Abstract
The adenomatous polyposis coli gene (APC) is a tumor suppressor gene that is inactivated in most colorectal cancers. Mutations of APC cause aberrant accumulation of β-catenin, which then binds T cell factor–4 (Tcf-4), causing increased transcriptional activation of unknown genes. Here, the c-MYC oncogene is identified as a target gene in this signaling pathway. Expression of c-MYC was shown to be repressed by wild-type APC and activated by β-catenin, and these effects were mediated through Tcf-4 binding sites in the c-MYC promoter. These results provide a molecular framework for understanding the previously enigmatic overexpression of c-MYC in colorectal cancers.
Most human colorectal tumors are initiated by inactivation of the APC tumor suppressor gene, located on chromosome 5q21 (1). APC is a cytoplasmic protein that can bind to and promote the degradation of β-catenin (2). Among β-catenin functions is the ability to bind members of the Tcf family of transcription factors and activate gene transcription (3). Accordingly, human colorectal tumors with APC or β-catenin mutations exhibit increased β-catenin/Tcf–mediated transcription (4, 5). However, the downstream targets of this β-catenin/Tcf-4–regulated transcription are unknown. This study was undertaken to define those targets and thereby gain clues to the mechanisms through which APC affects cellular growth.
To evaluate the transcriptional effects of APC, we studied a human colorectal cancer cell line (HT29-APC) containing a zinc-inducible APC gene and a control cell line (HT29–β-Gal) containing an analogous inducible lacZ gene (6). Both endogenous APC alleles in HT29 cells contain truncating mutations, and restoration of wild-type (WT) APC expression results in growth inhibition and apoptosis. Upon induction, APC protein is synthesized rapidly and reaches maximal levels by 9 hours (7). By 12 hours, a significant fraction of the cells display morphological signs of apoptosis. Because we were interested in identifying changes in gene expression that directly relate to restoration of APC function and not apoptosis, we analyzed the HT29-APC cells 9 hours after APC induction.
To evaluate changes in gene expression, we used serial analysis of gene expression (SAGE), a technique that allows the quantitative evaluation of cellular mRNA in an unbiased manner (8). In brief, the method is based on the use of short sequence tags [15 base pairs (bp)] generated from defined positions within each transcript. Expression levels are deduced from the abundance of individual tags in a sample. SAGE analysis of 51,622 and 55,846 tags from APC-induced and control cells, respectively, allowed identification of 14,346 different transcripts (9), most of which were expressed at similar levels in the APC-induced and control cells. Of the 30 tags showing significant differences in expression (10), 14 were overexpressed and 16 were repressed in APC-induced cells. Because biochemical studies have indicated that APC represses β-catenin/Tcf-4–mediated transcription (4,5), we focused on the latter transcripts. One of the three most highly repressed transcripts was a tag corresponding to the c-MYC oncogene (eight tags in HT29–β-Gal compared with zero in HT29-APC). This repression was confirmed at the mRNA and protein level by Northern (RNA) blot (Fig. 1A) and immunoblot (Fig. 1B) analysis, respectively. Repression of c-MYC mRNA and protein was evident within 6 hours after zinc induction and within 3 hours after the first detection of APC protein (Fig. 1).
c-MYC expression after APC induction. (A) Total RNA was isolated from the ZnCl2-treated cells at the indicated times and evaluated on Northern blots (10 μg of RNA per lane) that were hybridized with a probe for c-MYC or a control probe for elongation factor 1α mRNA (EF1). (B) Total cellular proteins from the same cells were separated by SDS–polyacrylamide gel electrophoresis and subjected to immunoblotting with a monoclonal antibody to c-MYC (9E10, Santa Cruz Biotechnology). An identical blot probed with a monoclonal antibody to p53 shows that equal amounts of protein were loaded in each lane.
These results suggested that APC might directly modulate c-MYC transcription through β-catenin/Tcf-4. To assess this possibility, we isolated a 2.5-kb genomic fragment encompassing the c-MYC promoter, inserted it upstream of a luciferase reporter gene (11), and then tested the construct for responsiveness to APC (12). This c-MYCpromoter region conferred significant transcriptional activity to the basal reporter gene when transfected into human colorectal cancer cells, and this activity was significantly repressed by APC (Fig. 2). Nested deletions of the promoter were used to map the APC-responsive region to a fragment containing nucleotides (nt) −1194 to −484 relative to the TATA box at the c-MYC major transcription start site (Fig. 2, A and B). Testing of restriction fragments spanning the promoter revealed two responsive regions, one located in fragment B (nt −1194 to −741) and the other in fragment C (nt −741 to −484) (Fig. 3, A and C).
