Interactions between SOX factors and Wnt/β-catenin signaling in development and disease

Abstract

The SOX family of transcription factors have emerged as modulators of canonical Wnt/β-catenin signaling in diverse development and disease contexts. There are over twenty SOX proteins encoded in the vertebrate genome and recent evidence suggests that many of these can physically interact with β-catenin and modulate the transcription of Wnt-target genes. The precise mechanisms by which SOX proteins regulate β-catenin/TCF activity are still being resolved and there is evidence to support a number of models including; protein-protein interactions, the binding of SOX factors to Wnt-target gene promoters, the recruitment of co-repressors or co-activators, modulation of protein stability and nuclear translocation. In some contexts Wnt signaling also regulates SOX expression resulting in feedback regulatory loops that fine tune cellular responses to β-catenin/TCF activity. In this review we summarize the examples of Sox-Wnt interactions and examine the underlying mechanisms of this potentially wide spread and underappreciated mode of Wnt-regulation.

Introduction

Secreted Wnt growth factors are conserved in all metazoans and regulate many aspects of development and disease. The canonical Wnt signaling pathway activates a nuclear complex of β-catenin and TCF/LEF (T cell factor/lymphoid enhancer factors) transcription factors, which regulate the transcription of a vast array of different target genes in diverse biological contexts including, embryonic patterning, tissue homeostasis and stem cell maintenance (Logan and Nusse, 2004; Clevers, 2006). Wnt signaling must be tightly controlled because inappropriately elevated β-catenin/TCF activity leads to cancer in many tissues. Although there are approximately nineteen Wnt ligands and ten Frizzled receptors in vertebrates, the transcriptional responses to canonical signaling are mediated by just one β-catenin and four TCF factors (TCF1, TCF3, TCF4 and LEF1), raising the question of how distinct Wnt-target genes are activated in different cell types. One explanation is that β-catenin and TCFs interact with other transcription factors to modulate their activity and facilitate target gene selection (Arce et al., 2006; Gordon and Nusse, 2006). The identity of these factors and determining how they regulate β-catenin/TCF activity has become the focus of intense interest.

Over the last ten years, the SOX family of transcription factors has emerged as such nuclear regulators of β-catenin/TCF activity. They were first demonstrated to regulate canonical Wnt signaling when SOX17 and SOX3 were found to physically interact with β-catenin and antagonize its activity in Xenopus embryos and tissue culture assays (Zorn et al., 1999). Since then a growing number of SOX proteins have been reported to interact with β-catenin and TCFs in a variety of biological contexts, suggesting that this may be a common means of modulating transcriptional responses to Wnt signaling. In some contexts Wnt signaling also regulates Sox gene expression, and an emerging theme is that SOX factors act as feedback regulators that fine tune Wnt signaling activity. Here we review Sox-Wnt interactions in different biological settings, comparing the underlying molecular mechanisms, and highlight some future directions to be explored.

Overview of Canonical Wnt/ β-catenin pathway

Wnt growth factors can stimulate several distinct intracellular transduction pathways. Non-canonical Wnt signaling refers to a collection of pathways that activate small GTPases and regulate the actin cytoskeleton, whereas the so-called “canonical” Wnt pathway, which is the focus of this review, is mediated by the protein β-catenin (Cadigan and Liu, 2006; Semenov et al., 2007). Most of the β-catenin in cells is associated with adherens junctions and is a structural component of the cadherin-associated cytoskeleton. However a small dynamic pool of β-catenin shuttles between the cytoplasm and the nucleus to transduce canonical Wnt signals (Kimelman and Xu, 2006).

In the absence of Wnt signaling cytosolic β-catenin is captured by a large group of proteins known as the destruction complex, including Axin, APC and the kinase GSK3, which phosphorylates serine residues in the N-terminus of β-catenin. Phosphorylated β-catenin is then recognized by the E3 ubiquitin ligase β-TrCP and targeted for proteosome degradation (Figure 1) (Kimelman and Xu, 2006). When Wnt signaling is inactive, TCF/LEF HMG domain proteins are bound to Wnt-target gene promoters in a complex with transcriptional co-repressors such as Tle/Groucho to keep target genes silent (Roose et al., 1998; Brannon et al., 1999; Hoppler and Kavanagh, 2007). Binding of Wnt ligands to cell surface receptors leads to the inactivation of the destruction complex. As a result cytosolic β-catenin accumulates and translocates into the nucleus, where it interacts with TCF/LEF transcription factors(Logan and Nusse, 2004; Clevers, 2006; Xu and Kimelman, 2007). Upon binding to TCF, β-catenin is thought to displace co-repressors, and recruit co-activators such as the histone acetyltransferase CBP/p300, to stimulate transcription (Figure 1) (Daniels and Weis, 2002; Daniels and Weis, 2005; Arce et al., 2006; Stadeli et al., 2006).

Figure 1. Summary of the canonical Wnt signaling pathway.

In the absence of a Wnt signal the destruction complex including Axin, APC and the kinase GSK3, phosphorylates (P) β-catenin. Phosphorylated β-catenin is recognized by the E3 ubiquitin (Ub) ligase β-TrCP and targeted for proteosome degradation. Wnt signaling inactivates the destruction complex, as a result β-catenin translocates into the nucleus, where it interacts with TCF/LEF transcription factors and co-activators such as CBP/p300, to stimulate transcription. In the absence of Wnt signaling, DNA-bound TCFs interact with transcriptional co-repressors such as, Tle/Groucho to keep target genes turned off.

