Wnt/Beta-catenin signaling regulation and a role for biomolecular condensates

Abstract

Wnt/β-Catenin signaling plays key roles in tissue homeostasis and cell fate decisions in embryonic and post-embryonic development across the animal kingdom. As a result, pathway mutations are associated with developmental disorders and many human cancers. The multiprotein destruction complex keeps signaling off in the absence of Wnt ligands and needs to be downregulated for pathway activation. We discuss new insights into destruction complex activity and regulation, highlighting parallels to the control of other cell biological processes by biomolecular condensates that form by phase separation to suggest that the destruction complex acts as a biomolecular condensate in Wnt pathway regulation.

eTOC

Wnt/β-Catenin signaling controls tissue homeostasis and cell fate decisions. The multiprotein destruction complex suppresses signaling in absence of Wnt ligand and is downregulated for pathway activation. Schaefer and Peifer discuss new insights into destruction complex activity and regulation, highlighting evidence that it acts as a biomolecular condensate in pathway control.

The cell is a complex place. As within a city, within the boundaries of a cell hundreds of different activities – from transcription to translation to metabolic reactions to signaling events – occur simultaneously in different places. To organize this complexity, cells dedicate particular locations to particular tasks. Some of this sequestration of activities is accomplished via membrane-bound compartments, ranging from the ER or Golgi to the smallest exocytic vesicle. These compartments allow segregation from the bulk cytoplasm, and interchange between compartments occurs via specialized transport systems. However, relying on specialized transport is insufficient to organize the vast volume of cytoplasm and nucleoplasm that is not encompassed by a membrane-bound organelle. To solve this problem, cells evolved an additional mechanism of organizing cellular compartments making use of physical properties of macromolecules that remove the need for a membrane enclosure. Some of these structures were large enough to merit recognition by cell biology’s pioneers (Gall, 2000)— for example, nucleoli or Cajal bodies, locations of ribosome or spliceosome assembly within nuclei, or the germplasm of animal eggs where determinants specifying germ cell fate reside.

In the past decade scientists recognized that these entities are examples of a much broader group of non-membrane bound cellular compartments that organize specific proteins and/or RNAs. They are key to diverse cellular processes including transcription, the DNA damage response, and cellular signaling (Banani et al., 2017; Holehouse and Pappu, 2018). Pioneering work on the C. elegans germline P granules and on signaling centers organized by SH3 domain proteins led to the idea that these structures assemble by “liquid-liquid phase separation” (Brangwynne et al., 2009; Li et al., 2012a). Multivalent interactions among their protein and/or RNA constituents lead to self-assembly, creating compartments separated from the bulk cytoplasm where the concentration of key players is exceptionally high, significantly speeding intricate reactions and/or processes (reviewed in Banani et al., 2017). The field emerged from concepts from soft-matter physics and polymer chemistry, which provide a biophysical basis and theoretical framework for this behavior. Critically, molecules can freely diffuse within, into and out of these structures, as they are not enclosed in a lipid bilayer and are often liquid-like in nature. This is thought to allow them to serve as centralized functional hubs for particular cellular processes, in which substrate molecules can enter, assemble, disassemble, or be modified, and products leave, and also as serve as storage depots for key players to be deployed at later times.

Structures like these recently were given the broad name “biomolecular condensates”, reflecting the broad range of cellular and molecular processes that occur within them. Condensates can display a range of physical properties, from liquid-like to more solid-like, and these properties can change over time. Here we focus on liquid-like condensates. These condensates have a number of defining properties (Banani et al., 2017; Fig. 1), though precise definitions are still being established. Each is a non-membrane bounded structure ranging up to micron scale that concentrates proteins and/or RNAs at a particular cellular site. They assemble by multivalent interactions mediated by multidomain proteins and/or RNAs with multiple protein or RNA interaction sites (Fig. 1). Many of the proteins involved contain “intrinsically disordered regions” – stretches of protein sequence that lack tertiary structure, are often not highly conserved in sequence, and self-interact or include within them interaction sites for other proteins (Fig. 1A-B). Intrinsically disordered regions are often tethered to folded domains (Mittal et al., 2018). Even after phase separation, protein components freely diffuse into and out of the condensate structures. Some condensates can transition to a more gel-like state (Wang et al., 2018), with reduced exchange with the bulk cytosol, a process that can contribute both to function and to pathogenesis. One key to understanding assembly of condensates is the ability to reconstitute phase separation behavior in vitro, with minimal components (Fig. 1D). Both in vitro and in vivo, liquid condensates can fuse and relax to minimize surface tension. The rapidly expanding universe of biological processes and structures encompassed under the biomolecular condensate umbrella and the challenge of defining the rules governing their assembly, disassembly, and function have made this one of the fastest growing areas of cell biology. As will be discussed in this review, the structures that regulate and transduce signals in the Wnt pathway, key to directing cell fate and maintaining tissue homeostasis during and after development, share many features with biomolecular condensates. In this perspective, we discuss the recent advances in the understanding of Wnt signaling and frame them in the context of the emerging idea that many key cellular processes are carried out in large non-membrane bound cellular compartments, an idea we think provides new insights into destruction complex function and regulation. Due to space limitations, we focus on Wnt/βCatenin signaling, not its variants, and on its core conserved components; other proteins with tissue- or animal phyla-specific roles will be neglected, though they are important for a full picture of Wnt signaling and its regulation (e.g. Adler and Wallingford, 2017; Green et al., 2014; Malinauskas and Jones, 2014)

Figure 1: Steps for building a biomolecular condensate.

A) Two separate proteins or RNAs (Green backbone and blue backbone) each contain multiple protein:protein or protein:RNA interaction sites sites (e.g. Orange, Purple, and Magenta circles and ovals) B) Homo and heteromeric interactions between molecules form multivalent linakges. C) These interactions induce condensate formation. D) Example of a biomolecular condensate in vitro. The droplets are formed from 8 uM Whi3 protein and 5 nM BNI1 RNA in 150 mM KCl after 4 hours incubation—picture provided Erin Langdon and Amy Gladfelter. E) Key properties of liquid-like biomolecular condensates.

The textbook model of Wnt signaling

The transformation of a fertilized egg into the body of an animal is among the most remarkable events in biology. Individual cells must choose fates based on their position, and then maintain those fates for a lifetime through tissue homeostasis. Cell-cell communication is critical for this, and a handful of cell-cell signaling pathways play especially important roles. Among these is the Wnt pathway (Nusse and Clevers, 2017), which directs cell fates from the initial establishment of the vertebrate body axes to the detailed architecture of the kidney or nervous system. The key developmental roles of these pathways mean that mutations in pathway components lead to congenital diseases, which in the case of the Wnt pathway include bone density and growth disorders (e.g. Robinow Disease) and progressive vision loss (Familial exudative vitreoretinopathy).

The same signaling pathways play critical roles in tissue homeostasis, maintaining proper cell numbers by regulating tissue stem cell proliferation. To ensure signaling occurs only at the right time and place, dedicated negative regulatory machinery has evolved to keep signaling completely off in the absence of ligand. In the Wnt pathway this is accomplished by the destruction complex, a multiprotein machine that targets the key Wnt-effector β-catenin βCat) for phosphorylation, and ultimate ubiquitination and destruction. Mutations in destruction complex proteins like Adenomatous polyposis coli (APC) occur in a wide variety of cancers and play the initiating role in virtually all colorectal tumors (Kandoth et al., 2013). As a result, mechanisms by which Wnt signaling is regulated are the subject of intensive research. Ultimately, developing ways to target activated Wnt signaling could provide new cancer therapies.

