Synergy Between Tumor Suppressor APC and the β-Catenin-Tcf4 Target Tcf1

See allHide authors and affiliations

Science  17 Sep 1999:
Vol. 285, Issue 5435, pp. 1923-1926
DOI: 10.1126/science.285.5435.1923


Mutations in APC or β-catenin inappropriately activate the transcription factor Tcf4, thereby transforming intestinal epithelial cells. Here it is shown that one of the target genes of Tcf4 in epithelial cells is Tcf1. The most abundant Tcf1 isoforms lack a β-catenin interaction domain. Tcf1 −/−mice develop adenomas in the gut and mammary glands. Introduction of a mutant APC allele into these mice substantially increases the number of these adenomas. Tcf1 may act as a feedback repressor of β-catenin–Tcf4 target genes and thus may cooperate with APC to suppress malignant transformation of epithelial cells.

The tumor suppressor geneAPC, first identified in a dominantly inherited disorder termed familial adenomatous polyposis, is mutated in the vast majority of colorectal cancers (1). APC's principal role is that of a negative regulator of the Wnt signal transduction cascade (2). APC resides in a large complex with axin, GSK3, and the Wnt effector β-catenin (3). In this complex, the serine kinase GSK-3β constitutively phosphorylates β-catenin at a set of regulatory NH2-terminal Ser/Thr residues, thereby targeting β-catenin for ubiquitination by β-TrCP and for subsequent proteasomal degradation (4). Wnt signaling stabilizes β-catenin. In the nucleus, β-catenin binds to Tcf/Lef transcription factors. The bipartite complex then activates transcription of Tcf target genes (5). In the absence of signaling, Tcf factors repress transcription by interaction with Groucho transcriptional repressors or with CBP (6).

Loss of APC leads to the nuclear accumulation of β-catenin, which constitutively binds to Tcf4 (7), a Tcf family member specifically expressed in epithelia of the intestine and mammary gland (8). In some colorectal cancers that carry wild-type APC as well as in several other types of cancer, dominant mutations alter one of the four regulatory NH2-terminal Ser/Thr residues of β-catenin. This also leads to the inappropriate formation of β-catenin–Tcf complexes in the nucleus (9).

Expression of Tcf1, a gene encoding another Tcf family member, is largely restricted to T lineage lymphocytes in adult tissues and cell lines (10). However, colorectal cell lines have also been reported to express appreciable amounts of Tcf1(11). Confirming the latter observation, we detectedTcf1 mRNA by Northern (RNA) blot analysis in five of six colorectal cell lines (12). Three of these areAPC mutants (SW480, HT-29, and DLD1), and two others (LS174T and HCT116), carry oncogenic mutations in β-catenin. The cell line that did not express Tcf1 (RKO) is wild-type for bothAPC and β-catenin, suggesting that Tcf1expression might normally be regulated by these genes. We also detected nuclear Tcf1 protein in normal human tissues: in proliferating intestinal epithelial cells and in the basal epithelial cells of mammary gland epithelium (13) (Fig. 1). The most abundant Tcf1 isoforms lack a β-catenin interaction domain (10). Because they retain their Groucho interaction domain, they are likely to act as negative regulators of Wnt signaling.

Figure 1

Tcf1 is expressed in (A) the nucleus of fetal and adult epithelial cells of the intestine (a 16-week fetal sample is shown) and (B) in basal epithelial cells of the mammary gland tissue. Bars, 0.1 mm.

To test whether Tcf1 is a target of Tcf4, we used a transfectant derived from the APC−/− HT29 cell line, which inducibly expresses wild-type APC (14). This transfectant previously allowed the identification of another Tcf4 target, c-Myc (15). APC expression was induced in HT29-APC cells for 20 hours. The cells remained attached and were >95% viable. Northern (RNA) blot analysis revealed a consistent four- to fivefold decrease in steady-state mRNA levels for Tcf1and c-Myc (Fig. 2A), but no changes in the levels of Ep-Cam and γ-actinmRNAs. This experiment indicated that Tcf1 is regulated by APC, and therefore by β-catenin–Tcf4.

