BIOLOGICAL OVERVIEW

The vertebrate Axin protein, the product of the mouse fused gene, binds to beta-catenin to inhibit Wnt signaling. A homolog of vertebrate Axin, Drosophila Axin (here termed Axin in conformity to FlyBase usage) has been identified. Axin knockout produces phenotypes that are similar to overexpression of the Drosophila Wnt gene wingless (wg). Overexpression of Axin produces phenotypes similar to loss of wg. Axin overexpression can modify phenotypes elicited by wg and another Drosophila Wnt gene, Wnt oncogene analog 2. Using immunoprecipitation of endogenous Axin embryonic protein, it has been shown that Axin interacts with Armadillo and Zeste-white 3. The loss-of-function and overexpression phenotypes show that Axin, like its mammalian counterpart, acts as a negative regulator of wg/Wnt signaling (Willert, 1999b). Together, Axin, APC and Shaggy/Zeste-white 3 form a protein complex termed the "ß-catenin destruction complex."

Two approaches have been used to clone Axin. Hamada (1999) performed a yeast two-hybrid screen of a Drosophila embryo cDNA library using the Armadillo repeat domain of Armadillo as target and identified Axin as an Arm-interacting protein. Willert (1999b) identified Axn by searching the EST database with the protein sequence of mouse Axin. An EST with significant homology to the DIX (a domain similar between Axin and Dishevelled) domain of Axin was identified and used to isolate the full-length clone from an embryonic Drosophila cDNA library (Willert, 1999b).

Loss-of-function mutant phenotypes can be mimicked by the injection of double stranded RNA. This method, referred to as RNA interference (RNAi), was initially successfully used in C. elegans and more recently in Drosophila. RNAi was used to disrupt the function of Axn during patterning of the embryonic cuticle, a process that requires Wg signaling. The wild-type embryo secretes a cuticle consisting of a repeated pattern of denticle belts with intervening naked regions. Loss of wg leads to a cuticle covered with denticles -- one lacking any naked areas. Overexpression of wg in the embryo leads to loss of all denticle structures, i.e. a naked cuticle. Disruption of AXN expression by RNAi leads to a similar naked cuticle, suggesting that Axn functions to down-regulate Wg signaling. Virtually all injected embryos had extra naked cuticle, ranging from partially to nearly completely naked. Control injection of either sense or antisense single stranded AXN mRNA does not cause any phenotypic changes. As expected, wg RNAi produces a partial wg-like cuticle. Thus, mimicking loss of Axn function by RNAi leads to phenotypes similar to overexpressing wg, consistent with the model that Axn is a negative regulator of Wg signaling (Willert, 1999b).

A P-element insertion near the beginning of the Axn gene has been shown to disrupt expression of the gene to produce a loss-of-function allele of Axn. Embryos lacking zygotic Axn are still wild type and only upon removal of the maternally contributed Axn gene product is a naked cuticle revealed. Thus, the RNAi experiments successfully disrupt the maternally contributed Axn gene product and produce a phenotype identical to that of a loss-of-function mutation in the gene (Willert, 1999b).

To address further whether Axn regulates Wg signaling, the UAS-Gal4 system was used to overexpress Axn in various tissues. The Daughterless (Da)-Gal4 driver expresses early during embryogenesis; when combined with UAS-wg, it produces a completely naked cuticle. Overexpression of Axn using the Da-Gal4 driver produces a loss of wg-like phenotype, a cuticle covered with denticles, consistent with overexpression of Axn blocking Wg signaling in the embryonic epidermis. To extend this study Axn was misexpressed in the wing using the 69B-Gal4 driver. wg is expressed in a narrow stripe along the presumptive wing margin where it is required for proneural achaete-scute complex gene expression and for the formation of margin bristles. Loss of Wg signaling along the wing margin, as in the case of dsh loss-of-function clones, leads to loss of these margin bristles and notches along the wing. Overexpression of Axn in the wing also produces this wing notching effect, a result consistent with Axn interfering with the Wg signaling pathway at the wing margin (Willert, 1999b).

Deletion of the RGS domain of mouse Axin produces a dominant negative Axin protein, as assayed by its ability to induce a secondary axis when overexpressed in Xenopus embryos (Itoh, 1998 and Zeng, 1997). An analogous mutation was constructed in Axn that deletes the entire RGS domain (DaxinDeltaRGS; deletion of amino acids 50 to 172). Surprisingly, when overexpressed, this Axn mutant produces the same phenotypes in the embryo and in the wing as does wild-type Axn overexpression. Thus, the RGS domain is dispensable in overexpression assays for Axn. Thus, the RGS domain, which interacts with the APC protein (Behrens, 1998 and Hamada, 1999), is dispensable for overexpressed Axin’s ability to block Wnt signaling. The discrepancy between the Xenopus and fly data with respect to the RGS deleted Axin may be explained in a number of ways. In analogy to APC, Axin overexpression in various organisms may produce distinct phenotypes. APC overexpression in mouse cell culture is known to down regulate Wnt signaling, while overexpression of APC in Xenopus embryos leads to axis duplication, a hallmark phenotype of genes that activate Wnt signaling. As another exception to the model of APC being a negative factor, the C. elegans APC homolog Apr-1 has been shown to positively regulate Wnt signaling. The reasons for the inconsistent phenotypes for APC loss- and gain-of-function are not yet understood (Willert, 1999b and references).

In support of the findings presented by Willert (1999b), Hart (1998) has shown that overexpression of human Axin lacking the APC binding domain (i.e. the RGS domain) promotes Axin’s ability to downregulate beta-catenin levels. Although both mutant forms of mouse Axin (Zeng, 1997) and Axn primarily delete the RGS domain, there are subtle differences in the deletions: DaxinDeltaRGS deletes the entire RGS domain (amino acids 50-172 in Axn corresponding to amino acids 89-216 in mouse Axin) while mouse AxinDeltaRGS deletes part of the RGS domain and some N-terminally flanking sequence (amino acids 124-227 in Axin corresponding to amino acids 87-183 in Axn). In sum, it remains to be established whether the interaction between Axin and APC is critical for Axin to inhibit Wnt signaling. A genetic analysis of the Axn gene and identification of mutations that disrupt this interaction will likely provide insights into this question (Willert, 1999b).

The above data clearly demonstrate that Axn loss-of-function or overexpression produces phenotypes similar to wg overexpression or loss-of-function, respectively. However, it fails to demonstrate that Axn can modify the effects of Wnt signaling. To examine whether ectopic Axn expression can interfere with the effects of Wnt signaling, wg was misexpressed in the eye and DWnt-2 was misexpressed in the ovary. When overexpressed in the eye, wg produces a glassy eye that is greatly reduced in size compared to a wild-type eye. Misexpression of Axn in the eye produces a weakly rough eye. When Axn and wg are both overexpressed, the morphology of the eye is intermediate to that of a wild-type eye and that of a wg eye, indicating that Axn can interfere with the Wg signal (Willert, 1999b).

In order to address whether Axin inhibits only Wg, or whether it can function as a more general inhibitor of Wnt signals, an examination was carried out to see whether Axn could interfere with a different Wnt signal. Four Wnt genes have been identified in Drosophila (see The World Wide Web Wnt Window ), but mutations have been described only for two of them: wg and DWnt-2. Loss of DWnt-2 produces a muscle migration defect in the male gonads, resulting in male sterility, and a lack of the characteristic pigment cells that migrate over the male testis. Ovaries are normally not surrounded by pigment cells, but misexpression of DWnt-2 in females can induce ectopic male-specific pigment cells. This dominant phenotype of DWnt-2 misexpression was used to address whether Daxin can block ectopic pigment cell formation. Overexpression of Axn strongly reduces the frequency of DWnt-2-mediated pigment cell formation in ovaries, demonstrating that Axn can block the DWnt-2 signal in addition to the Wg signal. Overexpression of a dominant-negative Pangolin/dTCF lacking the amino-terminal Arm binding domain (DN-dTCF) also blocks the ectopic DWnt-2 induced pigment cells in the ovary (Willert, 1999b).