Like most key signaling pathways regulating development, the primary output of the Wnt/βCatenin pathway is a change in the cell’s transcriptional program. This occurs through regulation of the levels of βCat, a co-activator of transcription (reviewed in Gammons and Bienz, 2017; Nusse and Clevers, 2017; Stamos and Weis, 2013). In the absence of Wnt signaling βCat levels are kept low by the βCat destruction complex (Fig. 2A). At the core of this complex are the tumor suppressor APC, the scaffold Axin, and two kinases, GSK3 and CK1. This complex recruits βCat, where it is sequentially phosphorylated by CK1 and then GSK3. Once βCat is phosphorylated, it is transferred to the Cullin-based E3 Ligase SCF^βTrCP, polyubiqitinated, and then recognized by the proteasome and degraded. As a result, Wnt-regulated transcription is OFF.

Figure 2: Textbook model of Wnt signaling.

A) In the absence of a Wnt ligand, the destruction complex (APC, Axin, CK1, GSK3) recruits βcat for phosphorylation. Once phosphorylated, βcat can be recognized by the E3 ligase, SCFβTrCP, ubiquitinated, and then passed to the proteasome for ultimate protein degradation. B) Wnt signaling induces the formation of the Wnt receptor complex of Wnt/Frizzled/LRP5/6/Dvl. This complex recruits Axin and induces down-regulation of the destruction complex. Levels of cytoplasmic βcat rise, allowing βcat to enter the nucleus, bind to TCF/LEF family of transcription factors, and co-activate transcription of Wnt target genes.

Wnt ligands bind both the 7 transmembrane Frizzled (Fz) and the single-pass transmembrane LRP5/6 receptors (Fig. 2B; reviewed in DeBruine et al., 2017; Nusse and Clevers, 2017). The Wnt/Fz/LRP complex recruits the cytoplasmic protein Disheveled (Dvl in mammals, fly Dsh). GSK3 phosphorylates LRP5/6 (fly Arrow), creating a binding site for Axin and recruiting it to the membrane. This downregulates destruction complex activity. The primary mechanism of downregulation is not yet clear, as data support diverse mechanisms ranging from disassembly of the complex, inhibition of GSK3 kinase activity, Axin degradation, sequestration of destruction complex core proteins (e.g., GSK3 sequestration in multivesicular bodies, or loss of E3-ligase interaction; reviewed in Tortelote et al., 2017). Destruction complex inhibition allows βCat levels to rise and it enters the nucleus, binding to T-Cell Factor (TCF)/Lymphoid Enhancer factor (LEF) family DNA binding proteins. TCF/LEF proteins bind Wnt regulated genes, initiating different multiprotein complexes in cells where Wnt signaling is ON or OFF (reviewed by Gammons and Bienz, 2017). Thus the ultimate output of Wnt signaling occurs when the TCF/LEF:βCat complex activates transcription of Wnt target genes. Although the field agrees on the main components of the destruction complex, it is becoming increasingly clear that the textbook model is missing key aspects of Wnt signaling and its regulation. Below we review how new research is providing fresh insights into the mechanisms of destruction complex function and regulation.

The Wnt-regulatory destruction complex—is it a biomolecular condensate?

Examining with hindsight the unfolding of our understanding of destruction complex components, regulation and function reveals striking parallels between many of its properties and those of biomolecular condensates. Two key non-enzymatic components, APC and Axin, are complex multidomain scaffolding proteins containing folded domains that bind other proteins along with long intrinsically disordered regions that have binding sites for other destruction complex proteins, including βCat (Fig. 3A, reviewed in Stamos and Weis, 2013). For example, human APC is predicted to have 50% disordered content (Piovesan et al., 2018). Axin has an N-terminal Regulator of G-protein signaling (RGS) domain that binds APC, a C-terminal DIX domain that mediates head-to-tail polymerization, and an intervening intrinsically disordered region containing binding sites for βCat and the two key kinases, as well as for the phosphatase PP2A (Behrens et al., 1998; Fagotto et al., 1999; Hart et al., 1998; Ikeda et al., 1998; Kishida et al., 1998; Sakanaka et al., 1999; Sakanaka et al., 1998; Sakanaka and Williams, 1999; Zeng et al., 1997; Fig.3A). Axin’s multiple binding sites allow it to bring βCat into proximity to the kinases GSK3 and CK1. APC has a conserved N-terminal region that mediates oligomerization (Kunttas-Tatli et al., 2014) and includes an Armadillo (Arm) repeat domain that can bind diverse partners. This is followed by a long intrinsically disordered region, embedded within which are multiple copies of two distinct types of binding sites for βCat (Fig. 3A), multiple binding sites for Axin, and other conserved sites for which the binding partners remain undetermined (reviewed in Stamos and Weis, 2013). The multivalent nature of APC and Axin and their intrinsically disordered regions are shared features with known components of biomolecular condensates, suggesting that they may also form a condensate.

Figure 3: Axin, APC, and Dvl are multidomain proteins with intrinsically disordered regions that accumulate in structured non-membrane bound puncta.

A) Cartoon of the structures of APC, Axin, Dvl, and an APC-Axin Chimera, highlighting domains mediating self-interaction as well as interaction sites with other proteins. Proteins found in condensates often have intrinsically disordered regions and can polymerize and/or oligomerize other components in the condensates. Domains and motifs are as labelled. Solid green lines indicate direct interactions. Dotted orange line indicates identified interaction regions. Circled arrows indicate regions of self-interaction. Wavy black lines indicate probable intrinsically disordered regions in the central parts of APC and Axin. The Chimera is composed of the essential regions of APC and Axin for βCat degradation. This chimeric minimized APC-Axin protein can restore Wnt regulation in APC mutant colorectal cancer cells (Pronobis et al., 2017). B-F. Drosophila APC2 and Axin constructs expressed in APC mutant SW480 colorectal cancer cells. B) Close up of APC and Axin puncta visualized using standard confocal microscopy shows colocalization, but no underlining structure (Image originally published in Pronobis et al., 2015: DOI: 10.7554/eLife.08022) C-F) SIM Imaging. C) When Axin is expressed alone, many small puncta are formed. D) Close-up of a puncta from B, revealing that Axin polymers form knots. E) Co-transfection of APC2 and Axin. They accumulate together in puncta. Note that puncta are larger and fewer than in C. F) SIM images of a punctum, revealing that APC2 and Axin each form intertwined cables with multiple potential interaction sites, thus revealing the internal structure of the destruction complex. (Image originally published in Pronobis et al., 2015.: DOI: 10.7554/eLife.08022)