Figure 2

Tcf1 expression is regulated by APC and β-catenin–Tcf4. (A) Zn2+-induced (+) expression of APC in HT29-APC colon cells results in reduced expression of Tcf1 and c-Myc, but not of Ep-Camand γ-actin control genes. Zn2+-induction of a control cell line, HT29-βGal (14), had no effect. (B) Schematic representation of the enhancer region upstream of the two Tcf1 promoters (arrows). GenBank accession number: AF163776. A reporter constuct was made containing the Tcf responsive region (TRR) with its own promoter and theluciferase gene. This TRR-luciferase reporter was used in all subsequent transfection studies. Gray boxes represent the two Tcf consensus sites in the TRR [TTCAAAGC and ATCAAAGG; Tcf consensus: (A/T)(A/T)CAA(A/T)GG (2)]. Black boxes indicate the first three exons ofTcf1. X, Xba I; S, Sac II; P, Pst I. (C) TRR responsiveness in IIA1.6 B cells. Tcf4/β-catenin readily activates the TRR-luciferase reporter, but not the Renilla control. Transfections in (C) and (D) were carried out as in (7). (D) The activity of the TRR fragment in LS174T colon cells is dependent on Tcf4. Increasing amounts of dominant-negative ΔNTcf4 (ΔN) inhibited TRR activity.

The human Tcf1 gene is transcribed from two closely spaced promoters (10). We sequenced 1.2 kb directly upstream of promoter I and found the region to be a CpG island containing two potential Tcf-binding motifs (Fig. 2B). The region acted as an enhancer, both in the context of promoter I and of a heterologous promoter (12). We tested the inducibility of the putative enhancer fragment by β-catenin and Tcf expression constructs in our “model” B cell line IIA1.6, which lacks endogenous Tcf/Lef factors (7, 16). The combination of β-catenin and Tcf4 transactivated the enhancer three- to fourfold in a transient reporter assay (Fig. 2C). Furthermore, expression of a dominant-negative Tcf4 (ΔNTcf4, which lacks the β-catenin interaction domain) inhibited enhancer activity in LS174T colorectal cancer cells (Fig. 2D).

Tcf1-deficient mice develop a progressive block in early thymocyte development (17). Nevertheless,Tcf1 −/− mice have functional peripheral T cells, are fully immunocompetent, and live for over a year (18). Prompted by a possible link between Tcf4 activity andTcf1 expression in the intestine, we performed autopsies onTcf1 −/− mice of various ages. Unexpectedly, we observed mammary gland adenomas and polyplike intestinal neoplasms in these mice (Fig. 3A). These lesions were never observed in littermates. Histological examination (13) of the intestinal and mammary gland lesions revealed typical epithelial polyps and adenoacanthomas, respectively, expressing high levels of cytoplasmic and nuclear β-catenin (Fig. 3, B and C). Significantly, the APC protein appeared absent in the intestinal adenomas (Fig. 3D).

Figure 3

Tcf1−/− mice develop spontaneous intestinal and mammary gland adenomas. (A) Fifteen percent of Tcf1−/− mice developed intestinal adenomas during the first year of life (top), whereas 25% of Tcf1−/− females developed mammary gland adenoacanthomas (bottom). (B) Immmmunohistological examination of spontaneous intestinal neoplasms demonstrated high levels of cytoplasmic and nuclear β-catenin. For comparison, see the nontransformed epithelium in Fig. 4, A and B. (C) Mammary gland lesions were typed as adenoacanthomas and found to express high levels of cytoplasmic and nuclear β-catenin. (D) Intestinal adenomas have lost expression of APC as analyzed by immunohistochemistry (open arrow), compared with surrounding normal epithelium (closed arrow).