The Adenomatous polyposis coli tumour suppressor is essential for Axin complex assembly and function and opposes Axin's interaction with Dishevelled

Most cases of colorectal cancer are linked to mutational inactivation of the Adenomatous polyposis coli (APC) tumour suppressor. APC downregulates Wnt signalling by enabling Axin to promote the degradation of the Wnt signalling effector β-catenin (Armadillo in flies). This depends on Axin's DIX domain whose polymerization allows it to form dynamic protein assemblies ('degradasomes'). Axin is inactivated upon Wnt signalling, by heteropolymerization with the DIX domain of Dishevelled, which recruits it into membrane-associated 'signalosomes'. How APC promotes Axin's function is unclear, especially as it has been reported that APC's function can be bypassed by overexpression of Axin. Examining apc null mutant Drosophila tissues, it was discovered that APC is required for Axin degradasome assembly, itself essential for Armadillo downregulation. Degradasome assembly is also attenuated in APC mutant cancer cells. Notably, Axin becomes prone to Dishevelled-dependent plasma membrane recruitment in the absence of APC, indicating a crucial role of APC in opposing the interaction of Axin with Dishevelled. Indeed, co-expression experiments reveal that APC displaces Dishevelled from Axin assemblies, promoting degradasome over signalosome formation in the absence of Wnts. APC thus empowers Axin to function in two ways-by enabling its DIX-dependent self-assembly, and by opposing its DIX-dependent copolymerization with Dishevelled and consequent inactivation (Mendoza-Topaz, 2011).

Drosophila apc null mutants wer used for stringent in vivo function tests to show that APC is indispensable in Drosophila tissues for Axin's activity in assembling functional degradasomes that destabilize Armadillo. The evidence suggests that the same is also true for APC mutant colorectal cancer cells in which Axin–GFP, if expressed at low levels, shows a marked tendency to fail in assembling functional degradasomes. This provides a new insight into how APC promotes the destabilization of Armadillo/β-catenin—namely by enabling Axin to assemble degradasomes. The failure of this assembly step explains why Axin fails to destabilize Armadillo/β-catenin, given that this function of Axin crucially depends on its DAX-dependent polymerization. It is emphasized that previous studies have shown that APC loss-of-function can be bypassed by Axin overexpression, which indicated a non-obligatory role of APC in the destabilization of β-catenin/Armadillo. However, these studies underestimated APC's essential role in this process since Axin was assayed under conditions of residual APC function, and possibly also since high Axin overexpression levels were used for complementation (Mendoza-Topaz, 2011).

Why does Axin fail to assemble functional degradasomes in the absence of APC? This is believed to be true due to a combination of two different effects of APC loss on Axin. First, Axin is destabilized in the absence of APC, so its levels may fall below the minimal cellular concentration required for DAX-dependent polymerization: note that the DIX domain auto-affinity is in the micromolar range, and so the DAX-dependent polymerization may not occur spontaneously at low cellular Axin concentrations, but might require a co-factor capable of clustering Axin, increasing its local concentration and nucleating polymerization. APC is a candidate for such a co-factor, given its relatively high cellular abundance and affinity to Axin which allow it to associate efficiently with Axin at physiological concentrations. APC might cluster Axin directly, by binding simultaneously to multiple Axin molecules through its multiple Axin-binding sites, or indirectly through additional factors (such as CtBP, itself capable of clustering APC. Notably, Axin would become independent of this co-factor if overexpressed at high enough levels, as this would allow it to overcome its low auto-affinity and to polymerize spontaneously (Mendoza-Topaz, 2011).

Second, the absence of APC (or of binding to APC) renders Axin prone to Dsh-dependent relocation from the cytoplasm to the PM, into signalosome-like particles. Since the recruitment of Axin into signalosomes normally blocks its function in promoting the phosphorylation and destabilization of β-catenin/Armadillo, its relocation to the PM might also explain its inactivity in the absence of APC. The evidence indicates that APC shields Axin from interaction with Dsh in the absence of Wnt signalling, to ensure Axin's function in the cytoplasmic degradasomes. This shielding function of APC may be particularly important in cells experiencing non-canonical Wnt signalling (promoting PM-association of Dishevelled), such as in third-larval instar wing discs.The observations in Drosophila tissues suggested that APC may compete with Dishevelled for association with Axin, which is strongly supported by evidence from co-expression experiments in mammalian cells (carried out in the absence of Wnt stimulation): these indicate that Axin cannot interact simultaneously with APC and Dvl2, and that APC is capable of displacing Dvl2 from Axin protein assemblies. The notion of a competition between APC and Dishevelled for their association with Axin is consistent with previous evidence from epistasis experiments in Drosophila embryos, which indicated that APC acts at the same level as Dishevelled rather than below it (Mendoza-Topaz, 2011).

Evidently, the function of APC that shields Axin from its interaction with Dishevelled is somehow antagonized by Wnt stimulation, which enables Dishevelled to bind to Axin and recruit it to the PM into signalosomes. Indeed, Wnt signalling may overcome the competition between APC and Dishevelled for their binding to Axin, allowing simultaneous interaction of all three proteins. Consistent with this, Axin–GFP appears to co-localize with E-APC in the Wg signalling zones of Drosophila embryos within the PM-associated signalosomes that are likely to also contain Dsh (although this has not been confirmed directly owing to the lack of a suitable Dsh antibody), suggesting that Wg signalling allows all three proteins to coincide in signalosomes (Mendoza-Topaz, 2011).

It is noted that a previous study in Drosophila uncovered a positive role of APC in antagonizing Axin (rather than promoting its function), thereby stimulating signalling through Armadillo. The key evidence supporting this rather unexpected conclusion was that the levels of Axin were upregulated in apc null mutant wing disc clones, as shown by immunofluorescence. This contrasts with the current result from apc mutant embryos, which shows much reduced levels of Axin–GFP, as judged by Western blotting. Although this quantitative biochemical approach is difficult to apply to apc mutant wing disc clones (owing to insufficient apc mutant material), the dramatic PM relocation of Axin–GFP observed in these clones may have been misled by the high levels of apical Axin, and mistaken for a general upregulation rather than simply a relocation (as has been shown for Axin–GFP). The current evidence reaffirms the negative role of APC in Wg/Armadillo signalling, demonstrating an essential function of APC in keeping Axin in the cytoplasm, where it enables it to assemble functional degradasomes (Mendoza-Topaz, 2011).

The main corollary of these results from Drosophila tissues is that APC promotes the DAX-dependent homopolymerization of Axin (required for degradasome assembly), and that it antagonizes the heteropolymerization between Axin and Dishevelled (mediated by DIX–DAX interaction). The latter is further supported by the evidence from co-expression experiments in mammalian cells that APC displaces Dvl2 from Axin puncta. This creates a mechanistic conundrum: APC binds to the N-terminal RGS domain of Axin, but appears to control the DAX-dependent interactions at its C-terminus. Although it is conceivable that APC achieves this at 'long range', given its unusually large size it is more likely that APC relies on additional factors, perhaps even on enzymes, to promote Axin's self-assembly at the expense of its heteropolymerization with Dishevelled. Future work will be required to determine the precise molecular mechanism by which APC enables Axin to assemble functional degradasomes and opposes its recruitment by Dishevelled, and how this is overcome during Wnt signalling (Mendoza-Topaz, 2011).