The localization of key destruction complex players is also striking when considered through the lens of phase separation. When Axin is expressed in many different cell types, both in vitro and in vivo, it forms large protein “puncta”, and recruits into them APC and other destruction complex proteins, thus increasing their effective local concentration (Figs. 3 and 4, e.g. Cliffe et al., 2003; Fagotto et al., 1999; Faux et al., 2008; Pronobis et al., 2015; Thorvaldsen et al., 2015). In cultured cells overexpressing Axin alone or with APC2, these can reach diameters of nearly a micron (Pronobis et al., 2015; Thorvaldsen et al., 2015), although when expressed at nearly endogenous levels in Drosophila embryos they are at the diffraction limit (Schaefer et al., 2018). Recent analysis by correlative fluorescence and electron microcopy confirmed these puncta are not enclosed in a membrane (Thorvaldsen et al., 2015). When Axin is expressed at endogenous levels, puncta are also seen (Faux et al., 2008). As we discuss in detail below, APC is required for puncta assembly in vivo, and puncta localization is regulated by Wnt signaling, consistent with puncta as active players in Wnt regulation (Cliffe et al., 2003; Mendoza-Topaz et al., 2011; Schaefer et al., 2018). Puncta formation and destruction complex function depend at least in part on the ability of Axin’s C-terminal DIX domain to oligomerize (Kishida et al., 1999; Sakanaka and Williams, 1999; Faux et al., 2008). The DIX/DAX (Dishevelled/Axin; referred to as DIX below) domain was initially defined because it is conserved with Dvl, a positive effector of Wnt signaling. Dvl also can form puncta, both when heterologously over-expressed in cells (Axelrod et al., 1998; Yang-Snyder et al., 1996) and in its endogenous state (Miller et al., 1999), though the nature, size and function of the endogenous puncta remains unclear. Like Axin, Dvl puncta formation also depends on its DIX domain (Schwarz-Romond et al., 2005). Strikingly, Dvl and Axin physically interact and co-localize in puncta (Fig. 3A) (Fagotto et al., 1999; Julius et al., 2000; Kishida et al., 1999). Dvl puncta are recruited to the membrane by the Wnt receptor (Axelrod et al., 1998; Miller et al., 1999; Yang-Snyder et al., 1996). Early work suggested that Dvl’s DIX domain associated with vesicles (Capelluto et al., 2002). However, subsequent work failed to reveal any co-localization of Dvl with vesicular markers. Instead, live imaging revealed that Dvl puncta are protein oligomers that can grow by fusion (Schwarz-Romond et al., 2005). Puncta containing Axin and APC can also fuse (Kunttas-Tatli et al., 2014). FRAP analysis further revealed that Dvl, Axin, and APC all freely diffuse into and out of puncta (Pronobis et al., 2015; Schwarz-Romond et al., 2007b). Together, these data suggest that Dvl and Axin puncta meet most of the criteria for biomolecular condensates (Fig. 3A) and demonstrate that puncta assembly is key for destruction complex function and regulation.

Figure 4: In vivo recruitment of APC2 into Axin:GFP puncta.

A) Model illustrating Wnt signaling in a Drosophila embryo. One row of cells per body segment produce and secrete the Wnt Wingless (Wg). It forms a graded distribution and stabilizes the fly βCat (Arm), leading to graded Arm levels across the body segment. B) Stage 9 Drosophila embryos expressing Axin:GFP at near endogenous levels. Anterior to the left. Axin:GFP localization is dependent on Wnt/Wg expression. In the absence of Wg, Axin:GFP is found in cytoplasmic puncta containing 10-100s of Axin molecules. In the presence of Wg, Axin puncta are located along the membrane and there is an increase in cytoplasmic GFP. Staining for endogenous APC2 shows that APC2 is recruited to both cytoplasmic Axin puncta and membrane localized Axin puncta. (Image originally published in Schaefer et al., 2018.: DOI: 10.1371/journal.pgen.1007339)

The destruction complex is an internally ordered structure that assembles by polymerization

Some biomolecular condensates form via a network of multivalent interactions without a strong underlying structural scaffold, while others can assemble into a more gel-like polymerized state (Banani et al., 2017). Early studies of the destruction complex suggested it is more structured, supporting a model of ‘signaling by reversible polymerization’ (Schwarz-Romond et al., 2007a). The DIX domains of Dvl and Axin polymerize by head to tail interactions, forming filaments that can be visualized by EM or X-ray crystallography (Schwarz-Romond et al., 2007a), similar to tubulin, actin, and septins. Critically, point mutations in their respective DIX domains that block polymerization reduce Dvl’s ability to promote Wnt signaling and attenuate Axin’s ability to inhibit Wnt signaling (Fiedler et al., 2011; Schwarz-Romond et al., 2007a). These data suggested that destruction complex puncta have an internal structure conferred by DIX domain polymerization. Evidence for this model was subsequently obtained using high-resolution microscopy. Structured illumination super-resolution microscopy (SIM) of Drosophila Axin and APC2 expressed in SW480 cells resolved the puncta into structured entities (Fig. 3B-F) (Pronobis et al., 2015). When expressed alone, Axin formed puncta with an internal structure that resembled toroids or knots, potentially representing DIX domain filaments (Fig. 3C-D). Co-expressing Axin and APC2 led to their co-recruitment in puncta and puncta assembly was enhanced, with the largest puncta on the order of micron size. Strikingly, intertwined homo-filaments of Axin and APC2 were resolved, further supporting the idea that regulated polymerization underlies destruction complex assembly (Fig. 3E-F). In parallel, scientists examined endogenous Axin puncta stabilized by inhibiting the enzyme Tankyrase, a known regulator of Axin levels, using both SIM and correlative fluorescence and electron microscopy (Thorvaldsen et al., 2015). These data also revealed micron scale puncta in which Axin, βCat, and Tankyrase formed an intermeshed network of filaments, which electron microscopy verified are not membrane-bounded. Together, these data suggest that the destruction complex is a biomolecular condensate with a structured scaffold, built around intertwined polymers of Axin and APC (Fig. 5).

Figure 5. A revised model of the destruction complex.

Polymers of Axin and of APC, mediated by polymerization of their respective DIX and ASAD/Arm repeat domains, intertwine. Polymers interact via the RGS:SAMP and Arm repeat-Axin motif interactions. Polymers concentrate ßcat, GSK3, and CK1 (not shown), accelerating phosphorylation.

A functional destruction complex contains many more than four proteins

Early work in Xenopus oocytes suggested that destruction complex assembly is limited by Axin’s very low protein abundance compared to all other destruction complex proteins—up to 5000-fold lower (Lee et al., 2003; Salic et al., 2000). This suggested that Axin was exquisitely rate-limiting, a factor built into many mathematical models of signaling (e.g,.Lee et al., 2003). Interestingly, recent work in both flies and mammals suggest that in many cell types Axin is expressed at levels similar to those of APC (Kitazawa et al., 2017; Schaefer et al., 2018; Tan et al., 2012). These data will power updated models and perhaps new insights of Wnt signaling and its regulation.