SOX Transcription Factors

Closely related to TCF proteins, the SOX transcription factors are also DNA-binding HMG domain proteins. There are approximately 20 different Sox genes in the vertebrate genome that are classified into eight subfamilies (SoxA-H) according to sequence homologies in their HMG domains (Bowles et al., 2000). SOX factors are found in all metazoans and regulate many of the same processes as Wnt signaling including tissue specification, organ development, stem cell homeostasis and cancer (Wilson and Koopman, 2002; Dong et al., 2004; Kiefer, 2007). In vitro, SOX and TCF proteins bind to similar but apparently distinct variants of the core DNA sequences 5′-(A/T)(A/T)CAA(A/T)G-3′ with variable flanking bases (Harley et al., 1994; Mertin et al., 1999; van Beest et al., 2000). In general SOX proteins have low DNA-binding affinity on their own and they acquire high affinity and specificity by interacting with other transcription factors such as POU, homeodomain, zinc finger and bHLH domain proteins (Kamachi et al., 2000; Wilson and Koopman, 2002; Wiebe et al., 2003; Wissmuller et al., 2006). SOX proteins can be either transcriptional activators or repressors depending on the cellular context and their associated interacting proteins. Upon DNA-binding the HMG domain induces DNA bending and it is thought that SOX proteins act in part by remodeling the local chromatin architecture(Kamachi et al., 2000; Scaffidi and Bianchi, 2001; Wilson and Koopman, 2002).

SOX-Wnt interactions in Development and Cancer

While it is clear that SOX proteins have many Wnt-independent functions, emerging evidence suggests that they have widespread and underappreciated roles in modulating Wnt signaling in development and disease (Table 1). In most reports SOX proteins repress Wnt transcriptional responses, however some Sox proteins appear to enhance Wnt-target gene expression (Figure 2). Finally the expression of some Sox genes are regulated by Wnts, and together these SOX-Wnt interactions appear to fine-tune the spatial and temporal activity of canonical Wnt signaling.

Table 1.

Summary of Sox-Wnt interactions

SOX	Effect on TOP:flash	Protein β-cat	Binding TCF	Context / Mechanism	Reference
SOXA
mSRY	Repress			Represses Wnt in sex determination - activates transcription of a Wnt-antagonist	Tamashiro et. al. 2008
hSRY	Repress	Yes		Represses Wnt in sex determination - binds β-cat, nuclear translocation	Bernard et. al. 2008
SOXB
mSOX1	Repress	Yes		Represses Wnt in neural differentiation - DNA and β-catenin binding	Kan et. al. 2004
mSOX2	Repress	Yes		Represses Wnt in bone formation - binds β-cat	Mansukhani et. al. 2005
hSOX2		Yes		Co-operates with Wnt in breast cancer - binds β-cat, promotes CyclinD1 transcription	Chen et. al. 2008
xSOX3	Repress	Yes		Represses Wnt in Xenopus axis formation - binds DNA and represses xnr5 promoter	Zhang et. al. 2003, Zorn et. al. 1999
spSOXB1	Repress	Yes		Represses Wnt in sea urchin development - mutual SoxB1, β-cat degradation	Kenny et. al. 2003, Angerer et. al. 2005
SoxNeuro	Repress	No	No	Represses Wnt in Drosophila cuticle patterning - acts at the level of TCF/Groucho	Chao et. al. 2007 Overton et. al. 2007
Dichaete				Represses Wnt in Drosophila cuticle patterning	Overton et. al. 2007
SoxC
hSOX4	Enhance	Yes	Yes	Enhances Wnt in colon cancer cells - binds β-cat and TCF, stabilize β-cat	Sinner et. al. 2007
mSOX11	Enhance			Enhances Wnt in 293T cells	Sinner et. al. 2007
SOXD
mSOX5	Enhance			Enhances Wnt in 293T cells	Sinner et. al. 2007
mSOX6	Repress	Yes		Represses Wnt in β-cell homoestasis - binds β-cat, recruits HDAC1	Iguchi et. al. 2007
hSOX13	Repress	No	Yes	Represses Wnt in lymphocyte development - binds TCF	Melichar et. al. 2007
SOXE
m/hSOX9	Repress	No	Yes	Represses Wnt in intestine and CRC - promotes expression of Tle/Groucho to repress Cyclin-D1 and c-Myc	Bastide et. al. 2007
	Repress	Yes		Represses Wnt in chondrogenesis - binds/represses β-cat, nuclear translocation - mutual SOX9 and β-cat degradation	Akiyama et. al. 2004, Topol et. al. 2009
mSOX10	Repress			Represses Wnt in 293T cells	Sinner et. al. 2007
SOXF
hSOX7	Repress	Yes		Represses Wnt in colon cancer cells - binds β-cat, promotes β-cat degradation	Takash et. al. 2001, Guo et. al. 2008, Zhang Y. et. al. 2008
xSOX17	Repress	Yes		Represses Wnt in Xenopus axis formation -Co-operates with β-cat to activate Sox-targets	Zorn et. al. 1999, Sinner et. al. 2004
hSOX17	Repress	Yes	Yes	Represses Wnt in colon cancer cells - binds β-cat/TCF, promotes β-cat degradation	Sinner et. al. 2007, Zhang W. et. al. 2008
dSOXF				Represses Wnt in wing formation - represses Wg transcription	Dichtel-Danjoy et. al. 2009

Figure 2. Sox proteins modulate β-catenin/TCF transcriptional activity.