Wnt pathway activation by ADP-ribosylation: APC and Tnks maintain basal Axin levels below a critical in vivo threshold to promote robust pathway activation following Wnt stimulation

Wnt/β-catenin signal transduction directs metazoan development and is deregulated in numerous human congenital disorders and cancers. In the absence of Wnt stimulation, a multi-protein 'destruction complex', assembled by the scaffold protein Axin, targets the key transcriptional activator β-catenin for proteolysis. Axin is maintained at very low levels that limit destruction complex activity, a property that is currently being exploited in the development of novel therapeutics for Wnt-driven cancers. This study used an in vivo approach in Drosophila to determine how tightly basal Axin levels must be controlled for Wnt/Wingless pathway activation, and how Axin stability is regulated. For nearly all Wingless-driven developmental processes, a three- to four-fold increase in Axin was found to be insufficient to inhibit signaling, setting a lower-limit for the threshold level of Axin in the majority of in vivo contexts. Further, both the tumor suppressor Adenomatous polyposis coli (APC) and the ADP-ribose polymerase Tankyrase (Tnks) were found to have evolutionarily conserved roles in maintaining basal Axin levels below this in vivo threshold, and separable domains were defined in Axin that are important for APC- or Tnks-dependent destabilization. Together, these findings reveal that both APC and Tnks maintain basal Axin levels below a critical in vivo threshold to promote robust pathway activation following Wnt stimulation (Yang, 2016).

The Wnt/β-catenin signal transduction pathway directs fundamental processes during metazoan development and tissue homeostasis, whereas deregulation of Wnt signalling underlies numerous congenital disorders and carcinomas. Two multimeric protein complexes with opposing functions -- the cytoplasmic destruction complex and the plasma membrane-associated signalosome -- control the stability of the transcriptional co-factor β-catenin to coordinate the state of Wnt pathway activation. In the absence of Wnt stimulation, β-catenin is targeted for proteasomal degradation by the destruction complex, which includes the two tumour suppressors: Axin and Adenomatous polyposis coli (APC), and two kinases: casein kinase α (CK1α) and glycogen synthase kinase 3 (GSK3). Engagement of Wnt with its transmembrane receptors, Frizzled and low-density lipoprotein receptor-related protein 5/6 (herein LRP6), induces rapid LRP6 phosphorylation, recruitment of Axin to phospho-LRP6, and assembly of the signalosome, which includes two other Axin-associated components, GSK3 and Dishevelled (Dvl). Signalosome assembly results in the inhibition of β-catenin proteolysis; consequently stabilized β-catenin promotes the transcriptional regulation of Wnt pathway target genes (Yang, 2016).

As a key component in both the destruction complex and the signalosome, Axin is tightly regulated. Under basal conditions, Axin is maintained at very low levels, and serves as the concentration-limiting scaffold for assembly of the destruction complex. Following Wnt exposure, the rapid association of phospho-Axin with phospho-LRP6 triggers Axin dephosphorylation, inducing a conformational change that inhibits Axin's interaction with both the destruction and signalosome complexes. Axin is subsequently degraded; however, Axin proteolysis occurs several hours after Wnt exposure, and thus does not regulate Axin's essential role during the initial activation of the Wnt pathway (Yang, 2016).

The mechanisms that rapidly reprogram Axin from inhibitory to stimulatory roles following Wnt exposure remain uncertain. In current models, Wnt stimulation induces Axin's dissociation from the destruction complex, thereby promoting its interaction with the signalosome. As Wnt stimulation induces Axin dephosphorylation, decreased phosphorylation was postulated to facilitate the dissociation of Axin from the destruction complex; however, recent work revealed that the interaction of Axin with LRP6 precedes Axin dephosphorylation, and that dephosphorylation serves to inhibit, rather than enhance this interaction (Kim, 2013) Furthermore, some findings have challenged prevailing models, providing evidence that Axin's interaction with the destruction complex is not diminished upon Wnt stimulation. Thus, whereas the rapid switch in Axin function following Wnt stimulation is essential for the activation of signalling, the underlying mechanisms remain uncertain (Yang, 2016).

During investigation of this critical process, an unanticipated role was discovered for the ADP-ribose polymerase Tankyrase (Tnks) in the reprogramming of Axin activity following Wnt exposure. As Tnks-mediated ADP-ribosylation is known to target Axin for proteolysis, small molecule Tnks inhibitors have become lead candidates for development in the therapeutic targeting of Wnt-driven cancers. This study identified a novel mechanism through which Tnks regulates Axin: by promoting Axin's central role in rapid Wnt pathway activation. Wnt stimulation was found to modulate Axin levels biphasically in both Drosophila and human cells. Unexpectedly, Axin is rapidly stabilized following Wnt stimulation, before its ultimate proteolysis hours later. In an evolutionarily conserved process, the ADP-ribosylated pool of Axin is preferentially increased immediately following Wnt exposure. ADP-ribosylation enhances Axin's association with phospho-LRP6, providing a mechanistic basis for the rapid switch in Axin function following Wnt stimulation. These results thus indicate that Tnks inhibition not only increases basal Axin levels, but also impedes the Wnt-dependent interaction between Axin and LRP6, suggesting a basis for the potency of Tnks inhibitors in Wnt-driven cancers. Thus, Tnks not only targets Axin for proteolysis independently of Wnt stimulation, but also promotes Axin's central role in Wnt pathway activation, which may be relevant to the context-dependent activation of Wnt signalling and the treatment of Wnt-driven cancers with Tnks inhibitors (Yang, 2016).

Wnt exposure induces biphasic regulation in the level of Axin, and a large increase in the level of ADP-ribosylated Axin immediately after stimulation. ADP-ribosylation enhances the interaction of Axin with phospho-LRP6, and promotes the activation of Wnt signalling. These findings lead to three major revisions of the current model for the role of Tnks in the activation of the Wnt pathway. First, Tnks serves bifunctional roles under basal conditions and after stimulation, revealing a remarkable economy and coordination of pathway components. Second, the results provide a mechanistic basis for the rapid reprogramming of Axin function in response to Wnt stimulation, and thereby reveal an unanticipated role for Tnks in this process. These findings suggest that Wnt exposure either rapidly increases the ADP-ribosylation of Axin or inhibits the targeting of ADP-ribosylated Axin for proteasomal degradation, through mechanisms yet to be elucidated. Finally, pharmacologic inactivation of Tnks was shown to diminish the interaction of Axin with LRP6, revealing a previously unknown mechanism through which small molecule Tnks inhibitors disrupt Wnt signalling, distinct from their known role in stabilizing the destruction complex inhibitors (Yang, 2016).

In the absence of Wnt stimulation, the concentration-limiting levels of Axin regulate its scaffold function in the destruction complex. As components of the destruction complex participate in other signalling pathways, the low levels of Axin were proposed to maintain modularity of the Wnt pathway. The new findings indicate that Axin levels are not only regulated in the absence of Wnt, but also regulated biphasically following Wnt stimulation. This sequential modulation of Axin divides activation of the pathway into an early, fast phase and a delayed long-term phase. During embryogenesis, the earliest expression of Wg triggers the rapid appearance of Axin in segmental stripes, which is a novel hallmark for the initial activation of the pathway. The findings reveal that Wnt exposure induces a rapid increase in the total level of Axin, and importantly, a preferential increase in the level of the ADP-ribosylated Axin. The early Axin stripes are absent in Tnks null mutant embryos and are also absent when the Tnks binding domain in Axin is deleted. Therefore, it is proposed that Axin ADP-ribosylation contributes to Axin stabilization and to the rapid response to Wg stimulation (Yang, 2016).