Textbook models of the destruction complex often represent it as a simple 4-protein complex of APC:Axin:GSK:CK1 (Fig. 1A), despite the fact that we knew for years that Axin polymerization is necessary for efficient βCat regulation. Defining the number of molecules in a functional destruction complex has been a challenge. Some destruction complex proteins localize to other locations where they have distinct roles --e.g., GSK3 regulates Wnt, Hedgehog, Insulin, PI3K, and Erk signaling (Cormier and Woodgett, 2017) and APC regulates the cytoskeleton (Nelson and Nathke, 2013). Thus, not every molecule of GSK3 or APC in the cell localizes to the destruction complex. Second, effective antibodies to key players were not available. This was particularly true for Axin. To overcome this, investigators over-expressed Axin and/or APC, hoping the larger complexes formed would serve as expanded representations of endogenous complexes, providing important insights. However, new reagents recently allowed scientists to look at Axin expressed at endogenous or near endogenous levels, using either a new antibody against fly Axin (Wang et al., 2016a) or epitope-tagged Axin expressed at near-endogenous levels (Schaefer et al., 2018; Wang et al., 2016a; Yang et al., 2016). This revealed that in Wnt-OFF cells Axin assembles into puncta similar but smaller than to those that assemble after overexpression in cultured cells, and these puncta recruit APC (Schaefer et al., 2018). This validates early work examining endogenous Axin in MDCK cells (Faux et al., 2008) and supports the idea that the destruction complex is a supermolecular machine containing tens to hundreds of molecules rather than a simple 4-protein complex. But how large is this complex? The ability to express GFP-tagged Axin at near endogenous levels provided insight. Fluorescence intensity measurements compared to GFP-labeled complexes of known molecular composition revealed that in Drosophila embryos, active destruction complex puncta contain on average ~260 Axin molecules (range ~60-930; Schaefer et al., 2018). Mass spectroscopy provides an alternate mechanism of putting numbers on destruction complex proteins—recent analyses suggest HEK293 cells each contain ~13,000 Axin proteins (Kitazawa et al., 2017)—however, with current technology this assessment requires many simplifying assumptions. Together, these data emphasize that the destruction complex is a large molecular machine, consistent with it being a biomolecular condensate.

Stabilizing destruction complex supermolecular assembly is a key factor in βCat destruction

Axin’s DIX domain is necessary for its self-polymerization but how is polymerization regulated? Newly synthesized Axin molecules can either nucleate a new Axin filament or add to an existing polymer. Several studies focused on APC’s role in forming or stabilizing Axin filaments. APC is required for assembly of Axin puncta and therefore active destruction complexes (Mendoza-Topaz et al., 2011). Co-expression revealed that APC2 stabilizes Axin assembly, as measured by destruction complex volume, increased complexity of Axin filaments, and decreased Axin turnover (Pronobis et al., 2015). Strikingly, as APC co-expression increased the size of Axin puncta, it simultaneously decreased the number of puncta, consistent with the idea that APC2 promotes Axin addition to existing polymers over nucleation of new polymers (Pronobis et al., 2015).

When the destruction complex was visualized, both Axin and APC2 appeared as intertwined filaments, suggesting that APC may also polymerize (Fig. 3E-F). Human APC1 has an N-terminal coiled-coil oligomerization domain (Joslyn et al., 1993), but this is not conserved in Drosophila family members. However, another N-terminal region of APC, the APC self-association domain (ASAD), is conserved in Drosophila, Xenopus, and humans, mediates Drosophila APC2 self-association (Kunttas-Tatli et al., 2014), and together with APC’s Arm repeats can mediate puncta formation (Pronobis et al., 2015). The ASAD and adjacent Arm repeats are required for destruction complex function (Kunttas-Tatli et al., 2014; McCartney et al., 2006; Roberts et al., 2012a). Interestingly, loss of APC’s oligomerization domain eliminated APC’s ability to stabilize Axin within the destruction complex (Kunttas-Tatli et al., 2014; Pronobis et al., 2015). These data suggest that APC polymerization is required to initiate and stabilize formation of functional destruction complexes.

APC’s stabilization of the destruction complex requires two different Axin:APC interactions mediated by different domains (Fig.3A; Pronobis et al., 2015). The first is via the well-established interaction of the Axin-RGS domain with APC’s SAMPS (Spink et al., 2000). Retention of at least one SAMP is essential for APC function in both mice (Smits et al., 1999) and flies (Roberts et al., 2011). The second APC:Axin interaction is between APC’s Arm repeats and a less well-defined region in Axin’s central intrinsically disordered region (Fagotto et al., 1999; Pronobis et al., 2015). Not all SAMP motifs are functionally similar. Fly APC2 has 2 SAMPs, and data suggest one recruits Axin while the other aids in efficient βCat destruction by an unknown mechanism (Kunttas-Tatli et al., 2015). Intriguingly, in colorectal cancer cells the SAMP interaction is dispensable if APC and Axin are fused into a single minimal polypeptide (Pronobis et al., 2017). The ability of APC to stabilize Axin in the destruction complex is further enhanced by a bridging interaction involving βCat’s ability to bind both APC and Axin, an interaction disrupted by the vertebrate-specific βCat binding protein ICAT (Ji et al., 2018). Together, these data support the idea that one key role of APC in the destruction complex is to stabilize Axin, increasing destruction complex size and thus its effectiveness.

Given this, it is also important to critically evaluate whether Axin polymerization is essential for all destruction complex activity, or whether it serves to enhance efficiency. Work using overexpression gave mixed results—in colorectal cancer cells Axin’s DIX domain is important for its ability to target for destruction (Kishida et al., 1999;.Sakanaka and Williams, 1999; Faux et al., 2008). In contrast, deleting the DIX domain blunts but does not eliminate the ability of Axin overexpression to inhibit Wnt signaling in Xenopus embryos (Itoh et al., 1998; Fagatto et al., 1999; Fukui et al., 2000; Fiedler et al., 2011). Loss-of-function mutations provide the most stringent test, and this has been addressed in Drosophila, where site-directed mutants were expressed into Axin null mutants. Strikingly, deleting Axin’s DIX domain does not eliminate function in destruction and in fact, led to some level of constitutive inhibition of Wnt signaling. In contrast, however, a point mutant in the DIX domain that blocks polymerization had significantly reduced rescue activity in vivo (Fiedler et al. 2011). This suggests two things. First, the DIX domain may not be absolutely essential for destruction complex function, but instead may enhance efficiency. This may reflect residual function of smaller protein complexes, or may reflect the ability of APC oligomerization to mediate formation of a biomolecular condensate even in the absence of the DIX domain. Second, these data reinforce the idea that the DIX-mediated interaction with Dsh is critical for inactivating destruction.

Like many biomolecular condensates, the destruction complex assembles via many multivalent interactions. This makes it surprisingly robust to removal of some but not all protein interaction motifs. For example, individually deleting most of APC’s βCat binding sites, Axin’s βCat binding site, or Axin’s RGS domain have only modest effects in vivo (Kremer et al., 2010; Kunttas-Tatli et al., 2012; Peterson-Nedry et al., 2008; Roberts et al., 2011; Yamulla et al., 2014). However, some binding sites are essential individually (Axin’s GSK3 binding site) or when deleted in concert (AxinΔRGSΔArm; Peterson-Nedry et al., 2008). This powered synthetic biology approaches to design a “minimal βCat destruction machine” which retains function in colorectal cancer cells (Pronobis et al., 2017). Future work assessing the limits of this robustness in the intact animal will be of interest.