A) Some SOX proteins, such as SOX17 repress β-catenin\TCF, while others like SOX4 enhance their transcriptional activity. B) An example of a TOP:flash transcriptional reporter assay. A TOP:flash plasmid containing optimized TCF binding sites upstream of a basal promoter driving luciferase gene expression is transfected into tissue culture cells. Co-transfection of β-catenin (+) stimulates transcription as measured by luciferase activity. Co-transfection of the indicated SOX factors along with β-catenin can either repress or enhance β-catenin stimulated TOP:flash activity. On their own SOX proteins have no effect on TOP:flash (not shown). The subfamily that each SOX belongs to is indicated on top of the histogram. Modified with permission from Sinner et al., 2007 MCB.

Xenopusgerm layer patterning

A functional interaction between SOX factors and Wnt signaling was first revealed in Xenopus embryos (Zorn et al., 1999). A maternal Wnt/β-catenin pathway is active on one side of the Xenopus blastula and this stimulates transcription of organizer genes, which define the future dorsal-anterior body axis (Heasman, 2006). Microinjection of molecules that inhibit this pathway suppress axis formation resulting in headless embryos. This is a common method to identify and characterize Wnt pathway components and it was in such a functional screen that over expression of Xenopus SOX17 and SOX3 were found to inhibit β-catenin activity.

During normal Xenopus development Sox17 and Sox3 are in the presumptive endoderm and ectoderm respectively (Hudson et al., 1997; Zhang et al., 2003), and on the dorsal side of the blastula their expression overlaps with active β-catenin. Loss of function studies found that in addition to impaired endoderm or ectoderm formation, the antisense depletion of SOX17 or SOX3 also resulted in increased Wnt-dependent gene expression (Zhang et al., 2003; Sinner et al., 2004; Sinner et al., 2006). This suggests that SOX17 and SOX3 coordinate Xenopus germ layer formation and patterning in part by restricting β-catenin/TCF stimulated transcription. Interestingly, the interaction between SOX17 and β-catenin also appears to enhance the ability to SOX17 to activate its target genes (Sinner et al., 2004).

Cell culture and reporter assays

Transcriptional reporter assays in tissue culture cells are commonly used to measure Wnt-stimulated transcription and to assay the impact of SOX proteins on β-catenin/TCF activity. In these assays a plasmid reporter construct known as “TOP:flash” which contains multiple copies of a TCF DNA-binding site upstream of a minimal promoter driving luciferase expression is transfected into fibroblasts (e.g. HEK 293T cells). When the Wnt-pathway is activated in these cells, for example by over expressing β-catenin, endogenous TCF factors activate transcription of the reporter, which can be quantitatively measure by luciferase activity (Korinek et al., 1997). In this assay, co-transfection of Xenopus SOX17 or SOX3 suppressed the ability of β-catenin to stimulate transcription of the TOP:flash reporter (Zorn et al., 1999).

Subsequently many other SOX proteins from a variety of species have been shown to repress β-catenin stimulated TOP:flash transcription in tissue culture (Figure 2B, Table 1) (Takash et al., 2001; Akiyama et al., 2004; Mansukhani et al., 2005; Sinner et al., 2007). Interestingly some SOX proteins, such as mammalian SOX4 have the opposite effect and enhance β-catenin stimulated transcription (Sinner et al., 2007). However, it is important to point out that these over expression experiments do not always correlate with the activity of SOX factors in vivo; suggesting that the tissue culture does not always reflect cell-specific contexts, or alternatively the cell culture experiments only measure one aspect of the SOX-Wnt interaction.

Genetic interactions in Drosophila

Genetic interactions between Wnt signaling and Sox genes have also been described in fruit fly. Mutations in the Drosophila SoxNeuro and Dichaete genes, members of the SoxB family, result in embryonic segmentation defects, reminiscent of elevated canonical Wnt signaling (Nambu and Nambu, 1996; Russell et al., 1996; Buescher et al., 2002; Overton et al., 2002). Genetic analyses confirmed that SoxNeuro and Dichaete pattern the fly embryo by spatially restricting Wnt signaling activity to discrete domains in the developing epidermis (Chao et al., 2007; Overton et al., 2007). Similarly Drosophila SoxF (dSox15) restricts Wnt signaling to a narrow stripe of cells in the developing wing imaginal disc, and in SoxF mutants inappropriately elevated Wnt signaling causes an over proliferation of the wing disc epithelium (Dichtel-Danjoy et al., 2009).

SoxB and vertebrate neural development

Canonical Wnt signaling and the SoxB1 genes (Sox1-3) also play important roles in vertebrate neural development; helping to sustain neural progenitors in an undifferentiated, self renewing state (Kamachi et al., 1998; Pevny and Placzek, 2005; Wang et al., 2006; Kiefer, 2007). In some contexts repression of Wnt signaling allows neural progenitors to exit the cell cycle and differentiate (Chenn and Walsh, 2002; Ille and Sommer, 2005). SOX1 may be involved in this process in mice (Pevny and Placzek, 2005). When over expressed in neural cell lines SOX1 suppressed the pro-proliferative activity of β-catenin, promoted cell cycle exit and initiated neuronal differentiation. Consistent with this SOX1 was able to bind β-catenin and inhibit its transcriptional activity in fibroblasts (Kan et al., 2004).

T-cell development

Genetic analyses in mice indicate that the segregation between the αβ and γδ T-cell lymphocyte lineage is controlled by the antagonistic interaction between TCF1 and SOX13. TCF1/β-catenin activity promotes the expansion of αβ cells, and transgenic over expression of Sox13 inhibited αβ T-cell proliferation and promoted γδ T-cell development (Melichar et al., 2007). In addition Sox13 over expression repressed endogenous TCF1 target gene transcription in cultured T-cells and inhibited β-catenin stimulation of a TOP:flash reporter in fibroblasts.

