PerspectiveGene Expression

A Wnt-fall for Gene Regulation: Repression

See allHide authors and affiliations

Science Signaling  30 Sep 2008:
Vol. 1, Issue 39, pp. pe43
DOI: 10.1126/scisignal.139pe43


A major endpoint of the canonical Wnt signaling pathway is a change in the transcription of target genes. The transcription factors lymphoid enhancer factor (LEF) and T cell factor (TCF) serve as the main gatekeepers of these changes by selecting genes to be targeted by the transcriptional coregulator β-catenin and by defining how target gene expression will be altered. Most research has focused on LEF/TCF:β-catenin–mediated activation of transcription, but there have been some reports that suggest that this complex also directly represses transcription. A recent study uncovered a new mode of repression of Wnt target genes in which recognition of a novel DNA element by TCF specifies that β-catenin acts as a transcriptional repressor.

LEF/TCFs and Wnt Target Gene Regulation

Wnts are secreted glycoproteins that drive signaling pathways contributing to cell fate determination, spatial-temporal patterning, and cell motility (1). Wnts are ligands of Frizzled:low-density lipoprotein (LDL) receptor–related protein receptor complexes on the cell membrane. In the canonical Wnt signaling pathway, the binding of Wnt to its receptor leads to the stabilization of a cytoplasmic pool of β-catenin (known as armadillo in Drosophila). Stabilized β-catenin translocates to the nucleus and interacts with the DNA binding transcription factors lymphoid enhancer factor (LEF) and T cell factor (TCF) (there are three mammalian TCFs: TCF-1, TCF-3, and TCF-4) to drive the expression of Wnt target genes. LEF and the three TCFs recognize a DNA sequence known as a Wnt response element (WRE: CTTTGWWS, Fig. 1), and it is primarily the binding of LEF or TCFs (LEF/TCFs) to these elements that enables Wnt signaling to change gene expression programs.

Fig. 1

LEF/TCFs regulate transcription by multiple mechanisms. (A) LEF/TCF transcription factors bend the CTTTGWWS sequence to influence interactions among proteins bound to nearby DNA sequences. DNA-bending–dependent interactions can create a strong transcriptional regulator, such as an enhancer, independently of Wnt and β-catenin. An example is the enhancer of the gene encoding the TCR α chain (19, 22). (B) LEF/TCF proteins recruit members of the Groucho family of repressors (TLE in mammalian systems) to silence gene transcription. An example is the repressor action of TCF-3 in stem cells (25). (C) The canonical mode of Wnt signaling directs LEF/TCF proteins to recruit β-catenin to target genes for activation. In this situation, the CTTTGWWS motif functions as a WRE. Two activation domains in β-catenin (“+”) within the N- and C-termini are involved. (D) There are a few examples of direct repression of gene transcription by LEF/TCF:β-catenin complexes binding to WREs (see text for references). The domains of β-catenin required for this repression (“–“) have not been determined. (E) In the study discussed here, Blauwkamp et al. discovered that recognition by Drosophila dTCF of an alternative sequence, AGAWAW, directs β-catenin to repress gene transcription (14). The structural features of TCF and the DNA site are unknown, as are the identities of the cooperating proteins that are likely to be involved. Although β-catenin is required to mediate gene repression, the N- and C-terminal activation domains are not necessary, which suggests that the repressor-acting domain is distinct.

The first pattern of gene expression linked to Wnts was the activated transcription of developmental genes such as en, distal-less, and labial in Drosophila (2). Soon after, target gene activation was identified in other developmental systems, and the collated list of activated genes has greatly expanded (for a current list, see R. Nusse’s excellent Wnt Web page: There are additional reasons for thinking of LEF/TCF:β-catenin complexes as activators. In the absence of Wnt signaling, LEF and the TCFs repress transcription by recruiting corepressors such as Groucho (Fig. 1B). This repression is opposed by stabilized β-catenin, which displaces Groucho and switches the activity of the complex from repression to activation (Fig. 1C).

The generation and wide dissemination of a Wnt reporter plasmid called TOPflash has also focused attention on transcriptional activation. Designed by van de Wetering and Clevers, this reporter plasmid has three multimerized LEF/TCF sites upstream of a minimal, low-activity promoter (3). TOPflash and its many second- and third-generation variants are sensitive indicators of activation because Wnt signaling triggers the assembly of LEF/TCF:β-catenin complexes on the multimerized sites of the reporter, thus detecting activities sometimes up to several-thousand-fold above baseline. Likewise, fusion of β-catenin to a heterologous DNA binding domain such as Gal4 leads to transcriptional activation of a reporter that contains multimerized Gal4 sites (3). The generous sharing of these reporters has been enormously useful and has advanced the field; however, as the molecular details of β-catenin action are more deeply probed, it is clear that the readouts of these reporters are monotonic. They do not capture the full spectrum of LEF/TCF:β-catenin activities, and, importantly for this discussion, they cannot be repressed below their already low basal activity state (4, 5).