The destruction complex serves a second role as a sink for cytoplasmic βCat, modulating Wnt-regulated transcription

Another mystery with regard to the destruction complex is why it needs multiple βCat binding motifs, both in APC and Axin (Fig. 3A). Presumably at least one binding site is needed to capture βCat and present it to the two kinases, but why so many binding sites? Most APC proteins have multiple copies of two distinct types of binding sites for βCat embedded in the central intrinsically disordered region, the 15- and 20-amino acid repeats (15R and 20R;Eklof Spink et al., 2001; Ha et al., 2004; Liu et al., 2006; Xing et al., 2004). Each 20R has a different affinity for βCat, with an affinity range of 100-fold (Liu et al., 2006), and phosphorylation of 20Rs by GSK3 dramatically increases their affinity for βCat (Ha et al., 2004; Liu et al., 2006). This led to the hypothesis that APC’s high affinity binding sites ensure capture and destruction of any newly synthesized βCat and thus maintain the very low levels of nuclear βCat needed to keep Wnt signaling tightly off in the absence of Wnt ligands. In contrast, the low affinity sites come into use when cells are exiting a phase of Wnt signaling and βCat levels are initially high. These latter sites would sequester βCat in the cytoplasm and speed destruction, hastening termination of Wnt-regulated gene expression (Ha et al., 2004; Krieghoff et al., 2006). This model was tested in colorectal cancer cells and in Drosophila, by systematically deleting 15R and 20R βCat binding sites. Strikingly, the highest affinity βCat binding sites are dispensable in targeting βCat for destruction—instead the binding sites collaborate in an additive way to fine-tune Wnt signals by cytoplasmic retention of βCat, supporting the sequestration hypothesis (Kunttas-Tatli et al., 2012; Roberts et al., 2011; Yamulla et al., 2014). In fact, a fly ApC2 mutant lacking all the 15R and 20Rs retains ability to restore APC function in colorectal cancer cells, although it is not is not fully functional in destruction in vivo in Drosophila (Yamulla et al., 2014). The dispensability of APCs βCat binding sites is likely because Axin also plays a role in cytoplasmic retention of βCat in Drosophila (Tolwinski and Wieschaus, 2001), via the single βCat binding site in Axin’s intrinsically disordered region (Xing et al., 2003). This suggests redundancy of the βCat binding sites in APC and Axin in ensuring βCat destruction, a designed synthetic minimal destruction complex containing essential regions of APC and Axin that restored βCat regulation in colorectal cancer cells solely utilized Axin’s βCat binding site (Pronobis et al., 2017). Existing data also suggest other, more complex roles for the multiple βCat binding sites. In vivo analysis of Drosophila APC2 mutants lacking βCat binding sites revealed an aspect of in vivo regulation that remains to be understood—rather than restoring the graded levels of βCat seen in wildtype, they led to a sharp ON/OFF transition (Yamulla et al., 2014), reminiscent of the behavior of certain mutations in Drosophila βCat (Orsulic and Peifer, 1996) and of Axin mutants lacking the βCat binding site (Peterson-Nedry et a., 2008). This is consistent with some sort of threshold feedback response. Together, these data reveal that the multiple βCat binding sites in the destruction complex provide it with the ability to regulate Wnt signaling in multiple ways, from complete inactivation to more subtle modulation.

Other conserved motifs in APC’s intrinsically disordered region also play key functions

In addition to the βCat and Axin binding sites, APC’s intrinsically disordered region also contains another highly conserved motif which did not have a known binding partner, variously called conserved sequence B (B) or the catenin inhibitory domain (CID, Fig. 3A). Strikingly the B/CID motif is essential for APC function in Wnt regulation in both flies and mammals (Kohler et al., 2009; Roberts et al., 2011), with its deletion causing defects in Wnt regulation equivalent to those of completely eliminating APC. Intriguingly, an immediately adjacent motif, 20R2, which resembles other 20Rs but lacks key residues that mediate binding to βCat (Kohler et al., 2008; Liu et al., 2006), is also absolutely essential for Wnt regulation. Consistent with their essential nature, the R2/B motif was one of the sequences that needed to be included in a synthetic, minimized chimera of APC and Axin that can restore βCat destruction in APC mutant colorectal cancer cells (Fig. 3A; Pronobis et al., 2017). Together, B and 20R2 may form a binding site for an unidentified partner. In fact, the R2/B motif may be the sequence targeted for removal in the protein truncations found in colorectal tumors (Kohler et al., 2009). Further examination revealed that 20R2/B regulates one of the two APC:Axin binding interactions: the interaction between APC’s Arm repeats and Axin’s mid-region. Intriguingly, the function of 20R2/B is regulated by phosphorylation by GSK3 (Pronobis et al., 2015). Together, these data led to a model in which phosphorylation of the 20R2/B motifs triggers a conformational change in APC, releasing one of the two APC;Axin interactions and allowing transfer of phosphorylated βCat to the E3 ubiquitin ligase, as part of a catalytic cycle. This model is consistent with other data, revealing that loss of GSK3 in fly embryos leads to βCat accumulation in the destruction complex (Schaefer et al., 2018), and suggesting that inhibiting βCat release to the E3 ligase is a key step by which Wnt signaling inactivates the destruction complex (Li et al., 2012b). This model also helps explain a paradox in the field—colorectal cancer cells are defective in βCat destruction but not in βCat phosphorylation (Yang et al., 2006). These data suggest that despite the multivalent nature of destruction complex interactions, some interactions may regulate key conformational changes that are essential for function.

These data left open the identity of the potential binding partner of the R2/B motif. In 2013 evidence emerged that α-catenin (αcat), like βCat a protein found both in cadherin-based cell-cell junctions and in the destruction complex (Rubinfeld et al., 1993), can bind the R2/B motif. Assays in cultured cells supported the idea that αcat facilitates βCat ubiquitination and proteolysis, and also suggested a possible inclusion in the nucleus with TCF/LEF DNA binding proteins (Choi et al., 2013). These data are intriguing, but the physiological role of αcat in Wnt regulation remains in question. Drosophila zygotic αcat mutants (Desai et al., 2013; Sarpal et al., 2012), zygotic βCat mutants deleting the αcat binding site (Orsulic and Peifer, 1996), and C. elegans αcat mutants (Costa et al., 1998) all lack apparent defects in Wnt signaling or its regulation. Continued work is needed to further clarify the relevant binding partner of R2/B and its role in destruction complex function. Another important challenge is to more fully define which parts of the intrinsically disordered regions of APC and Axin are essential for their functions, which act to enhance efficiency and which may be redundant—for example, a chimera covalently linking Axin and APC with portions of the intrinsically disordered regions of each protein deleted retains function after overexpression in colorectal cancer cells (Fig. 3A; Pronobis et al., 2017).