Pancreas

There is evidence in adult mice that SOX6 and canonical Wnt signaling also interact to regulate pancreatic β-cell homeostasis. In mice with obese-related insulin resistance, the pancreatic islets compensate by increasing insulin secretion and expanding the β-cell population, and this is correlated with adown regulation in Sox6 expression (Iguchi et al., 2007). Experiments in insulinoma cell lines suggest that SOX6 might normally attenuate β-cell proliferation by suppressing β-catenin stimulated transcription of the direct Wnt-target gene Cyclin-D1, a key component of the cell cycle(Iguchi et al., 2007).

Cartilage development

Sox9 regulates cartilage specific gene expression at multiple steps in mammalian chondrogenesis, and part of its function appears to involve Wnt repression (Lefebvre et al., 1997; Ng et al., 1997; Akiyama et al., 2002; Furumatsu et al., 2005). In humans Sox9 haploinsufficiency cause campomelic dysplasia, a disease characterized by extreme cartilage and bone malformation, as well as sex reversal (Wagner et al., 1994; Bi et al., 2001). Genetic studies in mice have shown that stabilization of β-catenin or deletion of Sox9 both inhibit chondrocyte differentiation (Bi et al., 1999; Akiyama et al., 2002; Chun et al., 2008), whereas transgenic SOX9 over expression or conditional deletion of β-catenin result in down regulation of Cyclin-D1, decreased chondrocyte proliferation and delayed bone formation (Akiyama et al., 2004; Kiefer, 2007). Consistent with these genetic studies SOX9 and β-catenin inhibited each other’s transcriptional activity in tissue culture assays (Akiyama et al., 2004). Thus the balance between antagonistic SOX9 and β-catenin activities appears to govern the expansion and timely differentiation of chondrocytes.

Sex determination

Sox9 also regulates mammalian sex determination along with the Y chromosome specific Sry, the founding member of the Sox gene family. Sry or Sox9 are both sufficient to initiate testis development in XX individuals and mutations in human Sry or Sox9 result in sex reversal, with female development in XY individuals (Koopman et al., 1991; Clarkson and Harley, 2002; Polanco and Koopman, 2007). Recent evidence suggests that SRY and SOX9 function in part by suppressing canonical Wnt signaling (Bernard et al., 2008; Tamashiro et al., 2008), which otherwise promotes ovarian fate and blocks testis development (Chassot et al., 2008; Maatouk et al., 2008; Tomizuka et al., 2008).

Intestinal homeostasis

SOX-Wnt interactions also regulate homeostasis of the adult intestinal epithelium, which is constantly renewed by resident stem cells. Wnt signaling maintains the stem cell population and promotes the proliferation of a transient amplifying population (Clevers, 2006). Attenuation of Wnt signaling is then required for these cells to exit the cell cycle and differentiate intoenterocytes, goblet cells or Paneth cells. A number of Sox genes are expressed in the intestinal epithelium (Blache et al., 2004; Sinner et al., 2007) and conditional deletion of Sox9 from the mouse intestine results in a failure of Paneth cell differentiation, an up regulation of the Wnt-target genes Cyclin-D1 and c-Myc, and an increase in epithelial proliferation causing intestinal hyperplasia (Bastide et al., 2007; Mori-Akiyama et al., 2007). This suggests that SOX9 suppresses Wnt signaling and allows cells to stop proliferating and differentiate. This balance between proliferation and differentiation must be tightly regulated because inappropriately elevated Wnt signaling can lead to intestinal cancer (Clevers, 2006).

Wnt-Sox interactions in cancer

Mutations in Wnt pathway components such as APC and β-catenin that lead to constitutive signaling are causally linked to the development of many types of cancer (Logan and Nusse, 2004; Clevers, 2006). Recent studies have demonstrated that Sox gene expression is also deregulated in a wide variety of human cancers (Dong et al., 2004; Reichling et al., 2005; de Bont et al., 2008; Haram et al., 2008), and there is evidence that SOX factors impact tumorogenesis by modulating β-catenin/TCF activity and the expression of oncogenic Wnt-target genes such as Cyclin-D1 and c-Myc.

Some Sox genes, such as Sox7 and Sox17 are epigenetically silenced in many human cancers and they appear to act as tumor suppressors (Zhang et al., 2008; Fu et al., 2009; Zhang et al., 2009). For example, over expression of SOX7 or SOX17 in human colon cancer cell lines was found to suppress the hyperactive β-catenin activity in those cells as well as reduce Cyclin-D1 expression and repress proliferation (Sinner et al., 2007; Guo et al., 2008; Zhang et al., 2008; Zhang et al., 2009).

Other Sox genes however, such as Sox4 are frequently over expressed or genetically amplified in tumors (Reichling et al., 2005; Aaboe et al., 2006; de Bont et al., 2008; Andersen et al., 2009; Medina et al., 2009; Scharer et al., 2009) and appear to act as oncogenes, in part by enhancing β-catenin/TCF activity. The stable transfection of SOX4 was found to transform prostate cells, (Liu et al., 2006), whereas antisense depletion of SOX4 from prostate or colon cancer cell lines inhibited Cyclin-D1 expression and reduced proliferation (Liu et al., 2006; Sinner et al., 2007). Finally there are cases where the same Sox gene behaves differently in different cancers. Sox2 for example is frequently over expressed in aggressive human breast carcinomas where it promotes β-catenin stimulated proliferation (Chen et al., 2008), whereas in gastric cancer, Sox2 is often down regulated and when over expressed in those cells represses Cyclin-D1 expression and proliferation (Otsubo et al., 2008).