It is postulated that the initial increase in levels of ADP-ribosylated Axin jump-starts the response to Wnt stimulation by enhancing the Axin-LRP6 interaction, whereas the subsequent decrease in Axin levels prolongs the duration of signalling by reducing destruction complex assembly. Thus, Wnt stimulation induces rapid increases in the levels of not only cytoplasmic β-catenin, but also ADP-ribosylated Axin. Previous work that coupled mathematical modelling with experimental analysis revealed that several Wnt signalling systems were responsive to the relative change in β-catenin levels, rather than their absolute value. This dependence was proposed to impart robustness and resistance to noise and cellular variation. The current data raise the possibility that a similar principle applies to changes in Axin levels on the Axin-LRP6 interaction, as the marked increase in ADP-ribosylated Axin levels following Wnt stimulation is evolutionarily conserved. Thus, the relative change in levels of ADP-ribosylated Axin may promote signalling following Wnt exposure by facilitating the fold change in β-catenin levels (Yang, 2016).

The current findings have relevance for the context-specific in vivo roles of Tnks in Wnt signalling suggested in previous studies. Tnks inhibition disrupts Wnt signalling in a number of cultured cell lines, but in vivo studies in several model organisms suggested that the requirement for Tnks in promoting Wnt signalling is restricted to specific cell types or developmental stages. In mice, functional redundancy exists between the two Tnks homologues, such that Tnks single mutants are viable and fertile, whereas double mutants display embryonic lethality without overt Wnt-related phenotypes. However, a missense mutation in the TBD of Axin2 that is predicted to disrupt ADP-ribosylation resulted in either activating or inhibiting effects on Wnt signalling that were dependent on developmental stage. Tnks inhibitors resulted in the same paradoxical effects, suggesting complex roles in mouse embryonic development. Analogously, treatment of fish with Tnks inhibitors resulted in no observed defects in Wnt-mediated processes during development; however, the regeneration of injured fins in adults, a process that requires Wnt signalling, was disrupted (Yang, 2016).

Similarly, the finding that Drosophila Tnks null mutants are viable (Wang, 2016a; Wang, 2016b; Feng, 2014) was unexpected, as Tnks is highly evolutionarily conserved, and no other Tnks homologues exist in fly genomes. Nonetheless, the current studies reveal that a less than twofold increase in Axin levels uncovers the importance of Tnks in promoting Wg signalling during embryogenesis. Therefore, it is postulated that Tnks loss can be compensated during development unless Axin levels are increased, but that the inhibition of Wg signalling resulting from Tnks inactivation cannot be attributed solely to increased Axin levels. Furthermore, Drosophila Tnks is essential for Wg target gene activation in the adult intestine, and exclusively within regions of the gradient where Wg is present at relatively low concentration. Thus, the context-specific roles of Tnks observed in different model organisms may reflect the mechanisms described herein, which reveal that the Wnt-induced association of Axin with LRP6 occurs even in the absence of Axin ADP-ribosylation, but is markedly enhanced in its presence. It is postulated that by enhancing this interaction, Tnks-dependent ADP-ribosylation of Axin serves to amplify the initial response to Wnt stimulation, and thus is essential in a subset of in vivo contexts (Yang, 2016).

The recent discovery that Tnks enhances signalling in Wnt-driven cancers has raised the possibility that Tnks inhibitors will offer a promising new therapeutic option. Indeed, preclinical studies have supported this possibility. Tnks inhibitors were thought previously to disrupt Wnt signalling solely by increasing the basal levels of Axin, and thus by increasing destruction complex activity. However, the current findings indicate that the degree to which the basal level of Axin increases following Tnks inactivation is not sufficient to disrupt Wnt signalling in some in vivo contexts. Instead, the results reveal that Tnks inhibition simultaneously disrupts signalling at two critical and functionally distinct steps: by promoting activity of the destruction complex and by diminishing an important step in signalosome assembly: the Wnt-induced interaction between LRP6 and Axin. On the basis of these findings, it is proposed that the efficacy of Tnks inhibitors results from their combined action at both of these steps, providing a rationale for their use in the treatment of a broad range of Wnt-driven cancers. Therefore, these results suggest that in contrast with the current focus on tumours in which attenuation of the destruction complex aberrantly activates Wnt signalling (such as those lacking APC), the preclinical testing of Tnks inhibitors could be expanded to include cancers that are dependent on pathway activation by Wnt stimulation. These include the colorectal, gastric, ovarian and pancreatic cancers that harbour inactivating mutations in RNF43, a negative Wnt feedback regulator that promotes degradation of the Wnt co-receptors Frizzled and LRP6 (Yang, 2016).

Wnt regulation: is exploring Axin-Disheveled interactions and defining mechanisms by which the SCF E3 ubiquitin ligase is recruited to the destruction complex

Wnt signaling plays key roles in embryonic development and adult stem cell homeostasis and is altered in human cancer. Signaling is turned on and off by regulating stability of the effector beta-catenin. The multiprotein destruction complex binds and phosphorylates beta-catenin, and transfers it to the SCF-TrCP E3-ubiquitin ligase for ubiquitination and destruction. Wnt signals act though Dishevelled to turn down the destruction complex, stabilizing beta-catenin. Recent work clarified underlying mechanisms, but important questions remain. This study explored beta-catenin transfer from the destruction complex to the E3 ligase, and test models suggesting Dishevelled and APC2 compete for association with Axin. This study found that Slimb/TrCP is a dynamic component of the destruction complex biomolecular condensate, while other E3 proteins are not. Recruitment requires Axin and not APC, and Axin's RGS domain plays an important role. Elevating Dishevelled levels in Drosophila embryos has paradoxical effects, promoting the ability of limiting levels of Axin to turn off Wnt signaling. When Dishevelled levels were elevated, it forms its own cytoplasmic puncta, but these do not recruit Axin. Superresolution imaging in mammalian cells raises the possibility that this may result by promoting Dishevelled:Dishevelled interactions at the expense of Dishevelled:Axin interactions when Dishevelled levels are high (Schaefer, 2020).

During embryonic development, cells must choose fate based on their position within the unfolding body plan. One key is cell-cell signaling, by which cells communicate positional information to neighbors and ultimately direct downstream transcriptional programs. A small number of conserved signaling pathways play an inordinately important role in these events in all animals. These include the Hedgehog, Notch, Receptor Tyrosine kinase, BMP/TGFβ, and Wnt pathways, which influence development of most tissues and organs. These same signaling pathways regulate tissue stem cells during tissue homeostasis and play critical roles in most solid tumors. Due to their powerful effects on cell fate and behavior, evolution has shaped dedicated machinery that keeps each signaling pathway definitively off in the absence of ligand (Schaefer, 2020).

In the Wnt pathway, signaling is turned on and off by regulating stability of the key effector β-catenin (βcat). In the absence of Wnt ligands, newly synthesized βcat is rapidly captured by the multiprotein destruction complex . Within this complex, the protein Axin acts as a scaffold, recruiting multiple partners. Axin and adenomatous polyposis coli (APC) bind βcat and present it to the kinases casein kinase 1 (CK1) and glycogen synthase kinase 3 (GSK3) for sequential phosphorylation of a series of N-terminal serine and threonine residues on βcat (Schaefer, 2020).

It has become increasingly clear that the destruction complex is not a simple four-protein entity. Instead, Axin directs assembly of destruction complex proteins into what the field originally described as 'puncta.' These are now recognized as examples of supermolecular, nonmembrane bound cellular compartments, referred to as biomolecular condensates. Condensate formation is driven by Axin polymerization via its DIX domain, by APC function, and by other multivalent interactions (Schaefer, 2020).