APC may play additional positive and negative roles in Wnt signaling

Current data support roles for APC in stabilizing the destruction complex, promoting βCat transfer to the E3 ligase, and sequestering βCat in the cytoplasm. However, APC may have additional functions. Nuclear roles for APC have been suggested (e.g. Sierra et al., 2006), but sequestering APC at a variety of cytoplasmic locations does not disrupt Wnt regulation in flies or mammalian cells (Roberts et al., 2012b) suggesting a nuclear role of APC is not essential. APC may also act at the level of the Wnt receptor, inhibiting baseline activity in the absence of Wnt ligands by promoting clathrin-dependent receptor endocytosis (Saito-Diaz et al., 2018)—however, this role is only exhibited by certain APC family members. Another intriguing hypothesis is that APC can also promote Wnt signaling (Tacchelly-Benites et al., 2018; Takacs et al., 2008). Genetic studies in which the levels of both Drosophila APC proteins, APC1 and APC2, were manipulated in parallel revealed that reducing APC2 function attenuated activated Wnt signaling in fly eyes induced by loss of APC1 (Takacs et al., 2008). The underlying mechanism was not fully defined, though effects on Axin stability and phosphorylation were suggested (Tacchelly-Benites et al., 2018; Takacs et al., 2008; Wang et al., 2016a). Interestingly, overexpressing APC2 in Drosophila embryos enhances accumulation of βCat only in Wnt-ON cells, supporting the idea that APC2 can aid in turning down destruction complex activity in response to Wnt signals (Schaefer et al., 2018). These data further illustrate the intricate nature of Wnt regulation.

Regulating a biomolecular condensate: Wnt signaling changes destruction complex localization and assembly via a second condensate, the signalosome

The data above summarize our knowledge of active destruction complexes. Another challenge is to define how its function is down-regulated by Wnt signaling. Many of us initially spoke of turning the destruction complex “OFF”, but this is inaccurate. Wnt signaling does not fully inactivate the complex—it retains, at least initially, the ability to phosphorylate βCat (Kim et al., 2013; Li et al., 2012b), such that rate of βCat turnover is reduced but not halted (Hernandez et al., 2012). In retrospect, this was apparent in early work in Drosophila, as mutational inactivation of GSK3 or APC function led to much higher levels of βCat accumulation than those seen in cells receiving Wnt signals (e.g. Ahmed et al., 2002; Akong et al., 2002). How is destruction complex activity repressed? After Wnt ligands bind their receptors, Axin is recruited into a second protein complex with many properties of a biomolecular condensate, the “signalosome” (reviewed in Gammons and Bienz, 2017). Axin recruitment to the Fz:LRP co-receptor is mediated both by direct interactions with the phosphorylated LRP5/6 tail (Davidson et al., 2005; Mao et al., 2001; Tamai et al., 2004; Zeng et al., 2005) and by a less well defined role of Dvl (Bilic et al., 2007; Cliffe et al., 2003). Consistent with this, Drosophila Axin (at endogenous or near endogenous levels) is found in cytoplasmic puncta in the absence of Wnt signaling, while in cells receiving Wnt signals Axin puncta are recruited to the plasma membrane (Schaefer et al., 2018). Interestingly, in Wnt receiving cells, the number of Axin molecules in puncta is reduced (Schaefer et al., 2018) while the cytoplasmic Axin pool is increased (Schaefer et al., 2018; Wang et al., 2016a; Yang et al., 2016) suggesting that after Axin is recruited to the membrane some change occurs that either inhibits Axin self-polymerization or inhibits its stability within condensates. But what is the nature of this change?

Wnt signaling causes a switch in the destruction complex mix and destabilizes it

One potential change involves a switch in binding partners. Dvl was one of the first proteins shown to be required for Wnt signaling and is a key regulator that inhibits destruction complex function in response to Wnt signaling (reviewed in Mlodzik, 2016). As noted above, Dvl and Axin both have a DIX domain. This shared domain mediates both self-polymerization and hetero-polymerization (Fiedler et al., 2011; Kishida et al., 1999; Schwarz-Romond et al., 2007a; Schwarz-Romond et al., 2007b; Smalley et al., 1999). Strikingly, the DIX domain is essential for Dvl function in promoting Wnt signaling in both Drosophila and mammalian cells, though not for its function in planar polarity (Yanagawa et al., 1995; Boutros et al., 1998; Axelrod et al., 1998; Li et al., 1999; Kishida et al., 1999; Julius et al., 2000).

Dvl is essential for Wnt receptor phosphorylation, Axin recruitment to the membrane and signalosome endocytosis, thus turning down destruction complex function (Bilic et al., 2007; Cliffe et al., 2003). Dvl binding to Fz, via Dvl’s DEP and/or PDZ domains (Axelrod et al., 1998; Wong et al., 2003), is followed by a conformational change that crosslinks Dsh polymers, increasing local concentration of Dvl at the receptor as it is endocytosed (Gammons et al., 2016). DEP-mediated Dvl cross-linking may drive Axin recruitment by increased avidity, driving Dvl:Axin hetero-polymerization. FRAP analysis revealed that Dvl:Axin co-assembly enhances Axin turnover in puncta (Schwarz-Romond et al., 2007b). Dvl destabilization of Axin puncta thus provides one mechanism by which Dvl could inactivate the destruction complex. In contrast, APC inclusion in Axin puncta stabilizes them, increasing destruction complex efficiency (Pronobis et al., 2015). Intriguingly, co-expressing Axin, APC, and Dvl2 in cultured cells revealed that APC:Axin:Dvl2 complexes are rare while Axin:APC or Axin:Dvl2 complexes are more frequent, consistent with a competition between APC and Dvl for interaction with Axin (Mendoza-Topaz et al., 2011). Competition for Axin binding is also consistent with the fact that protein levels of APC:Axin:Dsh in Drosophila embryos are all in the same order of magnitude (Schaefer et al., 2018). The possibility that DIX:DIX interactions between Dvl and Axin inhibit Axin is also consistent with fact that Drosophila Axin∆DIX is constitutively active in βCat destruction (Peterson-Nedry et al., 2008). One remaining question is how Axin decides between binding its different partners? Condensates form via multiple multivalent interactions. Perhaps changes in Axin posttranslational modifications in response to Wnt signaling, discussed below,alter the charge of the intrinsically disordered region, reducing interaction between Axin:APC or promoting Axin:Dvl interaction. Future research into the rules regulating the competition between assembly/disassembly of the destruction complex and that of the signalosome will provide essential insights.

One consequence of supermolecular assembly: GSK3 plays both positive and negative roles in the destruction complex via its access to many targets

GSK3, first discovered as a kinase regulating glycogen metabolism, plays pleiotropic roles in the cell, regulating multiple signaling pathways (Cormier and Woodgett, 2017). GSK3 was one of the first proteins with a known biochemical role to be placed in the Wnt pathway and the first negative regulator, a role defined via genetic analysis in Drosophila (Peifer et al., 1994; Siegfried et al., 1992; Siegfried et al., 1990; Siegfried et al., 1994). This raises the question of how is pathway specificity maintained? Recruitment into different supermolecular complexes provides a mechanism. Both CK1 and GSK3 are recruited by Axin into the destruction complex, where they sequentially phosphorylate βCat, priming it for destruction. However, subsequent work revealed that GSK3 plays many roles in the pathway, both positive and negative. Perhaps it is not surprising that recruiting an active kinase into a multiprotein complex allows it to phosphorylate many of its constituent proteins. Within the active destruction complex, GSK3 phosphorylates Axin to keep Axin “open” for βCat interaction (Kim et al., 2013). It also phosphorylates APC on its 20Rs to increase affinity for βCat (Ha et al., 2004; Liu et al., 2006; Xing et al., 2004), and on R2/B to facilitate βCat release to the E3 ligase (Pronobis et al., 2015). Diverse Axin phosphorylation events appear to all depend on APC (Tacchelly-Benites et al., 2018). However, GSK3 is also recruited into the Wnt signalosome, where it plays important roles. In response to Wnt signaling, CK1 and GSK3 phosphorylate the tail of LRP5/6, creating a binding site that facilitates Axin recruitment to the receptor complex for inactivation (Tamai et al., 2004; Zeng et al., 2005), a process that is visualized in Drosophila as GSK3-dependent recruitment of Axin puncta to the membrane (Cliffe et al., 2003; Schaefer et al., 2018). Intriguingly, the LRP5/6 phosphorylated tail then can act as a GSK3 inhibitor (Piao et al., 2008; Stamos et al., 2014; Wu et al., 2009), providing another mechanism by which Wnt activation turns down the destruction complex. GSK3 phosphorylation of Axin also allows it to be “open” for binding LRP5/6 (Kim et al., 2013). In these two roles GSK3 is a positive effector of Wnt signaling. Vertebrates have two GSK isoforms, and questions remain about the roles they play (or do not) in regulating this pathway. Genetic analysis demonstrated that the two isoforms are largely redundant for Wnt regulation, with single mutants having tissue-specific defects (Doble et al., 2007; Hoeflich et al., 2000), though recent studies with isoform specific inhibitors suggest possible differences in isoform function in Wnt regulation (Chen et al., 2017).