Mechanisms of SOX-Wnt interactions

The precise mechanism by which SOX proteins regulate β-catenin/TCF activity is still being resolved and various studies suggest a number of different mechanisms including, protein-protein interactions, the binding of SOX factors to Wnt-target gene promoters, the recruitment of co-repressors or co-activators and the regulation of protein stability. It appears that several of these mechanisms may operate together and some SOX proteins appear to employ different mechanisms in different cellular contexts.

SOX-β-catenin protein interactions

Many SOX proteins have been shown to physically interact with β-catenin (Table 1) and this appears to be critical for their ability to regulate Wnt signaling (Zorn et al., 1999; Takash et al., 2001; Akiyama et al., 2004; Kan et al., 2004; Mansukhani et al., 2005; Iguchi et al., 2007; Sinner et al., 2007). In most cases these interactions have only been demonstrated in vitro or from co-immunoprecipitations of over-expressed proteins in tissue culture. However, endogenous SOX1 and SOX17 have been shown to form a complex with β-catenin in extracts from mouse embryos and human colorectal cancer cell lines respectively(Kan et al., 2004; Sinner et al., 2007).

The central region of β-catenin consists of 12 armadillo repeats that together form a helical groove that mediates the interaction with over 20 different β-catenin-binding partners(Xu and Kimelman, 2007). In vitro studies indicate that SOX17, SOX9 and SOX6 can directly bind β-catenin in a region of the armadillo repeats that overlaps with the site where TCF proteins bind (Figure 3) (Zorn et al., 1999; Akiyama et al., 2004; Iguchi et al., 2007; Sinner et al., 2007). Structure function experiments have also mapped the β-catenin-binding regions in various SOX proteins (Figure 3). Human SRY binds β-catenin through a N-terminal domain (Bernard et al.), SOX6 interacts via a centrally located leucine zipper (LZ/Q) element (Iguchi et al.), and mammalian SOX7, SOX9 and SOX17 all bind β-catenin via their C-terminal regions (Zorn et al., 1999; Takash et al., 2001; Akiyama et al., 2004; Sinner et al., 2007). In the case of SOX17 sequences just carboxyl to the HMG box also appear to be important, suggesting that β-catenin contacts several regions of the protein (Zorn et al., 1999).

Figure 3. Summary of SOX, TCF and β-catenin interaction domains.

(A) The schematic shows the regions of various SOX proteins that have been shown to bind and/or inhibit β-catenin/TCF. The nuclear import signals (NLS), nuclear export signals (NES) and DNA-binding HMG domains are indicated. (B) β-catenin as a central domain of 12 armadillo repeats, which can bind many proteins including TCF, SOX, the inhibitor ICAT and E-cadherin. The N-terminal domain harbors Serine residues that are phosphorylated by GSK3, whereas the C-terminal region binds to p300, CPB (Xu and Kimelman, 2007). (C) TCF/LEF proteins indicating the β-catenin, Tle and SOX binding domains. Nemo-like kinase (NLK) phosphorylates specific residues in N-terminus of LEF1 and TCF4.

Since the amino acid sequence of different SOX proteins is rather divergent outside of the HMG domain it is possible that β-catenin recognizes a conserved tertiary structure. A candidate motif was identified in the C-terminal β-catenin binding regions of SOX17 and SOX7. They contain a short ~9 amino acid motif (DXXEFDQYL) conserved in many SOX factors including all SOXF family members. Mutations of this motif indicate that it is required for SOX7 and SOX17 to bind β-catenin and repress its activity in tissue culture assays (Sinner et al., 2004; Sinner et al., 2007; Guo et al., 2008). Interestingly this motif has similarity to a loosely conserved armadillo-binding sequence found in many β-catenin-binding proteins including TCFs and E-cadherin (Daniels and Weis 2002, Xu and Kimelman 2007).

SOX-TCF protein interactions

In vitro protein binding experiments have shown that mammalian SOX17 and SOX4 can also directly interact with TCF3, TCF4 and LEF1 via their respective HMG domain (Figure 3) (Sinner et al., 2007). This is consistent with studies showing that in addition to DNA-binding the SOX HMG domain can mediate interactions with many different transcription factors (Wiebe et al., 2003; Wissmuller et al., 2006). In T-cell development SOX13 appears to repress Wnt-targets not by binding to β-catenin but by interacting with TCF1 (Melichar et al., 2007). Interestingly SOX13 bound full length TCF1 from cell extracts, but not to an mRNA splicing variant of TCF1 that lacks the N-terminal β-catenin interaction domain. This suggests that SOX13 associates with the N-terminal of TCF1 and may inhibit Wnt signaling by excluding β-catenin.

Do SOX and TCF factors compete for β-catenin binding?

The observation that SOX and TCF proteins interact with overlapping armadillo repeats also supports the possibility that they might compete for β-catenin binding (Figure 4A). Indeed steric hindrance is an established means of regulating β-catenin-binding partners and is an important mechanism for inhibitory proteins such as ICAT to displace TCF/LEF from β-catenin (Xu and Kimelman, 2007). This competition model has been tested for mouse SOX9 and SOX17 with conflicting results. In vitro binding experiments found that SOX9 inhibited the association between TCF and β-catenin as predicted (Akiyama et al., 2004; Iguchi et al., 2007). Surprisingly, SOX17 was found to bind both β-catenin and TCF in a complex (Sinner et al., 2007). Detailed biochemical studies will be required to resolve these discrepancies and to determine whether competition or complex formation occurs in vivo.

Figure 4. Models of SOX-Wnt regulation.