APC- and β-catenin–responsive regions within the c-MYC promoter. (A) Map of the c-MYC promoter showing the restriction sites used for generating nested deletions (Del constructs) and fragments (Frag constructs). The horizontal lines represent the sequences in each reporter construct, which were placed upstream of a minimal promoter and luciferase cassette. P1 and P2 are start sites of transcription; P2 is the major start site. (B and C) SW480 cells were cotransfected with the indicated reporter plasmids plus an APC expression construct or a control plasmid. The bars represent luciferase activity in the cells transfected with APC relative to that in cells transfected with the control plasmid. Luciferase activity (mean ± SD) was measured in three separate experiments. The constitutive reporter activity (APC off) of the deletion constructs (Del-1 to Del-4) varied less than twofold, ranging from 3520 to 6859 as expressed in arbitrary luciferase light units. The constitutive activity of the Frag-A, -B, -C, -D, and -E was 364, 3050, 1063, 1754, and 976, respectively. (D) 293 cells were cotransfected with the indicated reporters plus a β-catenin expression construct or a control plasmid. The increase in luciferase activity in the β-catenin transfectants relative to the control transfectants is plotted on they axis. Data are presented as the mean ± SD determined from three separate transfections.
Tcf-4 binding elements (TBEs) within the c-MYC promoter. (A) Map of the c-MYC promoter, indicating the 2.5-kb region containing the APC- and β-catenin–responsive elements. The fragment containing the WT sequence of the promoter (TBE1/2) contains TBE sites near both ends. This fragment was engineered to contain mutations in either site 1 (TBE1m/2), 2 (TBE1/2m), or both sites 1 and 2 (TBE1m/2m), and each fragment was placed upstream of a minimal promoter and luciferase reporter. Reporters containing four copies of TBE1 (4×TBE1) or TBE2 (4×TBE2) or a mutant TBE2 (4×TBE2m), in the absence of any additional genomic sequences, were constructed similarly. (B) SW480 cells were cotransfected with the indicated reporter plasmids plus an APC expression construct or a control plasmid. Data are presented as in Fig. 2B. (C) 293 cells were cotransfected with the indicated reporters plus a β-catenin expression construct or a control plasmid. The increase in luciferase activity in the β-catenin transfectants relative to the control transfectants is plotted on the y axis. Data are presented as the mean ± SD determined from three separate transfections. (D) Electrophoretic mobility shift assay. Oligonucleotides containing TBE1 or TBE2 sequences (wt) or mutants (mt) with nucleotide substitutions at critical positions were end-labeled with [γ-32P]ATP and incubated with 0.5 μg of a GST–fusion protein containing the DNA-binding domain of Tcf-4. DNA-protein complexes were separated by electrophoresis and detected as “shifts” from the position of free probe. Unlabeled oligonucleotides (250 ng) were used as competitors (Comp.) in some reactions.
If the effects of APC on c-MYC transcription were mediated through inhibition of β-catenin/Tcf-4–regulated transcription, then the c-MYC promoter should be activated by β-catenin. Previous studies have shown that β-catenin/Tcf-4 transcription can be activated by exogenous expression of a mutant β-catenin gene in the human kidney cell line 293. The β-catenin construct used for these experiments was mutated at codon 33, rendering it insensitive to down-regulation by the endogenous WT APC in 293 cells (5). The c-MYC reporter was found to be significantly activated by β-catenin in this line. Using the nested deletion and restriction fragment constructs noted above, we found that the region of the c-MYC promoter that conferred β-catenin responsiveness was the same region (fragments B and C) shown to be APC-repressible in colorectal cancer cells (Fig. 2D).