Axin post-translational regulation plays complex roles

Axin has a central role in destruction complex downregulation, with models ranging from enhanced Axin degradation to its dissociation from GSK3 or changes in its assembly state (reviewed in MacDonald and He, 2012; Nusse and Clevers, 2017). These diverse data may reflect differences in different animals and/or tissues or, as we think more likely, may reflect initial versus more long-term mechanisms for elevating βCat levels and thus Wnt signaling. Many models of destruction complex inhibition involve Axin regulation via post-translational modifications, including phosphorylation/dephosphorylation (outlined above), ADP-ribosylation, or ubiquitination. One common consequence of long-term Wnt signaling is down-regulation of Axin protein levels (Kofron et al., 2007; Liu et al., 2005; Mao et al., 2001; Tolwinski et al., 2003; Yang et al., 2016). GSK3 phosphorylation of Axin can stabilize it, while Axin de-phosphorylation by PP2A leads to degradation (Willert et al., 1999; Yamamoto et al., 1999). It will be intriguing to examine this in light of recent work that highlights the role of kinases as “dissolvases” of other biomolecular condensates (Rai et al., 2018). Further since condensates form due to low affinity interactions between proteins, changes in phosphorylation could change the charge of Axin, APC or Dvl, to alter their relative affinities for one another.

Two other Axin post-translational modifications, ADP-ribosylation and ubiquitination, can also regulate Axin stability and assembly into condensates, and thus affect Wnt signaling. The poly(ADP-ribose) polymerase Tankyrase adds ADP-ribose to target proteins, which are then often ubiquitinated and destroyed (Hsiao and Smith, 2008; Mariotti et al., 2017). Tankyrase binds and ADP-ribosylates Axin (Huang et al., 2009), targeting it for ubiquitination by RNF146, a poly(ADP-ribose)-directed E3 ligase (Callow et al., 2011; Zhang et al., 2011; Zhou et al., 2011). Wnt signaling induces Axin ADP-ribosylation within 30 min and this enhances Axin’s ability to bind active LRP5/6 (Yang et al., 2016). Consistent with a role in destruction complex destabilization, Tankyrase inhibition stabilizes Axin puncta formation (de la Roche et al., 2014; Thorvaldsen et al., 2015; Waaler et al., 2011). Intriguingly, Tankyrase can also bind APC (Croy et al., 2016). However, things are not that simple, as Axin is still degraded after long Wnt signaling exposure in a Tankyrase-independent fashion (Wang et al., 2016b), suggesting Tankyrase is not solely responsible for Axin degradation. Further, Drosophila lacking Tankyrase are viable and fertile (Feng et al., 2014), and mice lacking both Tankyrase proteins, while embryonic lethal, survive to E10 without obvious defects in Wnt signaling. Thus while Tankyrase can fine tune destruction complex activity, as it does in the Drosophila intestine (Wang et al., 2016c), it is not an essential pathway regulator. Wnt signaling may also downregulate Axin levels by other means. The RING domain E3 ligase Seven in absentia (Sina) homolog SIAH 1/2 can ubiquitinate βCat, labeling it for proteasomal degradation independent of βTrCP (Liu et al., 2001; Matsuzawa and Reed, 2001). However, SIAH1/2 also directly binds Axin near its GSK3 binding site, competing with GSK3 for access to Axin (Ji et al., 2017). Perhaps after Wnt signaling inhibits GSK3 interaction with Axin, SIAH binds Axin and labels it for proteasomal degradation. Once again, however, it is important to note that the single Drosophila Sina family member, is adult viable without Wnt signaling obvious defects (Carthew and Rubin, 1990). This is a more general issue, as outside of the core destruction components APC, Axin and the kinases, many other identified destruction complex interactors have milder or tissue specific phenotypes. For example, ICAT mutant phenotypes are restricted to the forebrain, neural crest and kidney (Satoh et al., 2004).

While long-term Wnt exposure decreases Axin levels, recent studies in vivo in Drosophila suggest that Axin levels in the cytoplasm initially increase after Wnt signaling is activated, and only decrease several hours later (Wang et al., 2016a; Yang et al., 2016). Total embryonic Axin levels increase over the same time frame, but it is hard to rule out that this is not simply an effect of activating zygotic Axin expression in all cells. This degree of Axin elevation is at the threshold at which Axin can inhibit Wnt signaling (3-9×; Peterson-Nedry et al., 2008; Schaefer et al., 2018; Wang et al., 2016b). Why should Wnt signaling elevate Axin levels when this could enhance βCat destruction and thus inhibit Wnt signaling? Perhaps the increased pool of cytoplasmic Axin is largely “inactivated” Axin molecules that cannot form stable puncta. In fact, membrane-localized Axin puncta in Wnt-ON cells harbor only half the number of Axin molecules as cytoplasmic puncta in Wnt-OFF cells, while cytoplasmic Axin levels rise (Schaefer et al., 2018). Together, these data reveal many levels at which Axin is regulated by post-translational modification and open up questions for future research, defining which changes are the initial response to Wnt signals and which are adjustments allowing longer term modulation of signaling.