(A) SOX and TCF may compete for a limited pool of β-catenin. This may occur on DNA or independent of DNA-binding. (B) SOX and TCF may compete for DNA-binding to the same sites in target gene promoters. (C) SOX might stabilize TCF/Tle repressor complexes, and/or exclude β-catenin and co-activators. This could occur though SOX DNA binding or by a direct interaction between SOX and TCF (not shown). SOX might recruit (D) transcriptional co-repressors, or (E) co-activators to Wnt-target gene promoters, either though DNA-binding or by an association with β-catenin/TCF. (F) SOX proteins might recruit post-translational modifying enzymes to β-catenin and/or TCF. The resulting phosphorylations (P), ubiquitination (Ub) or SUMOylation (S) could impact protein stability, activity and/or subcellular localization. (G) Some SOX proteins promote the degradation of β-catenin (and TCF). In the case of SOX9 this appears to involve nuclear import of the GSK3-containing destruction complex. Other studies suggest and alternative degradation pathway. (H) Sox and Wnt-signaling can functionally interact in negative feedback loops, (upper) or in mutually repressive associations.

Structure function studies aimed at determining which SOX protein domains are required to repress β-catenin/TCF activity have come to varying conclusions (Figure 3). For some SOX proteins the β-catenin interaction domain is the most critical(Kenny et al., 2003; Mansukhani et al., 2005; Iguchi et al., 2007; Guo et al., 2008). However in other studies the HMG domain appears to be essential (Zhang et al., 2003; Tamashiro et al., 2008; Zhang et al., 2009), while in some reports (SOX1 and SOX17) both the HMG domain and β-catenin binding motif are required (Kan et al., 2004; Sinner et al., 2007). These apparent contradictions may reflect distinct mechanisms in different cellular contexts. For example in colorectal cells SOX9 requires the HMG domain to repress Wnt signaling (Bastide et al., 2007), whereas in chondrogenesis SOX9 appears to requires the β-catenin interaction domain (Akiyama et al., 2004).

Sox-Wnt regulation by DNA-binding

Although tissue culture experiments indicate that some SOX proteins can repress β-catenin activity independently of the DNA-binding, other studies have found that in vivo an association with Wnt-target gene promoters is essential. For example, while DNA-binding was not required for Xenopus SOX3 to repress β-catenin stimulated transcription of the TOP:flash reporter, DNA-binding was essential to inhibit expression of the endogenous Wnt-target gene xnr5 in Xenopus embryos (Zhang et al., 2003). An analysis of the xnr5 promoter revealed adjacent SOX and TCF DNA-binding sites that were bound by SOX3 and TCF3 respectively in embryonic extracts. Repression of xnr5 required SOX3 binding to the SOX site (Zhang et al., 2003). Although it is unknown exactly how SOX3 represses xnr5 transcription it is possible that SOX3 inhibits TCF3 DNA-binding or prevents its association with β-catenin. Alternatively SOX3 and TCF might form a high affinity DNA-binding complex that stabilizes the Groucho co-repressor on the xnr5 promoter (Figure 4C).

Genetic epistasis experiments in Drosophila indicate that SoxNeuro also represses endogenous Wnt-target genes at that the level of TCF, consistent with a model where the SOX stabilizes TCF-Groucho complexes (Figure 4C)(Chao et al., 2007). This genetic data does not rule out the possibility that SOX factors might also bind to Wnt-target gene promoters at a distance from TCF sites and regulate transcription independent of any interaction with β-catenin/TCF factors. In addition there is evidence that SOX proteins can repress Wnt signaling indirectly by regulating the expression of Wnt pathway components. For example in the mammalian intestine, SOX9 appears to activate the transcription of Groucho/TLE co-repressors (Bastide et al., 2007), while in Drosophila SOXF interacts with a wing specific enhancer to repress the transcription of the Wg gene encoding the Wnt ligand (Dichtel-Danjoy et al., 2009).

As SOX and TCF proteins bind similar DNA sequences, SOX proteins might also suppress Wnt-induced transcription by competing with TCF for the same promoter sites (Figure 4B). In vitro DNA-binding studies with optimized SOX and TCF DNA-binding sites argue against this model, and conclude that SOX proteins bind very poorly if at all to an optimized TCF consensus sequence and vice versa (Harley et al., 1994; Zorn et al., 1999; Zhang et al., 2003; Akiyama et al., 2004). However endogenous promoters are often complex and several well characterized Wnt-target genes including Cyclin-D1 have multiple divergent TCF binding sites in their proximal promoters, some of which are also predicted to be SOX sites(Tetsu and McCormick, 1999).

Chromatin immunoprecipitation (ChIP) studies have shown that SOX6 in pancreatic cells and SOX2 in human breast cancer cells can both associate with regions of the Cyclin-D1 promoter that contain TCF-binding sites. Interestingly SOX2 enhances β-catenin stimulated Cyclin-D1 transcription (Chen et al., 2008), whereas SOX6 represses its expression (Iguchi et al., 2007). It is unclear whether SOX6 or SOX2 directly bind to the TCF DNA sites. Both SOX2 and SOX6 appear to require β-catenin to associate with the promoter and mediate their effects, supporting the idea that they interact indirectly by binding to the β-catenin/TCF complex (Figure 4C, D). However, mutations in one of the known functional TCF-binding sites in the Cyclin-D1 promoter diminished the ability of SOX2 to associate with the DNA and activate transcription. Moreover, depletion of endogenous TCF4 did not alter the ability of SOX2 and β-catenin to cooperatively activate Cyclin-D1 transcription, arguing that SOX2 and β-catenin might directly bind the promoter independently of TCF (Chen et al., 2008). In vivo foot printing experiments may be necessary to rigorously test this model.