Analysis of the c-MYC promoter sequence revealed one potential Tcf-4 binding site (13) within fragment B (TBE1) and another within fragment C (TBE2) (Fig. 3A). To test the functional significance of these sites, we created fragments of the c-MYC promoter in which one or both binding sites were eliminated by nucleotide substitutions (14). Mutation of either TBE1 or TBE2 reduced the activity of the c-MYC promoter fragment by 50%. Importantly, deletion of both sites completely removed APC repression and β-catenin activation from the reporter, whereas deletion of either element alone did not abrogate responsiveness (Fig. 3, A to C).
We also tested the TBE1 and TBE2 elements in isolation. Constructs containing four tandem copies of either TBE1 or TBE2 upstream of a minimal promoter (15) conferred β-catenin responsiveness and APC repression to a downstream luciferase reporter (Fig. 3, A to C). In all cases, the responsiveness of the reporter containing TBE2 sites was greater than that obtained with TBE1 sites. Nucleotide substitutions within TBE1 or TBE2 that would be expected to abolish Tcf-4 binding abrogated responsiveness to APC and β-catenin (Fig. 3, A to C). Finally, to confirm the direct nature of the responsiveness, we tested the ability of Tcf-4 produced in bacteria to bind TBE1 and TBE2 (16). Tcf-4 bound both TBE1 and TBE2, as judged by electrophoretic mobility-shift assay, and this binding was abrogated by the same nucleotide substitutions that eliminated transcriptional responses (Fig. 3D).
On the basis of these data, we propose that in normal colorectal epithelial cells, WT APC prevents β-catenin from forming a complex with Tcf-4 and activating c-MYC. In colorectal tumors with APC mutations or activating β-catenin mutations, increased β-catenin/Tcf-4 activity leads to overexpression of c-MYC, which then promotes neoplastic growth. Consistent with this model, expression of a dominant-negative Tcf-4 in colorectal cancer cells with mutant β-catenin (HCT116) or mutant APC (SW480) significantly reduced the endogenous levels of c-MYC (Fig. 4). This model is also consistent with the powerful oncogenic activities of c-MYC (17) and provides an explanation for two long-standing quandaries. First, it has been extensively documented that c-MYC is overexpressed at the RNA and protein levels at both early and late stages of colorectal tumorigenesis (18). However, unlike in some other cancers, where the c-MYC gene is rearranged or amplified (19), genetic alterations of c-MYC are rare in colorectal tumors, and the cause of the overexpression has been unknown (20). The only clue to this mechanism has come from chromosome transfer experiments, in which it was shown that an extra copy of chromosome 5 can repress c-MYC transcription and inhibit neoplastic growth (21). This repression fits well with the molecular data presented here on APC which resides on chromosome 5q21.
Repression of c-MYC expression by a dominant-negative Tcf-4. Exponentially growing HCT116 and SW480 cells were mock- infected (Control) or infected with adenovirus encoding dominant-negative Tcf4 (DN-Tcf4) or β-galactosidase (β-Gal) at a multiplicity of infection of 100. Total cellular proteins were isolated 24 hours after infection and subjected to immunoblotting with a monoclonal antibody to c-MYC (C-19, Santa Cruz Biotechnology). An identical blot probed with a monoclonal antibody to β-catenin (Transduction Laboratories) shows that β-catenin is intact and that equal amounts of protein were loaded in each lane. The recombinant adenoviruses were constructed by means of the AdEasy system (26), the details of which are available upon request.
The second enigma involves the cyclin-dependent kinase inhibitor p16INK4a. Most tumor types exhibit genetic alterations of the p16INK4a growth-inhibitory pathway through direct mutation of p16INK4a, its neighbor p15INK4b, or its downstream targets Rb, cdk4, or cyclin D1 (22). Colorectal cancers are a notable exception, in that few mutations of any of the genes in this pathway occur (22, 23). The activation of c-MYC through APC inactivation would explain this, as c-MYC expression can bypass p16INK4a- and p15INK4b-mediated growth arrest (24).
Note added in proof: Consistent with a critical role for β-catenin/Tcf4–stimulated c-MYC expression in promoting intestinal cell proliferation, recent genetic studies in mice indicate that Tcf-4 is required to maintain the proliferative compartment in intestinal crypts (25).
↵* To whom correspondence should be addressed. E-mail: kinzlke{at}welchlink.welch.jhu.edu