Buried among the many studies of β-catenin–mediated gene activation is a handful of reports that suggest that Wnt directs LEF/TCF: β-catenin complexes to directly repress the transcription of target genes. The earliest example of negative actions by Wnt signaling came from studies of the Drosophila UbxB enhancer, whose activity is increased by weak Wingless (Wg, the Drosophila homolog of Wnt) signals but suppressed by strong Wg signals (6, 7). Other examples of Wnt-mediated repression include bona fide direct repression of Dpp transcription by dTCF:armadillo:Brinker in the Drosophila leg imaginal disc (8), LEF-1:β-catenin–mediated repression of the osteocalcin promoter in developing mouse chondrocytes (9), repression of the E-cadherin promoter in mouse epithelial stem cells (10), and repression of the p16INK4a promoter in melanoma (11). Inhibition of the transcription of these four genes requires the direct recruitment of β-catenin by either LEF or one of the TCFs to DNA. Other genes whose transcription is directly inhibited by Wnt signals, the RANKL promoter in osteoblasts and the antimetastasis KAI1 gene in cancer, may or may not involve both β-catenin and either LEF or TCF (12, 13). The mechanism by which β-catenin represses transcription is not understood, but in each of the four solidly documented cases LEF or one of the TCFs recruits β-catenin to a “classic” WRE (CTTTGWWS) (Fig. 1D). Here is where Blauwkamp, Chang, and Cadigan have added a new twist to the story of repression: A new sequence element appears to direct β-catenin–dependent repression (14).

A New Mode of Wnt-Mediated Gene Repression

In a microarray study of Wg target gene regulation in Drosophila, Blauwkamp et al. observed that, in addition to the expected activation of target genes, transcription of a small cohort of genes was strongly suppressed. Choosing one of these genes, Ugt36Bc, which encodes a component of the extracellular matrix, the authors asked whether repression was due to a decrease in transcription and, if so, whether the decrease was mediated directly by the LEF/TCF ortholog in Drosophila, dTCF, and armadillo. The answer to both of these questions was “yes.” Ubt36Bc transcription was repressed by Wg signaling, which required the direct binding and action of dTCF:armadillo complexes and not the induced expression of a gene encoding a repressor protein. Standard mapping studies combined with chromatin immunoprecipitation assays located a repressor region upstream of the Ubt36Bc transcription start site. This region was occupied by dTCF:armadillo complexes when expression of Ubt36Bc was repressed, and loss of armadillo or dTCF eliminated Wnt-mediated repression. That there is no classic WRE motif in the repressor region was no doubt perplexing, but deoxyribonuclease I (DNAase I) footprinting with recombinant dTCF protein showed that three sequences in the repressor region were protected. Alignment of these sequences revealed a DNA sequence motif containing AGAWAW, a notably short and degenerate sequence (Fig. 1E). Mutation of AGAWAW destroyed repressor action, but, even more striking, mutation to a traditional TCF-binding site (WRE: CTTTGWWS) converted the region into a transcriptional enhancer. Blauwkamp and Cadigan propose that AGAWAW sequences specify that dTCF:armadillo complexes act as transcriptional repressors rather than as activators. This model has its basis in only three sequences from one regulatory region and could certainly benefit from discovery of other “repressor” sites for validation or correction. Nevertheless, there are features of LEF/TCFs that make the model entirely plausible: Wnt signaling relies on famously flexible, context-dependent LEF/TCFs to access its target genes.

Flexible Features of LEF/TCF DNA Binding

The consensus DNA sequence for binding of LEF/TCFs has been defined through electrophoretic mobility shift assays and DNAse I footprinting of activating regions of target genes and through LEF/TCF enrichment of optimal binding sequences from large, random sequence libraries (3, 1517). These studies have repeatedly enriched for WRE sequences (CTTTGWWS) and have never uncovered AGAWAW, leading one to conclude that the element identified in Blauwkamp’s study is a weak binding site. Indeed, when the AGAWAW element alone was multimerized in a reporter plasmid, neither positive nor negative activity was detected. Stable, productive binding of dTCF to AGAWAW must be unconventional and must rely on context, a viable possibility given what is known about the DNA binding properties of LEF/TCFs.