Ubiquitination by E3 ubiquitin ligases is a key mechanism for regulating protein stability. Once the destruction complex templates βcat phosphorylation, the most N-terminal phosphorylated serine forms part of the core of a recognition motif for a Skp-Cullin-F-box (SCF)-class E3 ubiquitin ligase. This E3 ligase ubiquitinates βcat for proteasomal destruction. SCF-class E3 ligases include Cullin1 (Cul1), Skp1, F-box proteins, and Ring box (RBX) subunits, which work together to bind substrates and attach multiple ubiquitin moieties. Cul1 is the scaffold of the complex, at one end binding Rbx1 and its associated E2-Ubiquitin proteins and at the other end binding Skp1. Skp1(SkpA in Drosophila) links Cul1 and the F-box protein-in this case, βTrCP. βTrCP (Slimb in Drosophila) contains the substrate recognition domain of the E3 ligase. The βcat recognition site spans the WD40 repeats on the C-terminal end of βTrCP. This domain forms a propeller structure with a pocket that binds only to phosphorylated proteins. βTrCP can bind multiple phospho-proteins and thus regulate diverse cell signaling pathways (e.g., NFκB and Hedgehog signaling). After βTrCP-βcat binding, βcat is poly-ubiquitinated and can now be recognized by the proteasome. While down-regulation of βcat levels via protein degradation is a key function of the destruction complex, understanding of how βcat is transferred from the complex to the SCF E3 ligase is a key unanswered question (Schaefer, 2020).

Two classes of models seem plausible. In the first class of models, the E3 ligase is a physical entity separate from the destruction complex-this would fit with the many roles for the SCFSlimb E3 ligase, which binds and ubiquitinates diverse phospho-proteins, ranging from the Hedgehog effector Ci/Gli to the centrosome assembly regulator PLK4 . However, given the abundance of cellular phosphatases, this model has a potential major problem. Phosphorylated βcat released free from the destruction complex into the cytoplasm would likely be rapidly dephosphorylated, preventing its recognition by the E3 ligase. Consistent with this, earlier work revealed that APC helps prevent βcat dephosphorylation within the destruction complex. In a second class of models, the SCFSlimb E3 ligase might directly dock on or even become part of the destruction complex, either by direct interaction with destruction complex proteins or by using phosphorylated βcat as a bridge. In this model, once βcat is phosphorylated it could be directly transferred to the E3 ligase, thus preventing dephosphorylation of βcat by cellular phosphatases during transit. Immunoprecipitation (IP) experiments in animals and cell culture revealed that βTrCP can co-IP with Axin, APC, βcat, and GSK3, and that Wnt signals reduce Axin:βTrCP co-IP. However, these studies did not examine whether βTrCP or other components of the E3 are recruited to the destruction complex, leaving both models an option, especially if βTrCP acts as a shuttling protein between complexes (Schaefer, 2020).

A second set of outstanding questions concerns the mechanisms by which Wnt signaling down-regulates βcat destruction. Wnt signaling is initiated when Wnt ligands interact with complex multiprotein receptors, comprised of Frizzled family members plus LRP5/6. This receptor complex recruits the destruction complex to the plasma membrane via interaction of Axin with the phosphorylated LRP5/6 tail and with the Wnt effector Dishevelled (Dvl in mammals/Dsh in Drosophila). This leads to down-regulation of the destruction complex, reducing the rate of βcat destruction. Current data suggest destruction complex down-regulation occurs via multiple mechanisms, some rapid and others initiated more slowly. These include direct inhibition of GSK3 by the phosphorylated LRP5/6 tail, inhibition of Axin homo-polymerization by competition with hetero-polymerization with Dsh, competition between Dsh and APC2 for access to Axin, targeting Axin for proteolytic destruction, and blockade of βcat transfer to the E3 ligase. Recent work has explored the role of Dsh. Overall protein levels of Axin, APC2 and Dsh in Drosophila embryos experiencing active Wnt signaling are within a fewfold of one another, suggesting that competition is a plausible mechanism for destruction complex down-regulation (Schaefer, 2018). The competition model is also consistent with the effects of elevating Axin levels, which makes the destruction complex more resistant to turn-down. However, somewhat surprisingly, elevating Dsh levels had only modest consequences on cell fate choices, and Dsh only assembled into Axin puncta in cells receiving Wingless signals, suggesting that Dsh may need to be 'activated' by Wnt signals in order to effectively compete with APC for Axin and thus mediate destruction complex down-regulation. Candidate phosphorylation sites and kinases potentially involved in this activation have been identified. Intriguingly, when Axin, APC, and Dvl were expressed in mammalian cells, potential competition between APC and Dvl for interaction with Axin was revealed. This study examined in vivo the effects of simultaneously altering levels of Dsh and Axin, testing aspects of the competition model, and combined this with analysis of how Dsh and Axin affect one another's assembly into puncta in a simple cell culture model, using structured illumination superresolution microscopy (Schaefer, 2020).

Wnt signaling plays key roles in development and disease by regulating the stability of its effector βcat. In the absence of Wnt signals, βcat is phosphorylated by the Wnt-regulatory destruction complex, ubiquitinated by an SCF-class E3 ubiquitin ligase, and destroyed by the proteasome. Binding of Wnt ligands to their Frizzled/LRP receptors stabilizes βcat via the cytoplasmic effector Dsh. This study explored two important questions in the field: Is there a direct transfer of βcat from the destruction complex to the E3 ligase, and how does Dsh interaction with the destruction complex protein Axin regulate destruction complex function (Schaefer, 2020)?

Regulating the stability of βcat is the key step in Wnt signaling. The SCFSlimb E3 ligase was first identified as the relevant E3 regulating βcat levels in 1998. It specifically recognizes βcat after its sequential phosphorylation by CK1 and GSK3, and the most N-terminal phosphoserine is a key part of the binding site for the F-box protein Slimb/βTrCP. Phosphatase activity in the cytoplasm can rapidly dephosphorylate this residue, raising the question of how βcat is transferred to the E3 ligase without being dephosphorylated. Earlier work offered two clues. First, βTrCP can co-IP with Axin and APC, suggesting it may associate, at least transiently, with the destruction complex, providing a potential transfer mechanism. Consistent with this, stabilizing Axin using Tankyrase inhibitors led to colocalization of βTrCP and Tankyrase with the destruction complexes that assemble in response. However, it was not clear if this occurred by a direct interaction of βTrCP with destruction complex components via bridging by phosphorylated βcat or occurred because other components of the SCFSlimb E3 ligase were recruited more directly, with βTrCP recruited as a secondary consequence. A second clue emerged from analyses revealing that one role for APC is to prevent dephosphorylation of βcat while it is in the destruction complex, protecting the βTrCP binding site (Schaefer, 2020).

Two plausible models were suggested by these data. In the first, the entire SCFSlimb E3 ligase might be recruited to the destruction complex, allowing direct transfer of phosphorylated βcat between the two complexes. In a second model, βTrCP could serve as a shuttle, binding to phosphorylated βcat at the destruction complex and shuttling it to a place where the E3 assembled and ubiquitinated βcat (Schaefer, 2020).

This study explored interactions of the E3 ligase with the destruction complex using cell biological assays in SW480 cells. Ready recruitment was observed of the βTrCP homologue Slimb to destruction complex puncta by Axin, butrecruitment by APC2 was not observed, consistent with earlier assays by co-IP. Slimb recruitment did not require the βcat-binding site of Axin, making it less likely that recruitment occurs solely via bridging by βcat. However, it was enhanced by the RGS domain of Axin-future work to assess whether this involves a direct interaction or whether an indirect one is warranted. There are conserved residues in the RGS domain that are not necessary for the APC-Axin interaction, some of which form a pi helix, and it will be interesting to further explore the function of these residues. Both the region containing the N-terminus plus the F-box of Slimb and that including its WD40 repeats could be separately recruited into Axin puncta, suggesting it may be recruited by multiple interactions-in the case of the WD40 repeats, this could include bridging by phosphorylated βcat. Once again, direct binding assays in vitro would provide further insights, building on earlier assays suggesting a multipartite binding interaction. Superresolution imaging suggests the interaction between Slimb and Axin is intimate, consistent with direct binding. FRAP data, on the other hand, reveal that Slimb can come in and out of the complex, similar to the behavior of Axin and APC (Schaefer, 2020).