APC mutations in colorectal cancer target specific aspects of destruction complex function

Activating mutations in the Wnt pathway play roles in many cancers, including endometrial and liver cancer, but are most prominent in colorectal tumors, where they initiate oncogenesis (reviewed in Zhang and Shay, 2017). ~10% of colorectal tumors have gain-of-function βCat mutations disrupting phosphorylation and thus destruction; a few have loss-of-function Axin mutations, but >80% are APC mutant (Kandoth et al., 2013). These mutations have a very striking feature—unlike most tumor suppressors where selection favors homozygous null mutations, all (or virtually all) colorectal tumors have at least one APC allele expressing a truncated protein. Intriguingly, the truncations occur in a small region of the protein, the mutation cluster region (MCR; Kohler et al., 2008), leading researchers to explore what properties are lost or retained to favor selection. Most now accept the “just right” hypothesis (Albuquerque et al., 2002), which proposes that complete loss of APC function leads to such high levels of Wnt activity that oncogenic stress sensors trigger apoptosis. This suggests the truncated APC protein retains some function. What functions are retained and which lost? One critical feature lost in the truncated proteins are the SAMP motifs, the high affinity Axin binding sites. Mice mutant for one APC allele truncated to lose all SAMPs are tumor prone, whereas mice carrying an APC allele with a slightly longer truncation that retains one SAMP are not tumor prone (and in fact are homozygous viable! Smits et al., 1999). However, closer examination of the MCR suggests that this is more complex than the SAMPs -- the B/CID motif, which is just N-terminal to the last SAMP (Fig. 3A) and thus also disrupted in most or all tumor truncations also may play a role (Kohler et al., 2008), suggesting that selection may favor loss of both the SAMPS and B/CID. What function is retained by truncated APC to prevent selection for null mutations? Truncated APC proteins like those in tumors cannot promote βCat destruction but can still bind βCat (Roberts et al., 2011) and mediate its phosphorylation (Yang et al., 2006). Retaining βCat in the cytoplasm may thus be how destruction complexes containing truncated APC dampen but do not eliminate Wnt signaling. It will be intriguing to further explore assembly and function of destruction complex condensates carrying truncated APC, as deleting the SAMP motifs reduces but does not eliminate APC incorporation into puncta (Pronobis et al., 2015; Roberts et al., 2011).

The destruction complex is a multifunctional machine with other targets including the cytoskeleton

While best known for roles in Wnt regulation of βCat, destruction complex proteins also have other functions. For example, the destruction complex may phosphorylate and regulate the stability of other targets (e.g., Kim et al., 2009). This could provide a mechanism by which Wnt regulation of biomolecular condensates can control other key cellular pathways and thus integrate signaling events. One place this may occur involves connections between Wnt signaling and BMP signaling, another key developmental pathway. Wnt signaling, via its regulation of GSK3 activity, controls phosphorylation of the BMP transcriptional effector Smad, though whether this requires the destruction complex remains unclear (Fuentealba et a., 2007; Eivers et al., 2011; Demagny et al., 20014; Demagny and De Robertis, 2015). Similarly, Wnt inhibition of GSK3 can also stabilize mitotic effector proteins by preventing their phosphorylation and subsequent ubiquitination, promoting either cell growth or cell division in different contexts (Acebron et al., 2014; Huang et al., 2015). It will be exciting to define the full range of proteins affected by this βCat-independent branch of Wnt signaling.

Destruction complex proteins also can regulate the cytoskeleton. Human APC binds microtubules and the microtubule plus end protein EB1(Munemitsu et al., 1994; Smith et al., 1994; Su et al., 1995). Subsequent work suggested roles for APC in spindle orientation, chromosome segregation, and polarity of neurons and migrating cells, but most studies rely on overexpressing full-length or truncated APC (reviewed in Nelson and Nathke, 2013; Rusan and Peifer, 2008). In Drosophila and C. elegans, analysis of loss of APC function revealed roles in genome stability via regulation of centrosome migration and function of the formin Diaphanous (McCartney et al., 2001; Poulton et al., 2013; Webb et al., 2009), mitotic spindle orientation (Yamashita et al., 2003; Sugioka et al., 2018; Sugioka et al., 2011), and microtubule dynamics in neuronal dendrites (Mattie et al., 2010; Weiner et al., 2016), but cast doubt on suggested roles in axon or dendrite polarity (Rusan et al., 2008). At least some APC proteins also regulate actin dynamics, working together with Diaphanous (Breitsprecher et al., 2012; Jaiswal et al., 2013), to control focal adhesion turnover (Juanes et al., 2017). When considering cytoskeletal effects of APC loss, however, it is critical to rule out places where Wnt signaling itself regulate downstream cytoskeletal events (e.g., Elbaz et al., 2016; Eom et al., 2014; Hayden et al., 2007; Nakagawa et al., 2017; Yokota et al., 2009). Other destruction complex or signalosome proteins also have cytoskeletal roles. Mouse Axin has suggested roles in oocyte meiosis (He et al., 2016), axon and dendrite morphogenesis, and intermediate neuronal progenitor differentiation in the cerebral cortex (Chen et al., 2015; Fang et al., 2013; Fang et al., 2011). Once again, however, one must be cautious in differentiating Wnt-independent from Wnt-dependent roles. Dvl and Fz have well known roles in planar cell polarity and in cilia, but these appear to be largely independent of the destruction complex and Wnt/βCat signaling (reviewed in Adler and Wallingford, 2017). Finally, APC, Axin, Dvl and βCat can localize to centrosomes in at least some cell types (reviewed in Bryja et al., 2017; Mbom et al., 2013), but the physiological functions of this remain unclear. Both flies and mice lacking centrosomes develop without strong defects in Wnt signaling (Basto et al., 2006; Bazzi and Anderson, 2014), and basic spindle assembly functions of centrosomes do not depend on APC, Axin, Dvl or βCat. Defining Wnt-independent cytoskeletal roles remains a challenge for the field, as is determining whether these are independent roles for individual destruction complex proteins or whether the complex acts as a biomolecular condensate in this role as well.

Looking towards the future

The sections above outline some of the many questions raised by new insights into the regulation of Wnt signaling. Looking more globally, it is an exciting time for the Wnt signaling field and for the broader arena of signaling pathway and biomolecular condensate research. Virtually every week brings new examples of cellular structures assembled by phase separation (e.g. Du and Chen, 2018; Shan et al., 2018) or defining the mechanisms by which they assemble (e.g. (Harmon et al., 2017; Rai et al., 2018; Zeng et al., 2018). The new insights described above, supporting the perspective that the destruction complex and signalsome should be viewed as biomolecular condensates, open a wide variety of new directions and questions for research that can be addressed using ever more powerful tools ranging from in vitro reconstitution to in vivo genetic analysis to molecular modeling.

In vitro reconstitution is perhaps the most critical step, but this poses a number of challenges. APC and Axin are both large proteins with long intrinsically disordered regions, so expression of intact full-length versions will be a challenge. Newly developed approaches for protein expression in and purification from insect cells using nanobodies may help in this regard (Bekesi et al., 2018). Other questions can be addressed via a combination of reconstitution and cell-based assays. New CRISPR-based tools allow us to engineer genes and the proteins they encode at the endogenous loci, freeing us from the potential artifacts of overexpression, and allowing us to follow localization of Wnt regulatory proteins throughout the signaling process. In parallel, new super-resolution microscopy and correlative light and electron microscopy approaches provide the potential to look inside biomolecular condensates and tease out their structure.

These new tools will empower many new questions. What proteins nucleate formation of each supermolecular complex, and how do the core proteins choose between nucleating a new assembly or adding to an old one? How do post-translational modifications govern condensate assembly or disassembly? How do newly identified components of the Wnt receptor complex, like TMEM59 (Gerlach et al., 2018), Gpr124 and Reck (Eubelen et al., 2018) affect supermolecular assembly of the signalingsome and downregulation of destruction complex function? How do different supermolecular complexes, like the destruction complex and the SCF-E3 ligase, communicate with one another to control flow of molecules and information? How do proteins like APC, βCat, and GSK3, which play multiple roles within each cell, choose between different supermolecular assemblies? These and other questions will keep many of us happily employed for many years to come.