Recruitment of transcriptional co-activators and co-repressors

Differential recruitment of transcriptional co-activators or co-repressors is one mechanism by which SOX factors can either enhance or repress Wnt-target gene transcription (Figure 4D, E). For example, SOX6 recruits the histone de-acetylase HDAC1 to β-catenin complexes on the Cyclin-D1 promoter suppressing its transcription (Iguchi et al., 2007). On the other hand the SOX2 and β-catenin interaction might enhance Cyclin-D1 transcription by cooperatively recruiting co-activators. Both β-catenin and some SOX proteins, such as SOX4 and SOX9, have been shown to directly interact with CBP/p300 (Tsuda et al., 2003; Furumatsu et al., 2005; Pan et al., 2009).

The recruitment of co-activators may also explain how nuclear β-catenin can enhance the ability of SOX17 factors to transactivate SOX-target genes in Xenopus (Sinner et al., 2004). This cooperativity required the conserved (DXXEFDQYL) β-catenin-binding motif and experiments with Xenopus SOX17 and mouse SOX18 have shown these same 9 amino acids can act as a bonified transactivation domain (Sinner et al., 2004; Sandholzer et al., 2007). One possibility is that by binding to β-catenin this motif facilitates the recruitment of β-catenin-CBP complexes to Sox17-target gene promoters. Similar charged transactivation motifs have been identified in other transcription factors including p53, VP16 and NFκB (Sandholzer et al., 2007), suggesting that this domain may mediate protein-protein interactions in many contexts.

SOX and β-catenin degradation and nuclear translocation

Another mechanism by which some SOX proteins repress Wnt signaling is by promoting proteosome-mediated β-catenin degradation (Figure 4G). In most studies this requires a physical interaction with β-catenin and can also result in SOX protein turnover. For example, when co-transfected into fibroblastsSOX9 and β-catenin are both degraded by a ubiquitin mediated proteosome pathway (Figure 4F, G) (Akiyama et al., 2004; Akiyama et al., 2005). Cre-mediated deletion of a floxed-Sox9 allele from primary mouse chondrocytes resulted in increased levels of endogenous β-catenin protein (Topol et al., 2009), suggesting that this mechanism regulates chondrogenesis in vivo.

A recent study in tissue culture concluded that SOX9 inhibited β-catenin/TCF dependent transcription and promoted β-catenin degradation by two separate mechanisms that depend on different domains of SOX9 (Topol et al., 2009). The N-terminus and HMG domain were required for β-catenin degradation, whereas the C-terminus was required to repress β-catenin stimulated transcription without affecting its stability (Figure 3). They found that SOX9 associated with components of the destruction complex including GSK3β, Axin and the ubiquitin ligase βTrCP, and translocated the complex to the nucleus where it phosphorylated and degraded nuclear β-catenin (Figure 4G) (β-catenin degradation is normally thought to occur in the cytoplasm). This novel degradation mechanism required the nuclear localization signaling (NLS) in SOX9’s HMG domain, which interacts with the nuclear pore complex (Preiss et al., 2001). Interestingly, most, if not all SOX proteins have both nuclear import and nuclear export sequences in their HMG domains (Figure 3), and regulated nuclear-cytoplasmic shuttling appears to be an important mode of regulation (Gasca et al., 2002; Rehberg et al., 2002; Smith and Koopman, 2004). For example, SOX9 is located in both the cytoplasm and nucleus during early chondrocyte development, but once they differentiate SOX9 is exclusively nuclear (Kiefer, 2007).

SOX9 and SRY also exhibit dynamic nuclear-cytoplasmic shuttling during sexual differentiation (Smith and Koopman, 2004). An analysis of Sry mutations found in human clinical patients with sex reversal found that SRY mutants with impaired nuclear import could not efficiently repress Wnt signaling in tissue culture experiments (Bernard et al., 2008), suggesting that nuclear degradation of β-catenin might important for sex determination in vivo.

SOX7 and SOX17 over expression can also stimulate β-catenin degradation in colon cancer cell lines, however this appears to involve a GSK3-independent mechanism, as they promote turnover of a stabilized mutant β-catenin (S37A mutation) that cannot be phosphorylated by GSK3 and therefore cannot be recognized by the destruction complex (Sinner et al., 2007; Guo et al., 2008). SOX17 also promoted TCF protein turnover in these experiments. Both the β-catenin interaction motif and the HMG domain were required to promote β-catenin/TCF degradation. However, the DNA-binding activity of the HMG domain was dispensable, suggesting that the HMG domain functions either by facilitating TCF-binding or by mediating nuclear translocation (Sinner et al., 2007). Surprisingly SOX4 has the opposite effect and appears to stabilize β-catenin. SiRNA depletion of endogenous SOX4 from colon carcinoma cells reduced the high β-catenin levels normally found in these cells (Sinner et al., 2007). This suggests that SOX4 stabilizes β-catenin by antagonizing the same pathway that SOX17 and SOX7 use to promote β-catenin degradation.