In contrast to the strong recruitment of Slimb to destruction complex puncta, two other core components of the SCFSlimb E3 ligase, Skp1 and Cul1, were not avidly recruited. The occasional recruitment seen could reflect interactions with endogenous βTrCP in the puncta. Coexpression of SkpA or Cul1 with Slimb slightly enhanced recruitment, but this was still not as robust as the recruitment of Slimb itself. IP/mass spectroscopy data and earlier work are consistent with the presence of all three core SCFSlimb E3 ligase proteins in the destruction complex, but suggest they may be present at lower levels than core destruction complex proteins. One possibility is that Slimb/βTrCP usually acts as a shuttle, but its presence occasionally recruits the other E3 proteins. Another possibility is that the entire SCFSlimb E3 ligase docks on the destruction complex transiently to accept phosphorylated βcat, ubiquitinate it, and then transfer it to the proteasome. Consistent with this possibility, inhibiting Tankyrase not only stimulates association of βTrCP with Axin but also leads to recruitment of the proteasome itself to the destruction complex-intriguingly, proteasome inhibition reduces destruction complex assembly, though this effect appears to be indirect due to effects on Axin2 levels. Further analyses will be needed to discriminate between these possibilities (Schaefer, 2020).

Additional work is also needed to explore how βcat transfer to the E3 ligase is regulated. Direct targeting of βcat to the E3, by fusing the F-box of Slimb with the βcat-binding sites of Tcf4 and E-cadherin, is sufficient to stimulate βcat destruction, independent of the destruction complex, but in vivo the destruction complex plays a critical role. Several pieces of data are consistent with the idea that transfer of βcat to the E3 ligase is the step regulated by Wnt signaling, rather than phosphorylation of βcat, with APC having an important role. Further exploration of this process will be welcome (Schaefer, 2020).

It has been clear for more than two decades that Dsh is a key effector of Wnt signaling. However, its precise mechanisms of action are complex and not fully understood. Current data suggest that Dsh is recruited to activated Frizzled receptors via its DEP domain. Dsh then helps ensure the Wnt-dependent phosphorylation of LRP5/6, leading to receptor clustering, facilitating Axin recruitment, and thus inhibiting GSK3. Dsh homo-polymerization, via its DIX domain, and hetero-polymerization with Axin, along with DEP-domain dependent Dsh cross-linking, are then thought to lead to down-regulation of the destruction complex and thus stabilization of βcat (Schaefer, 2020).

Intriguingly, in Drosophila embryos Dsh, Axin, and APC are present at levels within a few-fold of one another. Many current models suggest that relative ratios of these three proteins are critical to the signaling outcome, with APC and Dsh competing to activate or inhibit Axin, respectively. Consistent with this, substantially elevating Axin levels in vivo, using Drosophila embryos as a model, renders the destruction complex immune to down-regulation by Wnt signaling. Subsequent work revealed that the precise levels of Axin are critical-elevating Axin levels by two- to fourfold has little effect, while elevation by ninefold is sufficient to constitutively inactivate Wnt signaling. One might then predict that elevating Dsh levels would have the opposite effect, sequestering Axin and thus stabilizing βcat and activating Wnt signaling. While very high levels of Dsh overexpression can have this effect, it was previously surprising to learn that sevenfold elevation of Dsh levels only had a subtle effect on Wnt signaling and thus had little effect on embryonic viability (Schaefer, 2018). The data further suggested that Dsh is only recruited into Axin puncta in cells that received Wg signal, in which puncta are recruited to the plasma membrane, even though seemingly similar levels of Dsh were present in Wnt-OFF cells (Schaefer, 2018). This opened the possibility that a Wnt-stimulated activation event, such as Dsh phosphorylation, might be required to facilitate Dsh interaction with Axin and thus Axin inactivation. In this scenario, elevating Dsh levels in cells without this activation event, for example, in Wnt-OFF cells, would not alter signaling output (Schaefer, 2020).

The simplest versions of the antagonism model, involving competition between formation of Axin/APC versus Axin/Dsh complexes, would also suggest that elevating Dsh levels should alleviate effects of elevating Axin. This was tested directly, expressing high levels of Dsh maternally and lower levels of Axin zygotically. It was anticipated that elevating Dsh levels would blunt the effects of elevating levels of Axin. Instead, a substantial surprise was in store: elevating levels of Dsh enhanced the ability of Axin to resist turndown by Wnt signaling, thus leading to global activation of the destruction complex and inactivation of Wnt signaling. This was true whether effects were assessed on cell fate choice, Arm levels, or expression of a Wnt-target gene. Intriguingly, the data were also consistent with the possibility that elevating Dsh levels may alter Axin:APC interactions in Wnt-ON cells-this might provide a clue to an underlying mechanism (Schaefer, 2020).

What could explain this paradoxical result? The current data do not provide a definitive answer but do open some intriguing possibilities and new questions. In the view of the authors, part of the explanation will be that Wg-dependent 'activation' of Dsh is required for it to interact with and thus down-regulate Axin. Consistent with this, Dsh phosphorylation can regulate its ability to homopolymerize. By elevating Dsh levels, the capacity of this activation system might have been exceeded. High levels of 'nonactivated' Dsh, while unable to interact with Axin, might still interact with other key proteins involved in destruction complex down-regulation, sequestering them in nonproductive complexes. For example, Dsh can bind CK1, which has complex roles in Wnt regulation With key proteins sequestered, the system might become less able to inactivate the slightly elevated levels of Axin present, thus leading to constitutive activity of the destruction complex. In this speculative scenario, it is not the relative levels of Axin and Dsh that are key but the relative levels of Axin and 'active Dsh' (Schaefer, 2020).

The results of our SIM experiments may also provide insights. The ability of Axin and Dsh to both homo- and hetero-polymerize means free monomers must make a choice. It is likely this is a regulated choice, though the mechanism of regulation remains unclear. The experiments with SW480 cells, while overly simple, may provide an illustration of how the homo-/hetero-polymerization balance can shift. In cells in which both Axin and Dsh were expressed at relatively low levels, puncta contained both proteins, and internal structure was consistent with some level of hetero-polymerization. In contrast, when levels of Dsh were significantly higher, Axin and Dsh tended to segregate into separate, adjoining puncta, suggesting the balance was shifted to homo-polymerization, though the polymers retained the ability to dock on one another. If similar events occur on elevating Dsh expression in Drosophila embryos, segregation could allow Axin to remain in functional destruction complexes, even in Wnt-ON cells, while Dsh localized to separate puncta sequestered other Wnt-regulating proteins, potentially explaining how elevating Dsh expression could paradoxically down-regulate Wnt signaling. Elevating Dsh levels may also lead it to preferentially associate with itself, as was suggested by the SIM data in SW480 cells-this could recruit endogenous Dsh away from its normal localization with the destruction complex, thus preventing it from participating in inactivating Axin. Defining the mechanisms that determine the relevant affinities of each protein for itself versus for its partner will be informative. Intriguingly, a similar docking was observed rather than coassembly behavior when the puncta formed by Axin and those formed by the Arm repeat domain of APC2 were imaged-this may be another example where relative affinities of proteins for themselves versus their binding partners differ (Schaefer, 2020).

