 inhibits cell growth and found that RUNX3 inhibits gastric epithelial cell growth by inducing p21 gene expression in response to TGF-β1. The involvement of RUNX3 in TGF-β-dependent p21 induction is revealed by the fact that exogenous expression of RUNX3 increased endogenous p21 expression in response to TGF-β1. Moreover, suppression of RUNX3 expression decreased TGF-β1-induced endogenous p21 expression (Fig. ). Furthermore, in mouse and human gastric epithelia, RUNX3 expression coincided with p21 expression (Fig. ). That p21 is a direct target of RUNX3 is indicated by the observations that the p21 promoter is activated by RUNX3 and that mutation of RUNX-binding sites within the promoter significantly decreased this activity (Fig. ). The p21 promoter was synergistically activated when RUNX3 and SMAD3 were coexpressed. This is the first report that shows RUNX3 is an essential component of the TGF-β-mediated p21 induction that is critical for TGF-β-mediated cell growth inhibition.
plays a key role in cell cycle arrest at the G1
stage by inhibiting CDKs and was originally identified as an important transcriptional target of p53. One of the critical targets of G1
cyclin CDKs is the retinoblastoma protein (pRb), which, in its hypophosphorylated state, binds and sequesters transcription factors required for the entry of the cell into S phase (2
). Our results demonstrate that RUNX3 also directly influences the cell cycle machinery by inducing p21
expression in response to TGF-β. The ability to induce p21
expression and to inhibit cell growth is essential for the tumor suppressor activity of p53. The TGF-β signaling cascade also induces p21
gene expression. Therefore, our finding that RUNX3 directly regulates p21
expression in response to TGF-β stimulation suggests that the tumor suppressor activity of RUNX3
is at least partly associated with its ability to induce p21
expression. This claim is further supported by the observation that the RUNX3-R122C mutant, which lacks tumor suppressor activity, failed to induce p21
expression in response to TGF-β1. Reduced DNA- and SMAD-binding affinity appears to be responsible for this loss of activity. It is not known whether the loss of transactivation activity of RUNX3-R122C is specific to the p21
promoter. However, the R122C mutation may not abolish all target gene expression since the runt mutant in Drosophila
that lacks all DNA-binding activity still retains significant in vivo activity (40
It is worth mentioning that p53 controls the cell cycle by transcriptionally activating p21 expression, which then inhibits pRb phosphorylation. Interestingly, RUNX3, which is frequently inactivated in human cancers, controls the cell cycle by a similar mechanism, although the stimulating signals differ, namely, cellular stress for p53 and TGF-β stimulation for RUNX3. Well-known tumor suppressors such as pRb, p53, p16INK4a, and p14ARF directly influence the cell cycle machinery, and RUNX3 is a new addition to this list.