Post-translational modifications

Post-translational modifications are also likely regulate SOX-Wnt interactions (Figure 4F). For example, phosphorylation of SOX9 can influence its nuclear entry (Chikuda et al., 2004; Smith and Koopman, 2004), whereas phosphorylation of TCF by Casein kinase can disrupt β-catenin interactions, and phosphorylation by Nemo-like kinase (NLK) inhibits TCF DNA-binding and protein stability (Arce et al., 2006; Hoppler and Kavanagh, 2007). SOX11 can also associate with NLK (Hyodo-Miura et al., 2002)raising the possibility that SOX11 might recruit NLK to Wnt target gene promoters to phosphorylate TCF and release it from the DNA. SOX, β-catenin and TCF can also be modified by ubiquitin and SUMO (small ubiquitin-like modifier) moieties (Gill, 2005; Savare et al., 2005; Taylor and Labonne, 2005; Hattori et al., 2006; Pan et al., 2006), which can impact their stability, activity and sub-cellular localization (Arce et al., 2006; Hoppler and Kavanagh, 2007). For example the E3 ligase PIASy can SUMOylate both TCF4 and SOX9 enhancing their stability and/or activity, while SOX9 ubiquitination appears to promote its degradation (Taylor and Labonne, 2005; Arce et al., 2006; Hattori et al., 2006).

Sox-Wnt regulatory feedback loops

In addition to SOX proteins regulating Wnt activity, in many biological contexts Wnt signaling regulates Sox gene expression. An emerging theme from various studies is it that these reciprocal interactions result in regulatory feedback loops that fine-tune cellular responses to Wnt signaling (Figure 4H).

Both Drosophila and sea urchin SOXB proteins provide good examples of mutually repressive interactions. In sea urchin embryos nuclear β-catenin is localized to the vegetal cells and is required for mesendoderm specification, whereas SOXB1 is critical for the ectoderm (Kenny et al., 1999; Angerer et al., 2007). SOXB1 locally suppresses β-catenin/TCF activity, while β-catenin in turn inhibits SOXB1 expression by two separate mechanisms. First β-catenin/TCF activity represses SoxB1 transcription and second, nuclear β-catenin promotes the degradation of SOXB1 protein (Angerer et al., 2005; Angerer et al., 2007). As a result SOXB1, which is initially present in all of the nuclei is progressively restricted to the ectoderm. Thus the opposing SOXB1 and β-catenin activities segregate the lineages. Similarly in Drosophila Wnt signaling down regulates SoxNeuro and Dichaete expression, whereas SoxNeuro and Dichaete proteins attenuate Wnt-target gene transcription (Overton et al., 2007). This mutual repression sets up discrete spatial domains of Wnt activity that maintain tissue boundaries in epidermis.

Drosophila SoxF and Sox9 in the mouse intestine interact with the Wnt pathway in a classical negative feedback loop (Figure 4H). Wnt signaling promotes SoxF expression and SOXF protein then represses Wg (Wnt) gene transcription (Dichtel-Danjoy et al., 2009) thereby restricting the secreted Wnt ligand to a narrow strip of cells preventing over proliferation of the wing epithelium. Similarly Sox9 expression in the mouse intestinal epithelium requires Wnt signaling, but SOX9 then locally attenuates Wnt-target gene transcription (Blache et al., 2004; Bastide et al., 2007) thus ensuring that Wnt-stimulated proliferation is kept in check.

Negative feedback loops such as this appear to be a common feature of the canonical Wnt pathway and β-catenin/TCF are know to directly activate the transcription of a number of Wnt-antagonists (i.e. Axin, Dkk1, Nemo kinase etc.) that feedback to attenuate Wnt activity (Logan and Nusse, 2004). In some instances the Wnt regulation of Sox gene expression is likely to be direct, for example, in mouse taste buds and the chick neural plate, Sox2 transcription is regulated by TCF DNA-binding sites in the Sox2 promoter (Okubo et al., 2006; Takemoto et al., 2006). This may explain for example why Sox2 levels are elevated in aggressive breast cancer, which correlates with a poor prognosis (Chen et al., 2008). Hyperactive Wnt signaling might promote Sox2 expression and then SOX2 cooperatively stimulates β-catenin activity in an auto-regulatory amplification.

Conclusions and Future Perspectives

A flurry of studies in the last 3–4 years has revealed that SOX-Wnt interactions can regulate development and disease in a variety of contexts. While Sox genes clearly have Wnt-independent roles, there are numerous instances in the literature where SOX and Wnt are implicated in the same biological process but a functional interaction has not yet been tested, such as the maintenance of embryonic stem cells.

The relationship between SOX factors and Wnt signaling is evolutionarily ancient, and the pathways interact at many levels. SOX factors appear to modulate β-catenin/TCF activity through a variety of mechanisms including; protein-protein interactions, DNA-binding, recruitment of cofactors, and protein stability. Moreover the same SOX protein, such as SOX9 appears to employ different mechanism in different contexts. Further studies are needed to determine whether these reflect distinct cell-specific functions, or whether different experimental systems have revealed different aspects of a common underlying mechanism. It will be important to examine protein complex formation on endogenous target genes promoters in vivo and to more fully explore the role of post-translational modifications.

In addition the role of SOX-TCF interactions are understudied and complicated by the fact that TCFs undergo extensive alternative splicing (Hoppler and Kavanagh, 2007). Considering that SOX and TCF factors are structurally related and that they both bind β-catenin, as well as recognize similar DNA-sequences and remodel local chromatin architecture, it is perhaps not surprising that they coordinately regulate Wnt-target genes. This suggests that SOX factors might also be considered to be Wnt-signaling effectors acting to fine-tune the temporal and spatial activity of this potent pathway.

The regulation of Wnt signaling may be a feature shared by all SOX factors. Considering that there are over 20 SOX proteins encoded in the vertebrate genome and each of these might: recognize slightly different DNA sequences, interact with different cell-specific transcription factors and modulate β-catenin/TCF activity in different ways -this could well account for a vast array of cell-specific transcriptional responses to canonical Wnt signaling.