Very recent work provides important new insights in this regard. Yamanishi (2019) determined the structure of the heterodimer of the DIX domains of Dsh and Axin and also measured their relative affinities for one another. Another study used cryo-electron microscopy to solve the structure of Dsh filaments and also measured affinities of DIX domains of Dsh and Axin. The results of the second group contrast, with the first group suggesting Dsh homodimerization is an order of magnitude more favorable than Axin homodimerization, while heterodimerization is intermediate in affinity, and the other suggesting Axin homodimerization is most favorable. Resolution of this will be important, as how Dsh acts to turn down destruction complex activity by heterodimerization is assessed. It also is interesting given in vivo observations that APC may help stabilize Axin homo-polymerization . These data also may help explain the results in SIM, where segregation of Dsh and Axin is favored in some circumstances. Defining the in vivo regulatory mechanisms that modulate homo- and heteropolymerization will be an important goal. Together the results leave more questions than answers but suggest that there are important features of Wnt signaling in vivo yet to be uncovered. Further cell biological and biochemical experiments in vivo, combined with new mathematical models of the suspected competition, will be extremely useful (Schaefer, 2020).

Endosomal Wnt signaling proteins control microtubule nucleation in dendrites

Dendrite microtubules are polarized with minus-end-out orientation in Drosophila neurons. Nucleation sites concentrate at dendrite branch points, but how they localize is not known. Using Drosophila, this study found that canonical Wnt signaling proteins regulate localization of the core nucleation protein gammaTubulin (gammaTub). Reduction of frizzleds (fz), arrow (low-density lipoprotein receptor-related protein [LRP] 5/6), dishevelled (dsh), casein kinase Igamma, G proteins, and Axin reduced gammaTub-green fluorescent protein (GFP) at branch points, and two functional readouts of dendritic nucleation confirmed a role for Wnt signaling proteins. Both dsh and Axin localized to branch points, with dsh upstream of Axin. Moreover, tethering Axin to mitochondria was sufficient to recruit ectopic gammaTub-GFP and increase microtubule dynamics in dendrites. At dendrite branch points, Axin and Dsh colocalized with early endosomal marker Rab5, and new microtubule growth initiated at puncta marked with Fz, Dsh, Axin, and Rab5. It is proposed that in dendrites, canonical Wnt signaling proteins are housed on early endosomes and recruit nucleation sites to branch points (Weiner, 2020).

Neurons extend long branched processes from a central cell body. This shape is incompatible with a centrosomal microtubule organizing center (MTOC). Mature neurons are therefore among the ranks of differentiated cells that have noncentrosomal microtubule arrays. It is particularly important to understand how neuronal microtubules are organized because the distance from the primary site of synthesis in the cell body to functional sites in axons and dendrites can be large and, therefore, place heavy demands on microtubule-based transport. In humans, slight disruptions in microtubule regulators or motors can manifest as neurodegenerative disease, underscoring neuronal reliance on perfectly orchestrated microtubule-based transport (Weiner, 2020).

If neuronal microtubules are not anchored to the centrosome, how are they organized? In all neurons so far examined, axonal microtubules have their dynamic plus ends oriented away from the cell body (plus-end-out). In dendrites of vertebrate neurons, microtubules are mixed polarity. In invertebrate neurons (Drosophila and Caenorhabditis elegans), axons have the same plus-end-out microtubule organization as vertebrates, but mature dendrites have almost all minus-end-out microtubules. In immature Drosophila dendrites, microtubules are mixed polarity and only gradually resolve to the minus-end-out mature arrangement. Thus, although the final arrangement of microtubules in vertebrate and invertebrate dendrites is somewhat different, they are the same during dendrite development (Weiner, 2020).

Although the arrangement of neuronal microtubules is clearly noncentrosomal, the source of axonal and dendritic microtubules has been controversial. Two major models for generating axonal and dendritic microtubules have been proposed. The first is that neuronal microtubules are nucleated at the centrosome, or perhaps elsewhere in the cell body, and then released for transport/sliding into axons and dendrites. This model has substantial support, including recent analyses with newer techniques. For example, live imaging of microtubules with plus-tip (+TIP) tracking proteins and photoconvertible αTubulin has provided evidence for directional transport of microtubules into and out of developing axons in mammalian and Drosophila neurons (Weiner, 2020).

The second model is that nucleation sites are found outside the cell body and that microtubules are generated locally in axons and dendrites. Evidence for this model came from the observation that centrosomal γTubulin (γTub), the core microtubule nucleation protein, decreases gradually over time and that centrosome ablation does not disrupt axon formation. Similarly, the centriole is not surrounded by γTub in Drosophila neurons in vivo, and it is dispensable for neuronal microtubule organization. One way to reconcile these two models is to assume that both are important and that, very early in neuronal development, microtubule sliding can dominate, whereas later in development and in mature neurons, microtubules are primarily locally nucleated (Weiner, 2020).

In some cell types, the Golgi complex recruits nucleation sites, and small Golgi outposts can be found in both mammalian and Drosophila dendrites. Thus, it was proposed that the Golgi might act as a noncentrosomal MTOC in dendrites. However, subsequent analysis of γTub and Golgi outposts, including a strategy to deplete Golgi from dendrites, called this proposal into question (Weiner, 2020).

Within Drosophila dendrites, γTub is concentrated at branch points. A previous study identified proteins that localize a different microtubule regulator, adenomatous polyposis coli (Apc) 2, to branch points. It was reasoned that some or all of this machinery might be used to position γTub to the same region. Whether any of the Apc2 localization proteins act upstream of γTub-green fluorescent protein (GFP) in dendrites was tested. Surprisingly, a subset of Wnt signaling proteins was required to localize γTub-GFP to dendrite branch points, regulate dendritic microtubule polarity, and nucleate microtubules in dendrites in response to axon injury. The required proteins include the seven transmembrane domain Frizzled (Fz) proteins (Wnt receptors), Arrow (Arr, a Wnt coreceptor), heterotrimeric G proteins, Dishevelled (Dsh), Casein kinase I (CK1)γ, and Axin. Axin seems to be the key output protein of this pathway because it was sufficient to recruit γTub to ectopic sites in dendrites. Within branch points, Fz, Axin, and Dsh were found on puncta that colocalized with Rab5. In addition, new end-binding protein 1 (EB1) comets at polymerizing microtubule plus ends initiated from puncta marked with Fz, Arr, Dsh, Axin, and Rab5. It is proposed that Wnt signaling proteins localize to early endosomes at dendrite branch points and function there to control local microtubule nucleation. Although it has previously been shown that Wnt signaling proteins can function from endosomes, identification of microtubule nucleation as an output of endosomal Wnt proteins is quite unexpected (Weiner, 2020).

It was particularly intriguing to find integral membrane signaling proteins required for noncentrosomal microtubule nucleation. Although Wnt signaling has been linked to microtubule plus-end regulation in axon growth cones and regulation of microtubule stability and spindle orientation, the only connection to the minus end is localization of some cytoplasmic Wnt signaling proteins like Axin to the centrosome in dividing cells. This study demonstrated that a Wnt signaling pathway acts upstream of microtubule nucleation in a postmitotic cell. Not only were many canonical Wnt signaling proteins required for γTub-GFP to accumulate at branch points, but Axin and Dsh themselves concentrated at branch points. In addition, the scaffolding protein Axin was able to recruit γTub-GFP and the nucleation activator Cnn to mitochondria when tethered to them. Moreover, reduction of Wnt signaling proteins phenocopied loss of γTub in two functional nucleation assays, indicating that most or all dendritic nucleation occurs downstream of this pathway. Although this pathway seems to be the major regulator of dendritic nucleation, neurons are quite resilient to its loss under baseline conditions, and the simple ddaE neurons have normal arbor shape. This is likely because parallel pathways can be used to generate new minus ends. For example, microtubule severing can be used to generate new plus and minus ends and amplify microtubule number. In many cell types, minus ends generated when a microtubule is severed are recognized by minus-end binding proteins in the calmodulin-regulated spectrin-associated protein (CAMSAP)/Patronin family. In C. elegans, γTub-mediated microtubule nucleation has been shown to act in parallel and quite redundantly with Patronin to regulate microtubule organization. Recent work has shown that Patronin-mediated minus-end growth is an important regulator of dendritic microtubules in Drosophila, so it is possible that microtubule severing in conjunction with Patronin recruitment to minus ends can compensate for nucleation under most normal circumstances. Consistent with this hypothesis, phenotypes from reduction of nucleation or Patronin become more evident after severe stress, including axon [29] or dendrite [77, 98] injury (Weiner, 2020).