LEF/TCFs bind to DNA through a high mobility group (HMG) box, a specific class of nucleic acid interaction domain, that is well known for its ability to bind to distorted DNA structures such as four-way junction DNA and other stem loop structures. Most of the interaction between the HMG box and these structures is electrostatic and sequence-independent (18). However, although the LEF/TCF subclass can bind to these distorted DNA structures very well, they also specifically bind to WRE sequences (CTTTGWWS) through van der Waals (nonelectrostatic) interactions (18). In so doing, LEF/TCFs use electrostatic and nonelectrostatic interactions to create their own high-affinity distortions of CTTTGWWS (Fig. 1). All LEF/TCF orthologs bend this sequence 90° to 130°, a right-angle bend to a near-complete foldback of the helix (19, 20). Bending the DNA creates specific three-dimensional architectures that facilitate synergistic interactions with proteins bound nearby, a β-catenin–independent function that was the first activity discovered for LEF/TCFs (Fig. 1A) (21, 22). Thus, one could envision that, in a reversal of roles, the degenerate AGAWAW sequence is distorted by a specific context of proteins binding nearby, making the nucleic structure more attractive for electrostatic-based binding even though the primary sequence may not be well conserved.

There is another feature of LEF/TCFs that makes unconventional but stable binding to DNA feasible. The C-terminal tails of these proteins contain a second DNA binding domain, which was discovered in an alternatively spliced E isoform of mammalian TCF-1 and TCF-4 but which is present in dTCF (17). This domain is called the C-clamp [formerly known as CRARF for the first five conserved amino acids (Cys-Arg-Ala-Arg-Phe) of the C-terminal tail] and is a 30–amino acid, cysteine-rich region next to the HMG box (3). The C-clamp exhibits moderate sequence specificity and affinity for a GC-rich element, but it also stabilizes the binding of the HMG box to other suboptimal sequences by providing additional points of contact with DNA. The interaction of dTCF with AGAWAW may require both an interaction with nearby proteins and the binding of the C-clamp domain to DNA. Whatever the mechanism, AGAWAW is unlikely to be an autonomous sequence motif.

How does AGAWAW specify repression and not activation of transcription? Again, context is important, partly because Blauwkamp et al. found that simple multimerization of the sequence was not repressive, partly because this sequence mediated activation in the absence of Wnt signaling, and partly because the region of armadillo involved in Wnt-directed repression was distinct from its clearly delimited activation domains (Fig. 1C versus Fig. 1E). Either coactivator complexes recruited by the activation domains of armadillo are present (but actively suppressed or converted in their action), or the AGAWAW context specifies that a different set of coregulators is assembled. Mutation or deletion of the N- and C-terminal activation domains in armadillo did not create a better repressor, so the latter model is slightly more favored.


One important implication of the Blauwkamp study is that attention to alternative Wnt outcomes such as gene repression leads to new discoveries of LEF/TCF:β-catenin actions. These modes of action could be important in diseases such as cancer, which often has aberrantly active Wnt signaling, and in stages of development in which morphogenic Wnt actions range from self-renewal of stem cells to terminal differentiation of progenitor cells. Indeed, the repressing actions of TCF-3 are an integral component of the core transcription circuitry in embryonic stem cells (2325). It seems worth revisiting the initial hypothesis that this action is β-catenin–independent because β-catenin can be coimmunoprecipitated with TCF-3 and Oct-4 from cross-linked chromatin extracts (23). In the case of cancer, direct suppression of the genes encoding the cell adhesion molecule E-cadherin and the cell cycle inhibitor p16ink4a and repression of the genes encoding RANKL and KAI1 are good examples that underscore how Wnt-mediated repression might be as relevant to tumor growth as is Wnt-activated expression of c-myc. Numerous groups have performed small-molecule screens for cancer agents with derivatives of TOPflash, but so far no small-molecule inhibitor has found its way into the clinic. Perhaps the repression mode of Wnt signaling has a special Achilles’ heel that can be better exploited as a target?

There is growing appreciation that simultaneous activation and repression of gene expression is a feature of multiple developmental signaling pathways (26). One simple explanation for why a pathway might direct one complex to carry out two opposing modes of gene regulation is that the fate or behavior of a cell is driven by modifying a set of genes in aggregate; some genes must be turned on while others are turned off. It would seem more efficient to change a cell in a single step rather than wait for indirect repression to occur. Another explanation is that Wnts are only one of the myriad signals bombarding cells at any one moment during development. Crosstalk between Wnt signals and other pathways sets up daedal patterns of gene expression. If, as the TOPflash-like reporters illustrate, the core activity of LEF/TCF:β-catenin complexes is to activate transcription, then Wnt signals might use this core activity either alone or sometimes in synergy with other pathways to turn on signature target genes. At the same time, crosstalk between Wnt signaling and other developmental pathways may set up a specific regulatory context at a separate set of target genes to convert the LEF/TCF:β-catenin complex into a transcriptional repressor. If this is correct, the exploration of these instances of repression should yield valuable mechanistic insight into Wnt crosstalk with other pathways.


View Abstract

Navigate This Article