One possible explanation for this tumor phenotype is that Tcf1 acts as a feedback transcriptional repressor of β-catenin–Tcf4 target genes and that disruption of this negative feedback loop would allow the formation of epithelial tumors much like the loss of APC. This notion predicts synergy between the loss of Tcf1 and of APC. To test this, we crossed the Apc allele Multiple intestinal neoplasia (Min) into the Tcf1−/− strain.Min/+ mice develop multiple polyps mostly in the small intestine (19). They infrequently develop extraintestinal neoplasia, notably adenoacanthomas in the mammary gland (20). Min/+Tcf1 −/− mice displayed a marked enhancement of the intestinal Min/+ phenotype (Table 1). Adenomatous polyps were observed throughout the entire intestinal tract. Although the intestinal polyps tended to be larger than those of Min/+ mice, they were of similar histology and did not show any sign of tumor progression (Fig. 4A). All intestinal neoplasms that were analyzed by immunohistochemistry (stomach, small intestine, and colon) expressed high levels of β-catenin (for example Fig. 4, B and F). In addition, all females carried adenoacanthomas of the mammary gland by 8 weeks of age, while a substantial number of older male mice developed similar lesions (Table 2 and Fig. 4C). Again, the lesions typically expressed high levels of β-catenin compared with surrounding nontransformed cells (Fig. 4D). About 60% of the mice at 4 months of age had adenomas and adenoacanthomas of the salivary glands (Fig. 4E). The enhanced neoplastic phenotype was not observed in Min/+Tcf1+/− littermates, ruling out any influence of genetic background. These observations reveal a strong genetic interaction between APC and Tcf1.

Figure 4

Histology and β-catenin expression in adenomas of Min/+Tcf1−/− mice. (A) Polyps in the small intestine represent adenomatous lesions with moderate dysplasia, but without invasive behavior as visualized by hematoxylin and eosin stain. (B) High level of β-catenin expression in the polyps is revealed by immunohistochemistry (closed arrow), as compared with normal epithelium (open arrow). (C) A mammary gland–derived tumor showed an intimate mixture of glandular and epidermoid tissue, typical for adenoacanthomas. (D) Cells loaded with β-catenin in the adenoacanthoma (closed arrow) illustrate the involvement of the Wnt signal transduction pathway. Nontransformed epithelium demonstrates normal levels of β-catenin (open arrow). (E) In an early salivary gland adenoacanthoma and (F) an adenomatous lesion of the stomach, abundant nuclear β-catenin (closed arrow) illustrates the involvement of the Wnt pathway.

Table 1

Min/+Tcf1−/− mice demonstrated a 10-fold increase in the formation of intestinal neoplasms compared with Min/+ mice. ND, not done (mice were killed at 4 months).

View this table:
Table 2

At the age of 2 months, allMin/+Tcf1−/− females (F) developed mammary adenoacanthomas. Min/+Tcf1−/− males (M) developed the same type of adenoacanthomas, although at lower incidence. Mammary gland neoplasms did not occur inMin/+Tcf1 + /− mice of either sex. Mice were killed at the indicated age.

View this table:

Insight into the nature of the genetic program activated by Tcf4 has come from a gene disruption experiment. Mice deficient in Tcf4 develop normally, but die shortly after birth due to the absence of cycling epithelial progenitor cells in the prospective crypts of the small intestine (21). β-catenin–Tcf4 signaling appears to activate or maintain a progenitor cell phenotype. In concordance with this, recent reports have identified c-Myc andcyclinD1 as target genes of Tcf4 (15,22). Our data indicate that Tcf1 expression in epithelial cells is similarly controlled by APC and β-catenin–Tcf4. The genetic evidence indicates that Tcf1 serves as a negative-feedback regulator in APC-related carcinogenesis. The DNA-binding HMG boxes of the four mammalian Tcf/Lef proteins are essentially identical, implying that they may regulate the same target genes (2). We propose a model in which the transcriptional activation of target genes such as c-Myc and cyclinD1 by β-catenin–Tcf4 is counteracted by repressor isoforms of Tcf1.

It will be of interest to analyze the status of theTCF1 locus in human breast or colon cancer. TheAPC (5q2.1) and TCF1 (5q31.1) loci are linked in humans (23); large deletions on chromosome 5q could simultaneously inactivate both genes. In mice, the genes reside on separate chromosomes. Int-1/Wnt-1 was originally cloned as a proto-oncogene activated in breast epithelium by mouse mammary tumor virus integrations (24). By analogy to our observations, Int-1/Wnt-1 transgenic mice develop neoplasms in mammary glands as well as in salivary glands (25). In light of this observation, and given that APC functions as a regulator of Wnt signaling in a large variety of tissues, it has been surprising that APC's tumor suppressor activity appears predominantly relevant for intestinal cancer. Our data suggest that Tcf1 cooperates with APC and that the combination of the two activities is particularly important for the prevention of mammary neoplasms in mice.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: h.clevers{at}


Stay Connected to Science

Navigate This Article