Although it was consistently found that partial loss of function (RNAi or heterozygous mutants) for fz, fz2, arr, dsh, Gao, Gas, and Axin reduced γTub localization and/or function, no evidence was found that β-catenin/Arm, the key transcription factor that is the output of canonical Wnt signaling, was involved. In addition, an Arm protein trap showed clear expression in epidermal cells but was not seen in dendritic arborization neurons (da neurons). Because Axin itself was sufficient to recruit γTub, there was no strong rationale for a transcriptional regulator to mediate signaling between fz/arr and microtubule nucleation. It is proposed that canonical Wnt signaling proteins are co-opted in dendrites to directly recruit nucleation complexes to endosomes. Because this is a variant of canonical Wnt signaling that unexpectedly seems not to involve β-catenin, this pathway is termed apocryphal Wnt signaling in reference to the Apocrypha, ancient writings found in only some versions of the Bible (Weiner, 2020).

The involvement of arr as well as dsh and Axin suggests that a signalosome might be involved in dendritic Wnt signaling. Signalosomes form when wnt ligands bind to Fz and LRP5/6 at the plasma membrane, triggering recruitment and multimerization of Dsh and Axin. The normal output of signalosome formation is release of β-catenin from the destruction complex and its subsequent stabilization and transit to the nucleus to activate transcription. Signalosomes assemble at the plasma membrane. Endocytosis generally seems to promote Wnt signaling, although in many contexts the signalosome itself is disassembled upon endocytosis. It is not clear whether signalosomes persist after endocytosis, though in some Drosophila cells, Dsh and Arr are localized to endosomes. In dendrites, puncta of Fz, Dsh, and Axin colocalized with Rab5, suggesting that a stable signaling complex is present on endosomes in mature neurons. The initiation of comets from these puncta indicates that endosomes are likely the key site where Wnt signaling proteins promote nucleation. Colocalization of tagged Golgi proteins with Rab5 suggests that the previous association between Golgi markers and nucleation could have been due to leakage into endosomes. In addition, the identification of plasma membrane proteins acting upstream of γTub in dendrites suggests a more general role for the Golgi in the cell body by controlling secretion of Arr and Fz (Weiner, 2020).

Wnt signaling receptors have been classically studied at the plasma membrane, where they bind extracellular ligands that can be autocrine or paracrine in nature. A requirement for arr and fz upstream of γTub in dendrites suggests that a Wnt ligand is likely involved. Failure of neuronal wntless knockdown to reduce γTub-GFP at branch points favors the hypothesis that the ligand may be secreted from a neighboring cell. In the embryo, wingless (wg)/Wnt-1 is made in a patch of epithelial cells adjacent to developing dendritic arborization neurons and helps pattern dendrite orientation in ddaE . It would be very interesting if surrounding cells influenced the microtubule cytoskeleton in mature neurons through Fz and Arr at the plasma membrane. This signaling pathway is particularly intriguing in the context of regeneration or during neurodegenerative disease. During axon regeneration, the initial injury response involves a nucleation-dependent increase in microtubule dynamics, which serves a neuroprotective role. Modulating Wnt signaling could therefore influence neuroprotection in dendrites. In addition, this study found that this pathway is required during dendrite regeneration to position nucleation sites in regrowing dendrites. Interestingly, G protein coupled receptors (GPCRs) represent 33% of all Food and Drug Administration-approved drug targets, and as part of this family, Fz presents a possible target (Weiner, 2020).

Local microtubule nucleation also occurs in axons. As Rab5 endosomes are present throughout axons, it will be interesting to determine whether Wnt signaling proteins can be recruited to axonal early endosomes and whether they recruit nucleation proteins in this part of the cell. It is also possible that a link between Wnt signaling, endosomes, and nucleation could exist more broadly in other cell types. Indeed, the localization of Axin to centrosomes suggests that even in mitotic cells, parts of this relationship are conserved. Intriguingly, endosomal membranes are concentrated around the centrosome, and Rab5 reduction disrupts mitosis, so it is possible that Wnt signaling proteins, endosomes, and nucleation function together at centrosomes (Weiner, 2020).

GENE STRUCTURE

Sequence analysis has identified two Axn species, termed A1 and A2. These forms of Axn are revealed by amino acid differences between two independent Axin isolates: Willert (1999b) and Hamada (1999). Form A2 (Willert, 1999b) includes the six amino acids SRSGSS while Form A1 of Hamada, 1999, does not.

PROTEIN STRUCTURE

Amino Acids - 739

Structural Domains

The Axin cDNA contains two alternative polyadenylation sites and an open reading frame encoding a 739 amino acid protein (form A1). Several clones (3 out of 5) contain an 18 basepair (bp) insert (form A2). This insert in Drosophila is in a similar, albeit not identical, region where in mouse, Axin contains a 108 bp insert to give rise to two different forms (form 1 and 2). The 18 nucleotides in the fly gene code for six amino acids, four of which are serine residues, directly amino-terminal to the DIX domain. No difference in the activities of the two isoforms of either mouse Axin or Drosophila Axn have been identified. The recently published sequence of Drosophila Axin (Hamada, 1999) is identical to form A2, but lacks two serine residues at position 644 and 645. The differences in sequences, including forms A1 and A2, may reflect naturally occurring polymorphisms in the Axn gene. Overall, the identity between mouse and Axn is 21% , although several domains are more highly conserved. The Axn protein is more distantly related (17%) to Conductin (Behrens, 1998), suggesting that it is the vertebrate Axin ortholog. Like the vertebrate Axin proteins, Axn contains an RGS domain near the amino terminus. These domains are found in a family of proteins that regulate G-proteins. Although no such G-protein signaling function has been attributed to the mouse Axin RGS domain (Mao, 1998), the RGS has been shown to be required for the interaction with the APC protein (Behren, 1998). Deletion of this domain in mouse Axin produces a dominant negative Axin protein, as assayed by axis duplication activity in Xenopus embryos (Itoh, 1998; Zeng, 1997). In contrast, overexpression of wild-type mouse Axin inhibits axis duplication, suggesting that interaction with APC is crucial for Axin’s ability to block the Wnt signaling pathway. The region between the RGS and DIX domains, where both GSK-3beta and beta-catenin binding sites are located in Axin and Conductin, contains stretches of lower similarity. Despite this low similarity, Axn does in fact interact with both Arm and Zw3. Hamada (1999) mapped the Arm binding domain in Axn to amino acids 459-538 by precipitating in vitro translated Arm with GST-Axn fusion proteins, a region which contains significant homology with Axin and Conductin. The Axn protein contains four potential nuclear localization signals, but the mouse Axin protein, that also contains a nuclear localization signal, has been localized predominantly to the cytoplasm (Behrens, 1998 and Fagotto, 1999) (Willert, 1999b).

date revised: 6 October 99

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.